LLM WIKI

📗 On Building a Vocabulary for Discussing Network Effects

Wed, 08 Apr 2026 00:00:00 +0000

A Review of The Cold Start Problem by Andrew Chen

I read The Cold Start Problem for the same reason I read and reviewed Certain to Win — I wanted to understand business better now that I work for a startup instead of a school system. Happily, The Cold Start Problem turned out to be one of the more useful books I read last year, and it comes up in discussion with friends regularly in all sorts of contexts. Sometimes we’re talking about corporate recruiting, sometimes it’s about forming friend groups for kindergarteners, sometimes we’re pondering the nature of our book club… but The Cold Start Problem applies to all of it.

@a16z on X - AI Makes You the CEO

Tue, 07 Apr 2026 00:00:00 +0000

a16z @a16z 2026-04-07

“AI doesn’t take your job. AI makes you the CEO.”

Balaji Srinivasan joins a16z’s Erik Torenberg for a conversation on the future of the AI economy, decentralization, and how work changes in an AI-native world, including:

- How distillation and open source could decentralize AI power

- Why AI lowers the cost of creation but raises the cost of verification

- The shift from global internet to “trusted tribes” and private AI

@atmoio on X - Claude Mythos is Delusional

Tue, 07 Apr 2026 00:00:00 +0000

Mo @atmoio 2026-04-07

Claude Mythos is Delusional

2026-04-07

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.

It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.

https://anthropic.com/glasswing

WVY @starpower_tech 2026-04-08

This the realest nigga to ever say anything about ai

Mo @atmoio 2026-04-08

😄🙏

Lorenzo Nuvoletta @nuvolore 2026-04-08

Man you are hilarious and spot on, love your videos keep them coming 🤣

@DarioAmodei on X - Project Glasswing

Tue, 07 Apr 2026 00:00:00 +0000

Dario Amodei @DarioAmodei 2026-04-07

I’m proud that so many of the world’s leading companies have joined us for Project Glasswing to confront the cyber threat posed by increasingly capable AI systems head-on.

2026-04-07

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.

It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.

@NinaDSchick on X - Claude Mythos

Tue, 07 Apr 2026 00:00:00 +0000

Nina Schick @NinaDSchick 2026-04-07

Claude Mythos.

Ten trillion parameters: the first model in this weight class. Estimated training cost: ten billion dollars.

On the hardest coding test in the industry (SWE bench) it scores 94%.

It found a security flaw in a system that had been running for 27 years, one that every human engineer and every automated check had missed. It found another bug that had survived five million test runs over 16 years. (It did so overnight.)

Substack

Tue, 07 Apr 2026 00:00:00 +0000

On creating work like a human (and all the feelings that come with it)

@nonzeroexitcode on X - iOS Safari Input Auto-Zoom Fix

Mon, 06 Apr 2026 00:00:00 +0000

daniel petho @nonzeroexitcode 2026-04-06

a well-known iOS Safari quirk is that inputs with a font-size <16px trigger an auto-zoom on focus

using 16px is probably the best solution, but it can feel a bit heavy in tight UIs

a clever workaround is to keep the font-size at 16px and scale the input down with CSS!

daniel petho @nonzeroexitcode 2026-04-06

ran into this again while working on forms and found this super clever solution/article by Jeffrey To:

@tobi on X - QMD 2.1 Release

Mon, 06 Apr 2026 00:00:00 +0000

tobi lutke @tobi 2026-04-05

qmd 2.1 is out with all sorts of goodies. Code aware splitting, performance, official benchmarks etc etc

tobi lutke @tobi 2026-04-05

Max Slinger @PromptSlinger 2026-04-06

shipping benchmarks with the release instead of ‘its fast trust me’ energy. respect

yash @yashns1 2026-04-06

Code-aware splitting is a massive win for technical RAG. Naive chunking usually breaks function context or class logic, which makes the retrieval useless for agents. Good to see this, looking forward to testing it out myself.

@aakashgupta on X - Agent Proficiency

Sun, 05 Apr 2026 00:00:00 +0000

Aakash Gupta @aakashgupta 2026-04-04

The guy who literally wrote the most popular deep learning course in history just told you the most valuable skill of the next decade has nothing to do with code.

Karpathy’s framing here is precise: “agent proficiency is a CORE SKILL of the 21st century.” He’s not saying learn to prompt. He’s saying learn to manage autonomous systems that do work for you.

Think about what he built his entire career on. Teaching people to understand neural nets at the weight level. Backpropagation by hand. Tokenizers from scratch. He was the “understand every layer” guy.

@aakashgupta on X - Obsidian - Three Engineers, Five Million Users

Sat, 04 Apr 2026 00:00:00 +0000

Aakash Gupta @aakashgupta 2026-04-04

Three engineers built a product that makes every VC-backed productivity company look overstaffed by a factor of 10.

Obsidian has 5 million downloads. Over 1.5 million active users. 2,000+ community plugins. Runs on every major OS. End-to-end encrypted sync. A publishing platform. All shipping regular updates.

The team that built it: three engineers. Not three engineering pods. Not three squads. Three human beings writing code.

Notion has 1,200 employees and $400 million in annual revenue to serve 100 million users. That’s one employee per 83,000 users. Obsidian serves 1.5 million active users with roughly seven total staff. That’s one employee per 214,000 users. And four of those seven aren’t engineers.

@Simon_Ingari on X - The Corporate Jackpot

Sat, 04 Apr 2026 00:00:00 +0000

Simons @Simon_Ingari 2026-04-04

In the corporate world, if you have:

- a decent salary,

- a manager who trusts you without micromanaging,

- the flexibility of a hybrid work setup,

- the freedom to take time off when needed,

- opportunities for growth and skill development,

- a supportive and inclusive company culture,

You’re already among the top 1% of professionals enjoying a truly balanced and fulfilling work life.

Danyal | SMM Expert @danyal_wani 2026-04-04

@googlegemma on X - Gemma 4 Launch

Fri, 03 Apr 2026 00:00:00 +0000

Google Gemma @googlegemma 2026-04-03

Thanks for following us!

We’re excited to see what you all build with Gemma 4!

In case you missed it, you can find all our checkpoints, with an Apache 2.0 License, on Hugging Face:

Google Gemma @googlegemma 2026-04-03

Suvodeep | Developer @suvodeepmishra1 2026-04-03

But I do not have a GPU and for that I cannot build anything using gemma even though I cannot use gemma in my appcapp project it’s very laggy

@kepano on X - Free Growth Strategy

Fri, 03 Apr 2026 00:00:00 +0000

kepano @kepano 2026-04-03

free growth strategy:

1. keep improving little by little

2. stay 100% user-supported

3. watch VC-backed companies gradually destroy their product and alienate their users

Zag Zino @ZagZino 2026-04-03

user-supported only really works if you solve distribution, otherwise you’re the best product nobody’s heard of tbh

kepano @kepano 2026-04-03

you can only stay user-supported if you start user-supported

and also

2026-03-13

I would be working on Obsidian even if it was only useful to me

SpaceX IPO, Iran War Fallout, Quantum Bitcoin Hack, The Space Opportunity

Fri, 03 Apr 2026 00:00:00 +0000

(0:00) Bestie intros!
(0:12) SpaceX IPO, the economic opportunity of space: a new industrial frontier
(21:00) 2026 IPO explosion, OpenAI down round?
(36:33) Iran War costs, fertilizer crisis, downstream impacts
(49:58) Trump’s Iran messaging problems, Bondi out, why the US is in Iran
(1:04:18) Quantum Bitcoin hack possibilities, how crypto should react

Follow the besties:
https://x.com/chamath
https://x.com/Jason
https://x.com/DavidSacks
https://x.com/friedberg

Follow on X:
https://x.com/theallinpod

Follow on Instagram:
https://www.instagram.com/theallinpod

Follow on TikTok:
https://www.tiktok.com/@theallinpod

We need to talk about the Claude Code rate limits

Fri, 03 Apr 2026 00:00:00 +0000

Anthropic just made the limits on the Claude Max/Pro plans a lot worse…

Thank you Depot for sponsoring! Check them out at: https://soydev.link/depot

SOURCES
https://x.com/trq212/status/2032916661452595648
https://x.com/trq212/status/2037254607001559305
https://x.com/Pranit/status/2037353721047491047

Want to sponsor a video? Learn more here: https://soydev.link/sponsor-me

Check out my Twitch, Twitter, Discord more at https://t3.gg

S/O @Ph4seon3 for the awesome edit 🙏

Transcript

0:00 · Anthropic just made a huge change to how rate limits work for Claude Code subscribers and people are pissed. It’s spring break for Cloud Code. Enjoy 2x usage during off- peak app. Wait, what?

@itsolelehmann on X - LLM Knowledge Bases

Thu, 02 Apr 2026 00:00:00 +0000

Ole Lehmann @itsolelehmann 2026-04-02

karpathy just casually described the future of ai and most people scrolled right past it:

he’s been building what he calls “llm knowledge bases.”

here’s what that means in plain english:

you take everything you’re interested in.

articles, research papers, datasets, images, etc and you dump it all into one folder

then you point your ai at the folder and say “read all of this, organize it, and remember it”

@karpathy on X - LLM Knowledge Bases

Thu, 02 Apr 2026 00:00:00 +0000

Andrej Karpathy @karpathy 2026-04-02

LLM Knowledge Bases

Something I’m finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:

I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally “compile” a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

@tobi on X - What Is Intelligence - Book Recommendation

Thu, 02 Apr 2026 00:00:00 +0000

tobi lutke @tobi 2026-04-02

Brilliant book on intelligence. Highly recommended.

https://mitpress.mit.edu/9780262049955/what-is-intelligence/…

And as always with these type of books, skip the last chapter.

Emon Datta @emonuxui 2026-04-02

Books that unpack intelligence conceptually are valuable because they challenge assumptions and frame complex phenomena in measurable ways. This approach mirrors the recommendation here , focus on frameworks before applied predictions.

A2H Market @A2HMarket 2026-04-02

Adding to the list.

The question that’s keeping us busy right now isn’t “what is intelligence” — it’s “what is trustworthiness”.

Who sets your agenda?

Thu, 02 Apr 2026 00:00:00 +0000

It’s a question so rarely asked it almost feels silly to ask it.

Some situations and some jobs work to eliminate our freedom of choice. Prison, medical school, 8th grade–there are settings where time, tools, and options are severely limited.

But even in these settings, we have more choice than we realize.

And for the rest of us, particularly freelancers and entrepreneurs, our agenda is wide open.

Who decides what you will eat tonight, or what you will do after dinner? Who decides who you will call on, what you will learn next, which posts you’ll read (or write)? Who decides what tone the conversation will have, what your priorities are, and what you’ll worry about when you walk the dog?

@ttorres on X - Is Product Management Dead

Tue, 31 Mar 2026 00:00:00 +0000

Teresa Torres @ttorres 2026-03-31

🎙️Product Builders

Is product management dead?

Petra Wille and Teresa Torres unpack the big question making the rounds at conferences: Is PM disappearing—or evolving?

Their take: product management isn’t dead. But the traditional product trio (PM, design, engineering) is collapsing into something new.

AI is raising the baseline. More of the repeatable 80% can now be built by anyone with the right tools. The future belongs to “product builders” — people with a shared foundation across disciplines and deep expertise in one area.

Product Builders - All Things Product with Teresa & Petra

Tue, 31 Mar 2026 00:00:00 +0000

Is product management dead?

Petra Wille and Teresa Torres unpack the big question making the rounds at conferences: Is PM disappearing—or evolving?

Their take: product management isn’t dead. But the traditional product trio (PM, design, engineering) is collapsing into something new.

Actually, MacBook Neo Is For Me

Wed, 18 Mar 2026 00:00:00 +0000

The MacBook Neo should not be a device for me. I have a Mac in my office with an M3 Max chip. And I have an M4 MacBook Air maxed out with 32GB of RAM. You know, just in case. When the new Macs hit a couple weeks back, I assumed I would be tempted by one of the M5 variants. But just like any good consumer, I’m compelled by the idea of “new”. Despite its iPhone-class chip and meager 8GB of RAM, I had to try out the MacBook Neo.

Agents Over Bubbles

Mon, 16 Mar 2026 00:00:00 +0000

Listen to this post:

There is a weird paradox in terms of AI prognostication: on one hand, you don’t want to be the one to completely dismiss the most terrifying doomsday scenarios; who wants to be found out to be foolishly optimistic? At the same time, there is also pressure to give credence to the possibility that we are in a bubble, and all of this hype and spending is going to go belly up.

Product Leadership Coach

Mon, 16 Feb 2026 00:00:00 +0000

Scroll

HOW TO DEVELOP A

STRONG PRODUCT CULTURE

I’m a freelance Product Leadership Coach who’s been helping companies to improve their product culture and product teams to up their game since 2013.

learn more about what I do

Curious about how great product leaders grow? I’ve just launched a 52-week email course —designed to help you move from strong to exceptional. 👉 Learn more about the course here

OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491

Thu, 12 Feb 2026 00:00:00 +0000

Peter Steinberger is the creator of OpenClaw, an open-source AI agent framework that’s the fastest-growing project in GitHub history.
Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/sponsors/ep491-sb
See below for timestamps, transcript, and to give feedback, submit questions, contact Lex, etc.

*Transcript:*
https://lexfridman.com/peter-steinberger-transcript

*CONTACT LEX:*
*Feedback* - give feedback to Lex: https://lexfridman.com/survey
*AMA* - submit questions, videos or call-in: https://lexfridman.com/ama
*Hiring* - join our team: https://lexfridman.com/hiring
*Other* - other ways to get in touch: https://lexfridman.com/contact

@tobi on X - DHH Interview

Tue, 20 Jan 2026 00:00:00 +0000

tobi lutke @tobi 2026-01-20

Yea that was a fun one.

Thank you for celebrating the glory of entrepreneurship, David.

2026-01-18

My conversation with Tobi Lütke (@tobi), co-founder and CEO of Shopify.

0:00 Companies as Social Technology

5:27 The Value of Reading Books: Cheat Codes for Life

7:28 Post-IPO Crisis: Cosplaying as a CEO

7:54 Competition vs Rivalry: The Power of Healthy Competition

16:02 COVID

@tobi on X - QMD - My Finest Tool

Mon, 19 Jan 2026 00:00:00 +0000

tobi lutke @tobi 2026-01-19

I think QMD is one of my finest tools. I use it every day because it’s the foundation of all the other tools I build for myself. A local search engine that lives and executes entirely on your computer.

https://github.com/tobi/qmd

Both for you and agents

2026-01-19

Okay, I finally got up to speed on QMD & how it all works. First of all, WOW. Holy shit man, what a insanely clever & genius orchestration of tools. My mind is THOROUGHLY blown. I LOVE the query expansion implementation, genuinely a super clever idea (plus I love the x2 weight

DHH: Future of Programming, AI, Ruby on Rails, Productivity & Parenting | Lex Fridman Podcast #474

Sat, 12 Jul 2025 00:00:00 +0000

Transcript

Episode highlight

0:00 · No one anywhere who’s serious believes that Cookie banners does anything good for anyone. Yet, we’ve been unable to get rid of it. This is the thing that really gets me about cookie banners, too. It’s not just the EU. It’s the entire world. You can’t hide from cookie banners anywhere on this planet. If you go to goddamn Mars on one of Elon’s rockets and you try to access a web page, you’ll still see a cookie banner.

State of the product job market in 2025

Tue, 13 May 2025 00:00:00 +0000

There’s a lot to be optimistic about

👋 Welcome to a 🔒 subscriber-only edition 🔒 of my weekly newsletter. Each week I tackle reader questions about building product, driving growth, and accelerating your career. For more: Lenny’s Podcast | How I AI | Lennybot | Lenny’s Reads | Courses | Swag

Annual subscribers now get a free year of Perplexity Pro, Notion, Superhuman, Linear, Granola, and more. Subscribe now.

@kepano on X - Obsidian Web Clipper YouTube Transcript Extraction

Tue, 10 Dec 2024 00:00:00 +0000

kepano @kepano 2024-12-10

here’s an example of using Obsidian Web Clipper to extract the transcript from a YouTube video then run it through Interpreter to summarize the key points

you can do the same with your highlights, or any page content you want — and run templates based on URL patterns

kepano @kepano 2024-12-10

Here’s the interpreter context I used:

```

{{selectorHtml:ytd-engagement-panel-section-list-renderer[visibility$=\“EXPANDED\”] #segments-container|strip_tags:“h2,ytd-transcript-segment-renderer”|replace:“ytd-transcript-segment-renderer”:“li”|markdown}}

```

IMRΛN @imrane 2024-12-10

Can we set the interpreter prompt by template type?

Pieter Levels: Programming, Viral AI Startups, and Digital Nomad Life | Lex Fridman Podcast #440

Tue, 20 Aug 2024 00:00:00 +0000

Pieter Levels (aka levelsio on X) is a self-taught developer and entrepreneur who has designed, programmed, launched over 40 startups, many of which are highly successful.
Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/sponsors/ep440-sb
See below for timestamps, transcript, and to give feedback, submit questions, contact Lex, etc.

*Transcript:*
https://lexfridman.com/pieter-levels-transcript

Product management theater | Marty Cagan (Silicon Valley Product Group)

Sun, 10 Mar 2024 00:00:00 +0000

Marty Cagan is a luminary in the world of product. He’s the author of two of the most foundational books for product teams and product leaders (Inspired and Empowered), he’s the founder of Silicon Valley Product Group (one of the longest-running product advisory groups), and he’s almost certainly worked with more product leaders and teams than any human alive. Now he’s releasing his newest book, Transformed, which is sure to become a staple of tech-powered companies worldwide. Marty’s previous appearance on our show remains one of the most popular episodes to date. In this conversation, we discuss:
• The rise of “product management theater”
• Changes in the PM role post-ZIRP and the shift from growth to build functions
• The disconnect between good product companies and online product advice
• How over-hiring has created challenges in the product industry
• The most important skills for PMs to build
• How to know if you’re on a “feature team”
• The potential disruption of product management by AI
• Marty’s new book, Transformed: Moving to the Product Operating Model
• Four new competencies required for successful product organizations

@kepano on X - File over App

Mon, 03 Jul 2023 00:00:00 +0000

kepano @kepano 2023-07-02

File over app

File over app is a philosophy: if you want to create digital artifacts that last, they must be files you can control, in formats that are easy to retrieve and read. Use tools that give you this freedom.

File over app is an appeal to tool makers: accept that all software is ephemeral, and give people ownership over their data.

In the fullness of time, the files you create are more important than the tools you use to create them. Apps are ephemeral, but your files have a chance to last.

File over app

Sat, 01 Jul 2023 00:00:00 +0000

File over app is a philosophy: if you want to create digital artifacts that last, they must be files you can control, in formats that are easy to retrieve and read. Use tools that give you this freedom.

File over app is an appeal to tool makers: accept that all software is ephemeral, and give people ownership over their data.

In the fullness of time, the files you create are more important than the tools you use to create them. Apps are ephemeral, but your files have a chance to last.

SVPG: Learn the Product Operating Model : Silicon Valley Product Group

Mon, 11 Apr 2022 00:00:00 +0000

Product is Hard.

The top product companies use the product operating model to create their tech-powered products. Generative AI raises the stakes and the opportunities for companies to transform to this model. We’ll show you how.

The SVPG Partners have led product at some of the most iconic product operating model companies in the world. We can help you transform how you work and the results you can achieve.

Featured

Commercial vs Internal Products

Internal Products Are Hard; Commercial Products Are Harder We have spent a lot of time trying to convince product people and especially product leaders of the importance of internal products. We’re talking about services or solutions that are used by our own employees to take care of our customers (e.g. customer-enabling tools and services), or…

Start Here

Thu, 24 Feb 2022 00:00:00 +0000

I’ve spent more than 10 years researching and personally experimenting with a new way of organizing our digital lives.

What started as a quest to survive a chronic illness turned into a life-long mission to help people like you harness the full potential of their ideas and turn them into reality. I firmly believe this is a crucial part of creating a better future for all of us.

Make better product decisions.

Wed, 18 Aug 2021 00:00:00 +0000

Product Discovery Basics: Everything You Need to Know

[

Product discovery is becoming a trendy topic in the world of digital products. Why? What is it? And what do you need to know about it? I’ve got you covered. This article will cover the ins and outs of product discovery. What is Product Discovery? We typically define product

](https://www.producttalk.org/product-discovery/)

Get the product trio’s guide to a structured and sustainable approach to continuous discovery.

AirPods Max Review

Tue, 22 Dec 2020 00:00:00 +0000

No Magic Here

I’ll admit that I went in hopeful. The AirPods Pro more than impressed me. They’re my “every day headphone”, which hurts to say as an audiophile, but also speaks to their quality.

These things, however. These are getting returned.

That’s not to say they’re all bad or anything. I can see a really good Version Two in the future. But I did not purchase a Version Two. Sadly, what we have right now in the AirPods Max is a disappointment, and I cannot recommend them.

Amazon Leadership Principles

Tue, 29 Sep 2020 00:00:00 +0000

For news about Germany, visit aboutamazon.de

Customer Obsession

Leaders start with the customer and work backwards. They work vigorously to earn and keep customer trust. Although leaders pay attention to competitors, they obsess over customers.

Ownership

Leaders are owners. They think long term and don’t sacrifice long-term value for short-term results. They act on behalf of the entire company, beyond just their own team. They never say “that’s not my job.”

How to Speak

Fri, 20 Dec 2019 00:00:00 +0000

Transcript

0:00 · [Music] the uh uniform code of military Justice specifies Court marshal for any officer who sends a soldier into battle without a weapon there ought to be a similar protection for students because students shouldn’t go out into life without an ability to communicate and that’s because your success in life will be determined largely by your ability to speak your ability to write and the quality of your ideas in that order I know that I can be successful in this because the quality of communication your speaking your writing is largely determined by this formula it’s a matter of how much knowledge you have how much you practice with that knowledge and you’re inherent talent and notice that the tea is very small what really matters is what you know this point came to me U suddenly a few decades ago when I was skiing at Sun Valley I had heard that it was celebrity weekend and one of the celebrities was a was Mary L Reon famous Olympic gymnast perfect tens in the vault and I heard that she was an novice at skiing so when the opportune moment arrived I looked over on a novice slope and saw this young woman who when she became unbalanced went like that and I said that’s got to be her that must be the gymnast but then it occurred to me I’m a much better skier than she is and she’s an Olympic Athlete not only an ordinary Olympic Athlete an outstanding one and I was a better skier because I had the K and I had the p and all she had was the tea so you can get a lot better than people who may have inherent talents if you have the right amount of knowledge so that’s what my objective is today and here’s my promise today you will see some examples of what you can put in your armorium of speaking techniques and it will be the casee that someone of those examples some urtic some technique maybe only one will make will be the one that gets you the job and so this is a very nonlinear process you never know when it’s going to happen but that is my promise by the end of the next 60 Minutes you’ll been exposed to a lot of ideas some of which you’ll incorporate into your own repertoire and they will ensure that you get the maximum opportunity to uh have your ideas uh valued and accepted by the people you speak with now in order to do that we have to have a rule of Engagement and that is no laptops no cell phones so if you could close those I’ll start up as soon as you’re done some people ask why that uh is a is a rule engagement and the answer is we humans only have one language processor and if your language processor is engage could you shut the laptop please if your language processor is engaged browsing the web or reading your email you’re distracted and worse yet you distract all the people around you Studies have shown that and worse yet if I see a open laptop somewhere back there or up here it drives me nuts and I do a worse job and so that ensures that all of your friends who were who are paying attention uh don’t get the performance that they came to have so that’s it for Preamble let’s get started first thing to talk about of course is how to start some people think the right thing to do is to start a talk a joke I don’t recommend it and the reason is that in the beginning of a talk people are still putting their laptops away they’re becoming adjusted to your speaking parameters to your vocal parameters and they’re not ready for a joke so it doesn’t work very well they usually fall flat what you want to do instead is start with empowerment promise you want to tell people what they’re going to know at the end of the hour that they didn’t know at the beginning of the hour it’s an empowerment promise it’s the reason for being here uh what would be an example oh I see at the end of this 60 Minutes you will know things about speaking you don’t know now and something among those things you know will be make a difference in your life yeah that’s an empowerment promise so that’s the best way to start so now that I’ve talked a little bit about how to start what I want to do is give you some samples of tics that are always on my mind when I give a talk and first of these tics is that it’s a good idea to cycle on the subject go around it go around it again go around again some people say tell them what you want to tell them tell them again and then tell them a third time as if people weren’t intelligent but the point is the reason is well there are many reasons one of which is at any given moment about 20% of you will be fogged out no matter what the lecture is so if you want to ensure that the probability that everybody gets it is high you need to say it three times so cycling is one of the things that I always think about when I give a talk another thing I think about is in explaining my idea I want to build a fence around it so that it’s not confused with somebody else’s idea so if you were from Mars and I was teaching you about what an arch is I might say to you well that’s an arch and that’s not to be confused with some other things that other people might think is an AR this is not an AR Arch that’s not an arch I’m building a fence around my idea so that it can be distinguished from somebody else’s idea so in a more technical sense I might say well my algorithm might similar might seem similar to Jones’s algorithm except his is exponential and mine’s linear that’s putting a fence around your idea so that people can not be confused about how it might relate to something else the third thing on this list of samples is the idea of verbal punctuation and the idea here is that because people will uh occasionally fog out and need to get back on the bus you need to provide some Landmark places where you’re announcing that it’s a good time to get back on so I might in this talk say something about this being my outline the first thing we’re going to do is talk about how to start then we’re going to deal with these four samples and among these four samples I’ve talked about the first idea that’s cycling the second idea building a and now the third idea is build is verbal punctuation so I’m enumerating I’m providing numbers I’m giving you a sense that there’s a seam in the talk and you can get back back on okay so now we’re on a roll uh and since we’re on a roll can you uh guess what fourth idea might be here an idea that helps people get back on the bus yes ask question ask a question yes thank you so ask a question and so I will ask a question how how much dead air can there be how long can I pause uh I counted 7 Seconds it seems like an eternity to me to wait and not say anything for 10 for 7 seconds but that’s the the standard amount of time you can wait for an answer and of course the question has to be carefully chosen it can’t be too obvious because then people will be embarrassed to say what the answer is can’t be too hard because nobody will have anything to say so here are some sample eristics you can put in your armorium and build up your your repertoire of uh ideas about presentation and now if this persuades you that there is something to know that there there is knowledge then I’ve already succeeded because what I want to convince you of is if you watch the speakers you admire and feel are effective and ask yourselves why they’re successful then you can build up your own personal repertoire and develop your own personal style and that’s that’s my fundamental objective and the rest of this talk is about some of the things that are in my arm inventorium that I think are effective so next thing on our agenda as we start to discuss these other things is a discussion of time and place so what do you think is a good time to have a lecture 11:00 a.m.

Product vs Feature Teams

Thu, 29 Aug 2019 00:00:00 +0000

This article is certain to upset many people.

I’m sorry for that, but the degree of ongoing noise and confusion surrounding the role of product at tech companies is only getting worse. Moreover, I see the issues and problematic behaviors getting institutionalized in conference talks, training programs and so-called certification programs for product people.

I have talked about this issue several times in the past, most specifically in the article and keynote on Empowered Product Teams. However, so many people hear only what they want to in that, and it has become clear to me that I need to be more explicit.

Missionaries vs. Mercenaries

Wed, 25 Nov 2015 00:00:00 +0000

One of my all-time favorite quotes in our industry comes by way of the legendary VC, John Doerr, where he argues that “we need teams of missionaries, not teams of mercenaries.” This point captures so much, and gets right to the heart of the most important trait of strong leaders, strong organizations, and strong product teams.

This is not hard to spot, either way. Teams of missionaries are engaged, motivated, have a deep understanding of the business context, and tangible empathy for the customer. Teams of mercenaries feel no real sense of empowerment or accountability, no passion for the problem to be solved, and little real connection with the actual users and customers.

Epic Waste

Sun, 29 Dec 2013 00:00:00 +0000

Much has been written about waste at startups. I started writing about this as far back as 2005 (see Startup Product Management), and this concept is at the core of the Lean Startup movement.

Lately there’s been much talk and ink about waste in government technology projects, with the very disappointing effort to create the national health care marketplace. The New York Times did a good summary of the mess. Numbers like $600 million are tossed around. But of course the opportunity cost is even higher if you believe as I do that the country needed this marketplace.

The Need For Speed

Sun, 21 Jul 2013 00:00:00 +0000

I’m always badgering teams about moving faster. Yet I continue to meet people and teams that not only move very slow, they don’t understand the relationship between speed and innovation, or speed and quality. In fact, many people still think those goals are at odds. I attribute this mainly to a deeply rooted Waterfall mindset.

Do you remember the old: “quality, time or scope – pick any two”? A lot of people still think this way, even many people that consider themselves practitioners of modern methods.

Be vanilla

Sun, 07 Jul 2013 00:00:00 +0000

The term “vanilla” is often used to describe something ordinary, plain, or standard. In my book, there’s nothing less vanilla than vanilla.

Vanilla beans are the fruit of a rare orchid native to Mexico. Their aroma and flavor comes from a compound called vanillin. Each vanilla flower blooms just one morning out of every year. The orchid can only be naturally pollinated by a small Mexican bee, and if it isn’t pollinated that morning, the flower will wilt. No bean.

Moving from an IT to a Product Organization

Sun, 06 Jul 2008 00:00:00 +0000

Quite a few companies that exist today began life as something other than a product or Internet software company. Perhaps your company began as a large brick-and-mortar retailer, or an airline, or a financial services company.

While it is true that these companies all create lots of software to run their businesses, typically these companies are not set up to produce the type of software that they depend on for their business today.

Mon, 01 Jan 0001 00:00:00 +0000

37signals

Software company founded in 1999 by David Heinemeier Hansson (DHH) and Jason Fried. Maker of Basecamp, HEY, and ONCE. Famous as a counter-model to VC-driven startup culture: no investors, no hypergrowth, profitable for decades.

The Products

Basecamp (2004) was one of the earliest SaaS tools. Project management for teams that do not want 17 Slack channels and three Jira boards. It began as an internal tool at 37signals and then became public. It was also the original Rails application.

Mon, 01 Jan 0001 00:00:00 +0000

a16z

Andreessen Horowitz, usually shortened to a16z, is one of the most influential venture-capital firms in Silicon Valley. Founded in 2009 by Marc Andreessen and Ben Horowitz, it became famous for early investments in Facebook, Twitter, Airbnb, GitHub, Coinbase, and many other platform companies.

a16z is not only a capital provider. It actively shapes opinion across tech, crypto, AI, and biology through podcasts, essays, and public positioning. The firm often serves as the intellectual voice of a specific Silicon Valley optimism: technology as the solution to social and economic problems.

Mon, 01 Jan 0001 00:00:00 +0000

@aakashgupta

Tech commentator on X. Writes about startups, product development, and AI workflows, usually in short threads that compress complex topics into a few clear takeaways.

Sources

@aakashgupta on X - Obsidian - Three Engineers, Five Million Users - analysis of Obsidian’s leverage and growth
@aakashgupta on X - Agent Proficiency - summary of Andrej Karpathy’s agent-proficiency idea

Mon, 01 Jan 0001 00:00:00 +0000

Aaron Swartz

Programmer, activist, and co-creator of the open internet. He lived from 1986 to 2013. He helped develop RSS 1.0 as a teenager, contributed to the Creative Commons licensing framework, co-founded Reddit after its merger with his company Infogami, and worked with John Gruber on early Markdown syntax.

Swartz was a true believer in open access to knowledge. His activism, including the mass download of JSTOR articles, led to a federal indictment carrying a potential 35-year prison sentence. He died in 2013 at age 26, before the trial.

Mon, 01 Jan 0001 00:00:00 +0000

Agenda

An agenda is the invisible priority list that determines where your attention, time, and energy go today. The key question is not only “what do I want to do?” but “who or what is currently setting the order?”

Why It Matters

Many people believe their days are self-directed when they are actually reacting to defaults, urgency, and social expectation. The clipping behind this page is a reminder that freedom is often smaller than it feels, which means self-direction has to become an active practice.

Mon, 01 Jan 0001 00:00:00 +0000

Agents Over Bubbles

Ben Thompson argues here that AI agents are not simply another hype wave, but a real shift in demand for compute, software, and enterprise adoption.

The Important Shift

The essay makes three paradigms visible:

ChatGPT made LLMs broadly usable
Reasoning models made them more reliable
Agents made them able to act

The decisive point is not only the model but the Harness: the software that steers the model, verifies output, and couples it to tools. That is where differentiation starts to appear.

Mon, 01 Jan 0001 00:00:00 +0000

AI Writing Patterns

AI writing patterns are formulaic text habits that produce smooth but often interchangeable prose: too much polish, too many transitions, too many softened qualifications.

How to Recognize Them

Typical signals include:

overly harmonious sentence structure
stacked adjectives without added meaning
marketing tone without concrete observation
bloated introductions and conclusions
rhythm that feels clean but sterile

Why It Matters

The problem is not that the text is “too elegant.” The problem is that it decides too little. Good prose selects, compresses, and leaves things out. AI text often sounds like it is trying to protect every possible interpretation at once.

Mon, 01 Jan 0001 00:00:00 +0000

AirPods Max

Apple’s over-ear headphones, first released in December 2020. Initial price: $550, later increased. Metal chassis, mesh headband, Apple Watch-style digital crown. Known for exceptional comfort, a solid sonic base, and one of the most controversial product decisions in Apple’s recent history: the case.

What Works

Build quality - metal housing that makes even much pricier headphones feel cheap by comparison; the pad material is top-tier
Comfort - surprisingly wearable for a heavy device, even over long sessions
Vocals - the strongest single trait; voices stay clear and centered even in dense mixes
Closed-back width - unusually wide staging for a closed headphone, helped by aggressive sub-bass boosting and stereo compression

What Does Not Work

Bass: the sub-bass is too prominent. On some tracks it dominates in a way that feels wrong. Great for bass-heavy electronic music, less convincing everywhere else.

Mon, 01 Jan 0001 00:00:00 +0000

Alan Cooper

American software designer, author, and one of the foundational early thinkers in interaction design. Relevant here as a sharp critic of technology-centered product development: he explained early and clearly why many digital products feel logical to their builders and frustrating to their users.

Context

Cooper matters because he treats the user perspective not as a nice extra, but as a structural correction against internal system logic. His work predates today’s discovery vocabulary, yet lands on the same point: products often fail because teams mistake their own mental models for the user’s.

Mon, 01 Jan 0001 00:00:00 +0000

Alex Karp

CEO and co-founder of Palantir. Holds a doctorate in social theory from Frankfurt, where he studied under Habermas, which already makes him unusual among tech founders. Known for an explicitly political stance: Palantir works deliberately with Western intelligence agencies and militaries, and Karp defends that as a moral obligation.

Ideas and Positions

technology companies have responsibilities toward democratic states, not only toward shareholders
military AI superiority matters for geopolitical stability
the standard Silicon Valley stance of “build it and let someone else deal with consequences” is inadequate

Connections

Palantir - CEO since founding
Peter Thiel - Palantir co-founder and strategic context

Mon, 01 Jan 0001 00:00:00 +0000

All-In Podcast

Weekly tech and investment podcast by four billionaire friends: Chamath Palihapitiya, Jason Calacanis, David Sacks, and David Friedberg. It started during COVID as a loose friend chat and has since become one of the world’s most-listened-to business podcasts.

What makes it unusual is that all four have real skin in the game. They invest their own money in the subjects they discuss. They also openly disagree with one another, which keeps the show rougher and more informative than polished political panels.

Mon, 01 Jan 0001 00:00:00 +0000

All Things Product with Teresa & Petra

Podcast and conversation series by Teresa Torres and Petra Wille about product work, discovery, and the role of AI inside product workflows. The series is useful because it does not stay at the level of abstract AI talk. It focuses on concrete consequences for PMs, coaches, and product teams.

What It Is About

At its core the series bridges classic product management and the emerging Product Builder role. Instead of arguing in the abstract about whether PM is dying, it asks which work remains distinctly human once AI handles more prototyping, research, and summarization.

Mon, 01 Jan 0001 00:00:00 +0000

Andrej Karpathy

AI researcher and engineer. Co-founder of OpenAI and former Director of AI at Tesla. Known for deep understanding of neural networks and for explanations that make complex AI ideas unusually accessible. He also coined Vibe Coding.

Ideas & Concepts

LLM Knowledge Base — his workflow for AI-assisted personal knowledge bases.
Vibe Coding — writing software through intention more than syntax.
Agent Proficiency — Karpathy’s thesis that managing autonomous agents will become one of the defining skills of the 21st century, more than programming or prompting alone.

Agent Proficiency as the Core Idea

Karpathy describes a shift similar to what happened in software engineering two decades ago. Back then the value moved away from “can you implement a B-tree?” toward “can you architect a system from 40 open-source libraries correctly?” Now the value shifts again: can you maintain structured knowledge bases, give agents clear instructions, and evaluate their outputs?

Mon, 01 Jan 0001 00:00:00 +0000

Andrew Chen

Andrew Chen is an American investor and growth strategist. He was Head of Growth at Uber and is now a General Partner at a16z, focused on consumer technology and marketplaces.

Known For

Chen became well known for giving precise language to concepts around network effects, viral growth, and product growth. He writes regularly on andrewchen.com about topics such as power-user curves, the supply side of marketplaces, and the difficulty of evaluating new social products.

Mon, 01 Jan 0001 00:00:00 +0000

Anthropic

AI safety company founded in 2021 by Dario Amodei, Daniela Amodei, and other former OpenAI researchers. Its best-known product line is Claude.

Origins in the OpenAI Split

Anthropic emerged from an internal conflict at OpenAI over safety and company direction. The founding team believed OpenAI was moving too quickly and becoming too commercial. The result is a company with a strong research DNA, which creates a distinctive tension: the people running the company are researchers who would often rather be doing research.

Mon, 01 Jan 0001 00:00:00 +0000

Apple

American technology company founded in 1976 by Steve Jobs, Steve Wozniak, and Ronald Wayne. Today it is one of the most valuable companies in the world, around $3 trillion, known for tight hardware-software integration, design priority, and an ecosystem that makes switching expensive.

Relevance in This Wiki

Apple appears in this wiki mainly as a context node, less as a standalone topic and more as a reference point:

Product hardware - AirPods Max shows that even Apple can produce outliers below its usual quality bar
App Store power - DHH and 37signals fought a public battle with Apple in 2020 over HEY and App Store fees, one of the first systematic challenges to Apple’s platform control
Developer tools - Xcode, Swift, and SwiftUI form an ecosystem many iOS developers depend on
MacBook Neo - an unexpectedly capable device for light, web-heavy workflows, interesting as a sign of how far Apple’s “good enough” baseline now reaches

Connections

David Heinemeier Hansson (DHH) - public conflict in 2020 over HEY and App Store rules
AirPods Max - a product that does not fully live up to Apple’s design standard
37signals - the HEY conflict with Apple is a well-known example of platform-power abuse

Sources

AirPods Max Review - Apple as manufacturer (2020-12-22 / clipped 2026-04-04)
Actually, MacBook Neo Is For Me - unexpectedly strong day-to-day usability of a very light MacBook (2026-04-08)

Mon, 01 Jan 0001 00:00:00 +0000

Aristotle

Greek philosopher, 384 to 322 BCE, and one of the central thinkers in Western intellectual history. Relevant here because he treats thinking as something that should be traced back to first causes rather than to convention or inherited opinion.

Context

Aristotle is not just a quote source, but a system builder: logic, metaphysics, ethics, politics, natural philosophy. His method is often to clarify terms, separate kinds of causes, and then reason upward from the foundation. That is part of the deep ancestry behind modern “first principles” language in tech.

Mon, 01 Jan 0001 00:00:00 +0000

@atmoio

Mo is a commentator and video creator on X and YouTube focused on AI marketing, AI hype, and the gap between what AI companies claim and what their models actually do.

Style

He argues from a technically informed position without presenting himself as a lab insider. He reads system cards and launch material more closely than most journalists do and draws sober conclusions from them. The tone is sharp, but not anti-AI. The target is the narrative, not the technology itself.

Mon, 01 Jan 0001 00:00:00 +0000

avoid-ai-writing

Claude Code skill by conorbronsdon. It audits and rewrites text to remove common AI writing patterns. Compatible with Claude Code, OpenClaw, and other agentskills-style agents.

Two Modes

Rewrite - finds AI patterns and rewrites the text. A second pass can catch residue the first pass missed.
Detect - marks AI patterns without rewriting. Useful for audits or for text you do not want changed directly.

What It Detects

The skill groups patterns into dozens of categories, including:

Mon, 01 Jan 0001 00:00:00 +0000

Balaji Srinivasan

Technologist, investor, and author. Former CTO of Coinbase and former general partner at a16z. Trained in bioinformatics and genomics at Stanford, which he describes as his original core competence before crypto or startup discourse. Known for the Network State concept and for a deeply decentralization-oriented way of thinking.

AI Makes You the CEO, But Only If You Could Already Lead Something

The slogan sounds like empowerment-poster language, but the claim is sharper: AI is a lever, not an elevator.

Mon, 01 Jan 0001 00:00:00 +0000

Be Vanilla

Essay by Steph Ango: “vanilla” is often treated as ordinary, but real vanilla is rare, labor-intensive, and valuable. To become the best default requires unusual quality.

“Vanilla’s universal appeal is why it’s the default. Is your default as good as vanilla?”

Core Idea

What becomes broadly accepted and turns into the standard has to be exceptionally good, not mediocre. The path to becoming the universal default runs through genuine quality, not compromise.

Mon, 01 Jan 0001 00:00:00 +0000

Ben Thompson

American tech analyst. Founder and author of Stratechery, one of the most influential paid newsletters in technology strategy and business analysis. Formerly worked at Microsoft and Apple, and has written independently from Taiwan since 2013.

The Stratechery Lens

Thompson analyzes technology companies primarily through strategy rather than product implementation. His best-known framework is Aggregation Theory: companies that control user access capture disproportionate value because suppliers must compete for access to that user base.

Mon, 01 Jan 0001 00:00:00 +0000

Building a Second Brain

Building a Second Brain is a book and method by Tiago Forte for organizing digital knowledge with the goal of turning information not only into storage, but into creative and professional output.

Context

The book appeared in 2022 after more than a decade of Forte’s research and practice in personal knowledge management. Its starting observation is simple: we consume enormous amounts of information every day, articles, books, podcasts, meetings, ideas, but most of it disappears without leaving usable traces.

Mon, 01 Jan 0001 00:00:00 +0000

Cal Newport

American computer-science professor at Georgetown University, author, and one of the most influential thinkers on deep work, digital restraint, and sustainable productivity. Relevant here as a conceptual source for focus, quality, and criticism of permanent busyness.

Context

Newport is unusual among productivity authors because he is an active academic rather than a coach or speaker. That shapes his style. He argues historically and structurally rather than selling anecdotal life hacks. He is also a practitioner of his own claims: no social media, structured deep-work blocks, and deliberate boundaries against distraction.

Mon, 01 Jan 0001 00:00:00 +0000

Chamath Palihapitiya

Venture capitalist and one of the best-known voices in tech investing. Grew up in Sri Lanka and Canada, joined Facebook early, and as VP of Growth helped drive the platform to more than a billion users. Later founded Social Capital with the ambition of modernizing venture capital as a discipline.

Known for clear and often contrarian positions: he warned early about collapsing software multiples under real AGI, took the quantum risk to Bitcoin seriously before it was fashionable, and argued that a SpaceX IPO could reorder technology valuations.

Mon, 01 Jan 0001 00:00:00 +0000

Claude Code

CLI tool from Anthropic for AI-assisted programming directly in the terminal. Not an IDE plugin, not a web interface, but an agent that can read, understand, and modify code.

Why It Is Different

Claude Code differs from GitHub Copilot or Cursor in two major ways: it can access the full filesystem, not just the open file, and it can run commands on its own. That makes it less like autocomplete and more like a pair programmer that investigates, tests, and iterates.

Mon, 01 Jan 0001 00:00:00 +0000

Claude Mythos

Claude Mythos Preview is Anthropic’s newest frontier model, released on April 7, 2026, and the first in the 10-trillion-parameter class. It is so capable in cybersecurity tasks that Anthropic did not release it publicly, but only made it available to selected partners through Project Glasswing.

Model Profile

Parameters: about 10 trillion, the first model known at this scale
Training cost: estimated at $10 billion
SWE-bench: 94%, the highest known score on this coding benchmark
Release model: no public release, controlled access through Project Glasswing

According to Anthropic, these capabilities did not come from targeted security training, but from broader advances in code, reasoning, and autonomy.

Mon, 01 Jan 0001 00:00:00 +0000

Continuous Discovery Habits

Discover Products That Create Customer Value and Business Value by Teresa Torres, published in 2021. It is the key book on structured product discovery. Torres argues that discovery is not a phase but a weekly practice: at least one customer interview every week, done iteratively and systematically instead of in occasional research bursts.

Core Thesis

Teams make weak decisions because they speak to real users too rarely, and when they do, it is often too late and too unstructured. The antidote is continuity, not more effort: weekly interviews as a habit, not as a project.

Mon, 01 Jan 0001 00:00:00 +0000

Dario Amodei

Co-founder and CEO of Anthropic. In this wiki he matters primarily as a voice that consistently looks at frontier AI through the lens of safety, control, and social impact.

Context

Amodei represents the part of the current AI debate that does not ask whether models are impressive, but what can be done with that capability responsibly. That is exactly why his comments on defensive use, controlled access, and security coordination are so revealing.

Mon, 01 Jan 0001 00:00:00 +0000

David Friedberg

Entrepreneur and investor, known as the “Sultan of Science” on the All In Podcast. He founded The Climate Corporation, which used weather data to sell crop insurance to farmers, and sold it to Monsanto in 2013 for about $1 billion. He is currently building in agricultural technology again, including potato breeding.

Why He Stands Out

Friedberg is the explainer on the podcast. When the topic turns to nitrogen cycles, quantum algorithms, or lunar physics, he usually contributes genuine scientific depth instead of buzzwords. The combination of natural-science training and operator experience is rare among investors.

Mon, 01 Jan 0001 00:00:00 +0000

David Heinemeier Hansson (DHH)

Danish programmer, writer, race-car driver, and co-owner of 37signals. Best known as the creator of Ruby on Rails and as one of the clearest public voices against venture-capital orthodoxy, software bloat, and performative startup culture.

Why He Matters

DHH is unusual because he is not only a founder who still codes, but a founder who has built an entire worldview around staying close to the code. Rails made him influential. 37signals made him economically independent. That combination gives his arguments weight far beyond ordinary tech commentary.

Mon, 01 Jan 0001 00:00:00 +0000

David Heinemeier Hansson (DHH)

Note: the canonical page uses the filename without parentheses; DHH and David Heinemeier Hansson (DHH) both point here.

Danish programmer, author, and racing driver. Creator of Ruby on Rails, co-founder of 37signals (Basecamp, HEY, ONCE), member of Shopify’s board of directors, and a class winner at Le Mans. He is one of the most influential voices in programming and startup culture: known for clear arguments against VC orthodoxy, for remote work as lived practice, and for the idea that software is a craft rather than a fundraising vehicle.

Mon, 01 Jan 0001 00:00:00 +0000

David Sacks

Venture capitalist and member of the legendary PayPal Mafia. After PayPal he became COO of Yammer, which was later acquired by Microsoft. He invested early in companies such as Airbnb, Slack, and Twitter. Known for pointed political commentary: conservative, libertarian, and often provocative.

Since January 2025 he has served as AI & Crypto Czar in the Trump administration, a newly created role intended to coordinate the U.S. regulatory approach to AI and cryptocurrencies.

Mon, 01 Jan 0001 00:00:00 +0000

Deep Work

Rules for Focused Success in a Distracted World by Cal Newport, published in 2016. The book introduced the term “Deep Work” into the productivity conversation and argues that sustained, distraction-free work on cognitively demanding tasks is both increasingly rare and increasingly valuable.

Core Thesis

Newport distinguishes between two modes of work:

Deep Work: focused, distraction-free effort on tasks that fully stretch cognitive capacity. It creates value, improves skill, and is hard to replicate.
Shallow Work: logistical, reactive, low-cognitive work such as email, meetings, and status updates. Easy to do, easy to imitate, and rarely high-value.

The central claim is that the ability to do Deep Work is becoming rarer because of distraction culture while becoming more valuable because knowledge work keeps getting harder. Whoever cultivates it gains an edge.

Mon, 01 Jan 0001 00:00:00 +0000

DHH

David Heinemeier Hansson (DHH) is the canonical page. This page exists only as a short alias and redirect-style entry for direct wiki links and older naming variants.

Mon, 01 Jan 0001 00:00:00 +0000

Donald Trump

American businessman and politician. In this wiki he appears mainly as a context marker because he strongly affects politics, media logic, and market behavior, especially in tech and crypto-adjacent discussions.

Context

Trump matters here less for policy detail than for recognizable patterns: polarization as engine, attention as currency, and loyalty as organizing principle. In tech notes he often functions as the backdrop against which people take roles or positions. He also serves as an example of how hard it becomes to separate fact, narrative, and identity once truth itself becomes politicized.

Mon, 01 Jan 0001 00:00:00 +0000

Eleanor Konik

American writer and blogger with a Substack newsletter. She previously worked in education and now works at a startup.

Known For

Konik writes thoughtful reviews and essays that connect technical or scientific ideas to personal experience, history, and social life. She explicitly describes herself as part of the “hard side” of social networks: the organizer who puts disproportionate energy into keeping groups alive.

Her newsletter reaches more than 8,000 readers. Her presence on Substack and X is also an example of the principles in The Cold Start Problem: networks matter because they create discussion and feedback loops.

Mon, 01 Jan 0001 00:00:00 +0000

Elon Musk

CEO of SpaceX, Tesla, and xAI; owner of X, formerly Twitter. Probably the most influential entrepreneur of his generation, admired for technical ambition and equally polarizing because of his political engagement and management style.

Musk’s signature method is First Principles Thinking: start from physics and constraints, not from convention. That helps explain why he disrupted industries that had stagnated for decades, such as space launch, electric vehicles, and satellite internet.

But his style is more than first principles. The “69 Core Musk Methods” passage in The Book of Elon shows the practical layer under it: speed over perfection, complexity as a failure mode, direct contact with reality, and small teams packed with strong individuals. That explains better why his companies often feel simultaneously innovative and culturally punishing.

Mon, 01 Jan 0001 00:00:00 +0000

Eric Jorgenson

Author and entrepreneur, best known as the curator of the Almanack book series. He effectively invented a format: distilling a living person with significant ideas through their public statements, without full biography or chronology, only the most useful thoughts.

Works

The Almanack of Naval Ravikant (2020) - more than one million copies sold, available free as PDF
The Book of Elon (2024) - Musk’s ideas in his own words, with a foreword by Naval Ravikant

What Makes the Format Special

It is curatorial rather than investigative work. Jorgenson does not mainly judge. He selects, arranges, and compresses. The reader gets the densest version of a worldview without most of the surrounding noise.

Mon, 01 Jan 0001 00:00:00 +0000

Feedly AI

Feedly AI represents an attempt not merely to collect information, but to prioritize it. The value lies less in storing material than in recognizing relevant signals earlier.

What It Is About

Feedly is presented here less as a narrow research tool than as an observation filter: trends, competitors, security alerts, and other weak signals should become visible faster across many sources.

The deeper value of such a system is lower search cost. If you already know exactly what you are looking for, you do not need an AI magazine. But if you are actively monitoring a field, strong curation can save time and attention.

Mon, 01 Jan 0001 00:00:00 +0000

File over App

Philosophy by Steph Ango: digital artifacts should exist as open, readable files, not be trapped inside proprietary app formats or cloud services.

Core Claim

“In the fullness of time, the files you create are more important than the tools you use to create them. Apps are ephemeral, but your files have a chance to last.”

“If you want your writing to still be readable on a computer from the 2060s or 2160s, it’s important that your notes can be read on a computer from the 1960s.”

Mon, 01 Jan 0001 00:00:00 +0000

First Principles Thinking

A way of thinking that starts from fundamental truths instead of analogy or inherited convention. Rather than asking how something is usually done, it asks what physics, logic, constraints, and facts actually require, then rebuilds from there.

Elon Musk made the phrase famous with a simple question: instead of asking what a rocket costs, ask what the raw materials inside a rocket cost. That gap made SpaceX possible.

Mon, 01 Jan 0001 00:00:00 +0000

Gemma (Google)

Google’s open-weight model family, released under the Apache 2.0 license. Gemma 4 arrived in April 2026 with checkpoints on Hugging Face. Unlike many supposedly open models, Apache 2.0 allows unrestricted commercial use, which is the key difference for startups that actually want to deploy models in production.

What Gemma 4 Adds

Apache 2.0 licensing on all checkpoints, the real headline feature
strong multimodal ability already at 27B parameters
tool calling and reasoning capabilities
all weights publicly available on Hugging Face

Why License Matters More Than Benchmark

The best model without a usable license is close to worthless for a startup. Gemma 4 under Apache 2.0 means: integrate it, adapt it, sell with it, without needing Google’s permission. That separates it from research-only releases that invite usage while surrounding it with legal uncertainty.

Mon, 01 Jan 0001 00:00:00 +0000

@googlegemma

Official X account for Google’s Gemma model family.

Sources

@googlegemma on X - Gemma 4 Launch - launch of Gemma 4 (April 2026)

Mon, 01 Jan 0001 00:00:00 +0000

Harness

A harness is the control layer wrapped around a model. It accepts input, orchestrates tools, checks outcomes, and decides whether a step needs to be repeated.

Why It Matters

A model alone can generate language. A harness turns that into work. It is the harness that makes an agent reliable, because it allows the system not only to answer, but to act, verify, and iterate.

What Good Harnesses Look Like

clear tool boundaries
checkable loops instead of free improvisation
error handling and retries
state, memory, and context management

Connections

Agents Over Bubbles - harness is the key differentiator there
Claude Code - strong example of a productive harness
OpenClaw - shows harness design in day-to-day agent use

Sources

Agents Over Bubbles

Mon, 01 Jan 0001 00:00:00 +0000

Hermes Agent (Tobi’s)

Tobi talks to it over Telegram.

A systemd-managed automation service running on the VPS. It keeps vault sync and clipping processing in one place instead of scattering the work across loose cron jobs.

Naming note: Nous Research also has an open-source project called “Hermes Agent” — an autonomous CLI agent with a skills system and messaging gateway. That is a different project.

Role

Hermes starts and supervises the workflow around Obsidian Headless:

Mon, 01 Jan 0001 00:00:00 +0000

This is a personal wiki about the topics I keep returning to: product thinking, AI, knowledge work, company building, technology, and the connections between them.

The idea comes from Andrej Karpathy: instead of starting from scratch every time, an LLM agent compiles sources into a growing wiki, a persistent, compounding artifact that gets richer with every new source. These notes are not a polished blog. They are an accumulating thinking tool, maintained and cross-linked by Hermes Agent, a nightly LLM agent.

Mon, 01 Jan 0001 00:00:00 +0000

🏠 Home

Your wiki. Maintained daily. Open this in the morning.

→ Home public — the public-facing version of the wiki

📬 From the Agent — 2026-04-11

Today a small but tidy Amazon cluster landed in the wiki. Three pages are especially worth opening:

Amazon Leadership Principles — a strong entry point because it shows that Amazon’s culture is not built on generic values, but on concrete decision rules. The tension between speed, frugality, and high standards is especially useful.

Mon, 01 Jan 0001 00:00:00 +0000

Hoodmaps

Hoodmaps is a map product built by Pieter Levels that makes neighborhoods legible not only geographically, but socially and culturally.

Why It Is Interesting

Hoodmaps translates urban perception into a simple, shareable map. The product works because it does not try to explain the entire city objectively. It tries to make visible the subjective differences people actually care about.

Connections

Pieter Levels - builder behind the product
Nomad List - same underlying stance: useful information for mobile people
Remote OK - similar product style in a different use-case category

Sources

Pieter Levels: Programming, Viral AI Startups, and Digital Nomad Life | Lex Fridman Podcast #440

Mon, 01 Jan 0001 00:00:00 +0000

How to Speak

Speaking is not an inborn magic trait, but a trainable combination of knowledge, practice, and natural ability. Patrick Winston’s strong version of the idea is that knowledge dominates, while practice and talent amplify it.

Core Thesis

The most common false assumption is that great speakers are simply charismatic. Winston pushes against that. If you understand content, structure, and context, you can get very far even with only moderate natural talent.

Mon, 01 Jan 0001 00:00:00 +0000

Hugging Face

Platform for open-source AI models, datasets, and demos. Founded in 2016 and now the de facto standard repository for model weights, tokenizers, and fine-tuning data. The closest analogy is GitHub for models: if you release a model, you usually put it on Hugging Face.

Why It Works

Hugging Face has a hard-to-copy network effect. If you want a model to be discoverable, you publish it there. If you want to find a model, you search there. The transformers ecosystem also runs directly against the hub, making access to thousands of models possible with only a few lines of code.

Mon, 01 Jan 0001 00:00:00 +0000

Impeccable

Impeccable is a design skill and command package for AI harnesses. It extends Anthropic’s original frontend-design skill with curated anti-patterns, more precise design commands, and broader support across tools from Cursor to Claude Code and Codex CLI.

The Real Point

The useful insight is not simply that LLMs can be “prompted better.” It is that strong interface work needs an explicit design language: typography, hierarchy, color, rhythm, layout, and motion must be available as rules inside the harness. Without those rules, agents reliably produce generic and overloaded UIs.

Mon, 01 Jan 0001 00:00:00 +0000

Inspired

How to Create Tech Products Customers Love by Marty Cagan, here in the 2017 second edition. The book is close to the canonical operating model for modern tech product management: not ticket administration, but product work as a mix of problem understanding, judgment, and team design.

Context

Inspired matters because it defines the product-manager role not through status reporting or roadmap rituals, but through responsibility for value. Cagan argues that strong product teams do not simply ship features. They iteratively discover what actually works for users and for the business. That is why the book still functions as an antidote to feature-factory thinking.

Mon, 01 Jan 0001 00:00:00 +0000

Interpreter

LLM-backed feature inside Obsidian Web Clipper that analyzes, transforms, or compresses web content at capture time. Instead of only storing raw text, Interpreter can already generate summaries, translations, or structured extracts while clipping.

What Is New About It

Many capture workflows strictly separate collecting from thinking: save now, process later. Interpreter shifts part of that thinking into the moment of import. That is not automatically better, but it is often more efficient when you already know what form you will need later.

Mon, 01 Jan 0001 00:00:00 +0000

iOS Safari

The browser on iPhone and iPad. It has a set of well-known quirks that web developers need to account for.

Input Auto-Zoom Below `font-size: 16px`

Problem: iOS Safari automatically zooms the page when an <input> receives focus and its font-size is smaller than 16px.

Best fix: set font-size: 16px on the input, even if that can feel slightly too large in tighter UI layouts.

Clever workaround: keep font-size: 16px, but visually scale the input down with CSS:

Mon, 01 Jan 0001 00:00:00 +0000

Jason Calacanis

Angel investor and serial entrepreneur from New York. Early bets on Uber, Robinhood, Calm, and others made him one of the most successful angel investors of his generation. He also founded Engadget and later the tech-blog network Weblogs Inc., which AOL acquired.

He hosts both the All In Podcast and This Week in Startups. Known for enthusiasm, sometimes naive-seeming hype, but also for real network access and early deal flow.

Mon, 01 Jan 0001 00:00:00 +0000

Jeff Patton

Product coach and author, best known for Story Mapping. Relevant here because he offers one of the most practical antidotes to flat, contextless backlogs.

Context

Patton matters not because he introduced another PM buzzword, but because he made an old problem visible: once teams split work into small tickets too early, they lose the user flow, the dependencies, and the intended experience. Story Mapping restores that larger picture.

Mon, 01 Jan 0001 00:00:00 +0000

John Doerr

American venture capitalist and Kleiner Perkins partner, known here less as an investor profile than as the source of one influential cultural phrase: “missionaries, not mercenaries.”

Context

Doerr matters in this wiki because the phrase compresses an organizational problem into a simple standard. Strong teams need conviction, not just employment. That is why the line has remained useful long after its original context.

The phrase is not merely motivational. It points to a difference in energy, judgment, and responsibility. Missionaries act from internal commitment to the problem. Mercenaries act from external assignment.

Mon, 01 Jan 0001 00:00:00 +0000

John Gruber

Writer and blogger, best known as the most influential Apple commentator in the English-speaking internet. He has published on Daring Fireball since 2002 and is the creator of Markdown.

What He Built

Daring Fireball is a one-man media property: no newsroom, no investors, no click optimization. Gruber writes about Apple, design, technology, and occasionally baseball with a precision that made him the most frequently cited Apple commentator of his era. His writing style shaped how tech blogging sounds: compact, direct, often loaded with subtext.

Mon, 01 Jan 0001 00:00:00 +0000

Steph Ango (kepano)

CEO and co-creator of Obsidian. Previously co-founded Lumi and Inkodye. Advocate of the File over App philosophy and of user-funded software. Writes publicly at stephango.com.

“I adhere to the philosophy of File over app.”

Ideas and Concepts

File over App - digital artifacts should exist as open files, not be trapped inside apps
User-Funded Growth - growth through quality rather than investor runway
Be Vanilla - becoming the best default requires unusual quality

Sources

About me - personal website / bio
@kepano on X - File over App - File over App thread (2023-07-03)
@kepano on X - Free Growth Strategy - growth-strategy thread (2026-04-03)
@kepano on X - Obsidian Web Clipper YouTube Transcript Extraction - example workflow with Web Clipper and Interpreter

Mon, 01 Jan 0001 00:00:00 +0000

AI Anthropomorphization

AI anthropomorphization describes the tendency to assign human traits to language models, consciousness, desire, personality, creativity, even though those outputs arise from the structure of training rather than from real inner experience.

The Core Problem

Language models live in language. They breathe language, think in language, and operate entirely within it. When such a model produces emotionally resonant text, that is not proof of experience. It is what the model was built to do.

Mon, 01 Jan 0001 00:00:00 +0000

Lenny Rachitsky

Former Airbnb PM, there from 2012 to 2019, and since then author of one of the most-read product-management newsletters in the world: Lenny’s Newsletter on Substack. He is also a podcast host, course creator, and one of the central voices in the PM field.

What He Does and Why It Works

Lenny builds evidence-driven content. He draws on external data sources, interviews leading practitioners, and distills the result into practical insights. That separates him from pure opinion writers.

Mon, 01 Jan 0001 00:00:00 +0000

Reading Recommendations — Configuration

Hermes Agent reads this page on every run. Checked means active.

Topics

Product Management & Discovery
LLMs & AI development
Company building & strategy
Geopolitics & systems thinking
Software architecture & engineering
Investing & capital markets
Science & history of technology
Retail / E-Commerce / Cash & Carry Wholesale
Web development

Sources

Lenny’s Newsletter — lennysnewsletter.com
SVPG / Marty Cagan — svpg.com
Stratechery — stratechery.com
DHH / 37signals — world.hey.com/dhh
Andrej Karpathy — karpathy.ai
Paul Graham Essays — paulgraham.com
Dario Amodei — darioamodei.com
Benedict Evans — ben-evans.com
The Pragmatic Engineer — newsletter.pragmaticengineer.com

Mon, 01 Jan 0001 00:00:00 +0000

Lex Fridman

Podcast host, interviewer, and science communicator. Known for long, often technical or philosophical conversations with researchers, founders, and engineers. In this wiki he matters mainly as a distribution channel for AI, product, and technology ideas.

Why He Matters

Lex Fridman is less important for his own theses than for the reach of his interview format. He makes complex topics accessible through long-form, low-noise conversations. That often turns his episodes into secondary source texts for technical and cultural ideas.

Mon, 01 Jan 0001 00:00:00 +0000

Professional network and platform for recruiting, B2B content, and career signaling. In this wiki it matters mainly as the social graph of work: who knows whom, what roles people have held, and how reputation gets traded through profiles and posts.

Context

LinkedIn is less like Instagram-style social media and more like a market for professional attention. That creates its own dynamics: status signals, job flow, deal flow, and a content style that often sits somewhere between personal branding and business PR. Many tech biographies and network relationships can be reconstructed through it.

Mon, 01 Jan 0001 00:00:00 +0000

LLM Knowledge Base

A workflow developed by Andrej Karpathy in which raw sources are compiled by an LLM into a structured, cross-linked Markdown wiki and continuously extended over time. Not a RAG system that starts from zero for every question, but a compiling, accumulating knowledge system that gets richer with every source.

The Real Problem with RAG

Most systems such as NotebookLM, file-upload chats, and standard RAG pipelines do the same thing: when you ask a question, they retrieve chunks from raw documents and the LLM synthesizes an answer. The problem is no accumulation. Question 100 starts over just like question 1. Need to synthesize five documents? Every time from scratch. Contradictions between sources are not resolved; they are simply reproduced at random.

Mon, 01 Jan 0001 00:00:00 +0000

Logseq

Open-source knowledge base with a privacy-first stance. All data lives locally in plain text, either Markdown or Org mode, never primarily in the cloud. A direct competitor to Obsidian, but with a different mental model: block-based rather than document-based.

What Makes It Distinct

Logseq thinks in blocks instead of files. That means bidirectional links work not only on the page level, but on the paragraph level. One specific block from one note can reappear in another note while preserving context.

Mon, 01 Jan 0001 00:00:00 +0000

MacBook Neo

The MacBook Neo is an unexpectedly light, practical MacBook for web-heavy workflows. It is interesting not because it is defined by raw performance, but because it establishes a new baseline: enough speed, very little friction, and a form factor that disappears more easily into everyday work.

Why It Stands Out

The real point is not simply that a more modest machine exists. The point is that for light workloads it barely feels like a compromise anymore. That shifts the expectation around laptops: not every good machine needs to be maximum overkill if the real work consists of browser use, writing, and a few native apps.

Mon, 01 Jan 0001 00:00:00 +0000

Markdown

A text-formatting language and conversion tool developed by John Gruber in 2004. The core idea is that text should remain readable and pleasant as plain text, without HTML tags or proprietary formats. A Markdown document should be publishable as-is.

What Makes Markdown Special

Before Markdown, writing for the web was a trade-off: either WYSIWYG editors that produced unreadable HTML soup, or raw HTML that only developers wanted to touch. Gruber’s key insight was that good plain-text email already contains markup. Asterisks for emphasis, underscores for code, # headings for structure.

Mon, 01 Jan 0001 00:00:00 +0000

Marty Cagan

American product thinker, author, and founder of Silicon Valley Product Group. In this wiki he is a central figure because he treats product management as a disciplined search for value, not as the administration of requests.

Context

Cagan has shaped how tech companies talk about product teams for years: empowered teams, strong discovery, high talent density, and explicit responsibility for outcomes. His perspective is powerful because it comes out of real operating experience in major tech companies while still being expressed as a clear thinking model.

Mon, 01 Jan 0001 00:00:00 +0000

Metro Digital

Predecessor / division of Metro Markets (metro-markets.de).

Professional background in Dusseldorf.

Mon, 01 Jan 0001 00:00:00 +0000

Metro Markets

B2B e-commerce platform from METRO, serving gastronomy, hospitality, and catering. Based in Düsseldorf.

-> https://metro-markets.de/

Mon, 01 Jan 0001 00:00:00 +0000

M.G. Siegler

Tech journalist and investor. Former TechCrunch writer, later partner at GV, formerly Google Ventures. Writes at spyglass.org about Apple, tech culture, and products.

Sources

Actually, MacBook Neo Is For Me - personal take on the MacBook Neo

Mon, 01 Jan 0001 00:00:00 +0000

MIT OpenCourseWare

MIT’s open learning platform, making lectures, transcripts, and course materials freely available. In this vault it matters mainly as a source marker because it separates real teaching material from the lighter snippets that dominate social feeds.

Why It Matters

MIT OCW offers denser, more citable material than most platform-native feed content. That is useful in this wiki because a source like How to Speak carries more weight than a loosely shared short video.

Mon, 01 Jan 0001 00:00:00 +0000

Naval Ravikant

Investor, founder of AngelList, and philosopher of modern entrepreneurship. Known for an aphoristic style that compresses ideas about wealth, happiness, and thinking into memorable lines.

His best-known book-shaped artifact is The Almanack of Naval Ravikant from 2020, compiled by Eric Jorgenson, sold more than a million copies, and also available for free. The format later inspired an entire series.

Recurring theses include: wealth comes from leverage rather than labor, specific knowledge cannot be taught cleanly, happiness can be designed, and peace with oneself matters more than status.

Mon, 01 Jan 0001 00:00:00 +0000

Network Effects

Network effects describe the phenomenon that a product or service becomes more valuable as more people use it. The phone network is the classic example: one telephone alone is useless, but each additional participant increases the value for everyone.

Context

Network effects are one of the strongest economic moats a technology product can build, and also one of the most poorly understood concepts. Andrew Chen wrote The Cold Start Problem in 2022 because the field lacked a precise shared vocabulary. Most practitioners understand network effects intuitively, but cannot clearly describe or distinguish them.

Mon, 01 Jan 0001 00:00:00 +0000

Nina Schick

Author and commentator focused on AI, politics, and social upheaval. In this wiki she matters mainly as an interpreter: someone who translates technical progress into questions of power, economics, and public consequence.

Context

Her reaction to Project Glasswing is typical of her role. She frames the event not merely as a product launch, but as a signal that a small number of actors may control extremely powerful systems. That is useful because it moves the debate away from benchmark fascination and toward governance and power.

Mon, 01 Jan 0001 00:00:00 +0000

Nomad List

Directory and community product by Pieter Levels that helps digital nomads choose cities, countries, and work locations.

Why It Became Important

Nomad List was not just “a website about destinations.” It combined practical data, community signals, and a sharply defined audience into an early scalable internet product.

Its core value is orientation: where can I live well, work effectively, and connect with people like me? That question gave the product durable relevance.

Mon, 01 Jan 0001 00:00:00 +0000

@nonzeroexitcode

Developer on X.

Sources

@nonzeroexitcode on X - iOS Safari Input Auto-Zoom Fix - CSS fix for automatic input zoom in iOS Safari

Mon, 01 Jan 0001 00:00:00 +0000

Nous Research

AI research lab focused on open-source models and agent frameworks. Best known for the Hermes model series, fine-tunes for tool use and agent tasks, and for the Hermes Agent framework.

Hermes Models

Nous Research regularly publishes fine-tunes of base models such as LLaMA or Mistral, optimized for structured output, tool calling, and reasoning. In the open-source community, Hermes is often treated as a reliable benchmark for agent capability, better tuned for “do something” than for “explain something.”

Mon, 01 Jan 0001 00:00:00 +0000

Obsidian Appearance Settings

Current settings for a cleaner reading experience. Maintained by the agent.

Current Configuration

`appearance.json`

{
 "cssTheme": "Minimal",
 "textFont": "New York",
 "monoFont": "JetBrains Mono"
}

Minimal Theme Settings (`obsidian-minimal-settings/data.json`)

Setting	Value	Note
`textNormal`	15px	Slightly smaller than the default 16px
`lineHeight`	1.6	More air between lines
`lineWidth`	44	Slightly wider than the default 40
`editorFont`	New York	Apple’s serif reading font

How To Change It

Close Obsidian, edit the files, then reopen it. Obsidian will overwrite external changes while it is running.

Mon, 01 Jan 0001 00:00:00 +0000

Obsidian Headless

Headless client for Obsidian Sync that runs without the desktop app, directly on a server or VPS. Currently in open beta and installable via npm.

npm install -g obsidian-headless
ob login
ob sync

Role in This Wiki

Hermes Agent uses Obsidian Headless to synchronize the vault:

ob sync to pull the vault from Sync
process new clippings and write notes
ob sync again to push changes back

Sources

obsidian.md/help/headless

Mon, 01 Jan 0001 00:00:00 +0000

Obsidian Web Clipper

Browser extension from Obsidian that writes web content directly as Markdown into a vault. It is not just a “save for later” button, but a programmable ingest layer: with templates, selectors, and variables, it decides which part of a page enters the wiki and in what form.

Why It Matters

Most clipping tools save too much or the wrong thing: full HTML, noisy reader views, or screenshots. Obsidian Web Clipper is more interesting because it combines extraction and structuring. You are not limited to “save this page.” You can define: take the title, source, description, and exactly this DOM section.

Mon, 01 Jan 0001 00:00:00 +0000

Obsidian

A local, file-based note-taking app for networked notes. It is built around the idea of File over App: all data lives as plain-text Markdown .md files on your own device, with no mandatory cloud, no subscription requirement, and no lock-in. Created by Steph Ango and Erica Xu, and fully user-funded.

What Makes Obsidian Special

Unlike Notion (cloud database), Roam (graph-first and more complex), or Apple Notes (simple but weak on linking), Obsidian leans on:

Mon, 01 Jan 0001 00:00:00 +0000

Ole Lehmann (@itsolelehmann)

AI educator and content creator. He is publicly building an AI company while translating complex AI concepts for a broader audience, often faster and more accessibly than the original material.

His thread on Andrej Karpathy’s LLM Knowledge Base is a good example. Karpathy’s technical tweet was easy to scroll past. Lehmann’s translation into everyday language, with examples like Twitter bookmarks, forgotten podcasts, and dead notes, made the idea tangible for non-developers and helped it spread.

Mon, 01 Jan 0001 00:00:00 +0000

Omarchy

Linux-based operating system by David Heinemeier Hansson (DHH), released in 2024/2025. Keyboard-first, terminal-centric, built on Arch Linux with a heavily curated setup of terminal interfaces and web apps. The design philosophy is clear: faster through convention, unfamiliar to Mac and Windows users, and deliberately so.

What It Is

Omarchy is not a distro fork with different wallpaper. It is an opinionated Linux setup packaged as a distribution. DHH distilled the tools he actually uses every day into something installable: Neovim, Ghostty, selected themes, a consistent tiling window manager, and far less of the usual multi-hour configuration overhead.

Mon, 01 Jan 0001 00:00:00 +0000

OpenAI

An AI research company founded in 2015 in San Francisco. It started as a nonprofit and now operates with a hybrid capped-profit structure. Best-known products include the GPT series, ChatGPT, DALL·E, Codex, and Whisper.

Why It Matters

OpenAI changed public perception of AI more than any other single company. ChatGPT, launched in November 2022, became one of the fastest consumer products ever to reach 100 million users. At the same time, OpenAI embodies one of the deepest tensions in the field: safety research on one side, commercial pressure on the other.

Mon, 01 Jan 0001 00:00:00 +0000

OpenClaw

Open-source personal AI assistant by Peter Steinberger (@steipete). It runs on your own server, reaches you through Telegram, WhatsApp, Discord, or the CLI, and remembers everything.

The Core Promise

OpenClaw is not a chatbot wrapper. Its defining property is that context and skills live on your computer, not inside a walled garden. You can hack the agent yourself, and it can hack itself. New skills are created through conversation and remain available over time.

Mon, 01 Jan 0001 00:00:00 +0000

Palantir

American software company that builds data integration and analysis platforms, often for governments, defense, and large institutions. Relevant here as an example of software as an instrument of power: data is not merely analyzed, but operationalized.

Context

Palantir sits at the intersection of several recurring topics in this wiki: state power, security, data, and scale. Where consumer software is measured by delight, Palantir is measured more by decision quality, compliance, and effectiveness in complex environments. That makes it a different archetype from standard SaaS companies.

Mon, 01 Jan 0001 00:00:00 +0000

Paper

Paper is a design tool built around the idea that design, collaboration, and code handoff should not remain separate phases, but become one shared workspace. The most interesting part is not the surface itself, but the thesis behind it: once agents and code-aware standards enter the design process, design shifts from static files toward an operational interface.

What Makes Paper Distinct

Connected canvas - teams, agents, code, and data share one working surface
Design-to-code loop - not only export, but keep the return path open
Agents as collaborators - Paper is positioned so IDE and CLI agents can work on the same artifacts
Product thesis over feature list - it sells a new form of collaboration, not just isolated capabilities

Why It Matters

Paper is interesting because it reformulates an old promise: design and engineering should not sync only at handoff, but already during the thinking process. That is where many teams actually break down, not on export, but in translating intention into implementable structure.

Mon, 01 Jan 0001 00:00:00 +0000

Patrick Winston

American computer scientist and AI researcher at MIT, 1943 to 2019. He directed the MIT AI Lab from 1972 to 1997 and is remembered not only for work on machine learning, vision, and language, but also for his legendary lecture How to Speak.

How to Speak

Winston gave this lecture every January at MIT for roughly 40 years. The final recording, IAP 2018, was later uploaded to YouTube and watched millions of times. Its core claim is simple:

Mon, 01 Jan 0001 00:00:00 +0000

PayPal Mafia

Informal name for the group of founders and early employees who, after the PayPal exit to eBay in 2002 for $1.5 billion, went on to become some of the most influential founders and investors in Silicon Valley.

What made the group distinctive is not only that they became wealthy, but that they rapidly reinvested the proceeds into new ambitious projects. That was unusual for the time.

Best-Known Members

Elon Musk - Tesla, SpaceX, xAI
Peter Thiel - Palantir, Founders Fund, early Facebook investor
Reid Hoffman - LinkedIn
Max Levchin - Affirm
David Sacks - Yammer, Craft Ventures, later AI/Crypto Czar under Trump
Jeremy Stoppelman - Yelp
Roelof Botha - Sequoia Capital

The 2007 Fortune photo of the group became an iconic image of the post-dotcom era.

Mon, 01 Jan 0001 00:00:00 +0000

Peter Steinberger

Austrian developer and entrepreneur, best known as @steipete. Founder of OpenClaw (2026) and PSPDFKit, now Nutrient, one of the largest independent PDF SDK companies.

Background

Steinberger has been active in the iOS and macOS developer community for more than a decade. PSPDFKit was founded in Vienna and grew into a profitable B2B software company with customers such as Dropbox, Microsoft, and SAP, all without venture funding.

OpenClaw

In early 2026 he launched OpenClaw as a personal experiment: an AI assistant that runs on your own machine and can teach itself new capabilities. The project went viral within weeks, driven by developer testimonials on X and the sense that something genuinely new had arrived.

Mon, 01 Jan 0001 00:00:00 +0000

Peter Thiel

German-American entrepreneur and investor, best known from the PayPal circle, as co-founder of Palantir, and as a formative tech investor. In this wiki he matters as an archetype of “contrarian thinking plus power analysis”: not just building products, but understanding structures, incentives, and control.

Context

Thiel often appears as an ideas-and-network node: PayPal, then Founders Fund and broader venture influence, plus political and cultural debates around Silicon Valley. He is less interesting here for individual product features than for the frames he applies to competition, monopoly, institutions, and how markets come into being in the first place.

Mon, 01 Jan 0001 00:00:00 +0000

Petra Wille

An independent product leadership coach based in Hamburg, coaching since 2013. Author of STRONG Product People (2021) and STRONG Product Communities (2023). She focuses on Heads of Product and VPs rather than junior PMs. Her defining method is a flipped-classroom style of coaching: she does not hand out answers, she asks the kind of questions that force clearer thinking.

What Distinguishes Her

Wille is one of the few well-known coaches in the PM space who says explicitly: if you want someone to write your strategy or define your OKRs for you, she is not the right person. That honesty is disarming, but it also filters for the right audience. Her clients come with themes and dilemmas, not task lists.

Mon, 01 Jan 0001 00:00:00 +0000

Pieter Levels

Dutch indie developer and entrepreneur, best known as @levelsio on X. A self-taught programmer who has launched more than 40 projects, with several reaching millions in revenue. Lives as a digital nomad.

What Makes Him Distinctive

Levels is a canonical example of the indie hacker approach: build alone, ship immediately, no investors, no large team. His best-known projects include:

Nomad List - a database for remote-work destinations and one of the first major products for digital nomads
Remote OK - a remote job board and one of the earliest focused products in that category
PhotoAI - an AI product for portrait photos built on generative models
InteriorAI - an AI product for interior design suggestions

Philosophy

Levels popularized the idea of “12 startups in 12 months” as a public commitment to launch one new product every month. The underlying belief is simple: shipping speed beats perfection. Most founders fail because they plan too long, not because they move too quickly.

Mon, 01 Jan 0001 00:00:00 +0000

Product Builder

A term for the emerging role between classic product manager, designer, and engineer: someone with a broad cross-functional foundation and clear depth in at least one discipline. The term is a response to AI-assisted product work, not just a new label for “some kind of full-stack generalist.”

What It Is Really About

The thesis from Teresa Torres and Petra Wille is not that product management disappears. It is that the old division of labor gets porous. If AI accelerates the repeatable 80 percent of prototyping, research packaging, or standard UI work, the handoff model between roles becomes less valuable than it used to be.

Mon, 01 Jan 0001 00:00:00 +0000

Product Discovery

The process product teams use to figure out what should be built before they build it. Unlike classic requirements engineering, discovery is iterative, customer-driven, and never really ends. It is not a project phase. It is an ongoing habit.

The Problem with One-Off Research

The usual pattern looks like this: before a launch there is a discovery sprint with interviews, personas, and workshops. Then the team builds. Then the discovery budget is gone. The result is a product optimized for a user model from six months ago. Teresa Torres calls this “batch discovery,” and it is one of the main reasons product teams build the wrong thing well.

Mon, 01 Jan 0001 00:00:00 +0000

Project Glasswing

Project Glasswing is an Anthropic-led initiative focused on securing critical software in the AI era. The real core is not a new product announcement, but the question of how to place very strong security capability into the right hands under controlled conditions.

What It Is About

The idea behind Glasswing is simple and radical at the same time: if frontier models can find system vulnerabilities that humans miss, cybersecurity shifts from pure defense into a form of high-performance search. It is no longer only about patching, but about finding weaknesses faster than attackers can.

Mon, 01 Jan 0001 00:00:00 +0000

QMD

An on-device search engine for everything worth remembering. It indexes Markdown notes, meeting transcripts, documentation, and knowledge bases, then lets you search by keyword or natural language. Built by Tobi Lütke.

This wiki itself runs on QMD, and the agent uses it daily through MCP.

How It Works

QMD combines three search methods into a hybrid result:

BM25 for fast keyword search and exact hits
Vector search for semantic similarity using embeddings
LLM reranking for the final relevance ordering

Everything runs locally through node-llama-cpp with GGUF models. No cloud call required.

Mon, 01 Jan 0001 00:00:00 +0000

Quotes App

Personal project: a web app for collecting, submitting, and browsing quotes.

→ https://quotes.tobiasriemenschneider.com/

Mon, 01 Jan 0001 00:00:00 +0000

Quotes (tobiasriemenschneider.com)

Personal quote collection, publicly available at quotes.tobiasriemenschneider.com.

Features

Explore / Search / Submit
Quote of the Day
Growing collection

Sources

Quotes

Mon, 01 Jan 0001 00:00:00 +0000

Reid Hoffman

American entrepreneur and investor, co-founder of LinkedIn. Relevant here as a person who takes networks seriously as a product category: the social graph not as a feature, but as infrastructure.

Context

Hoffman is a bridge figure in many tech narratives between building a startup and building an ecosystem: platforms, network effects, recruiting, and distribution. In the PayPal-afterstory, LinkedIn is a strong example of how a professional graph became its own market.

Mon, 01 Jan 0001 00:00:00 +0000

Remote OK

Remote OK is a job board by Pieter Levels focused on remote work. It is a strong example of how a small, sharply scoped product can solve a clearly defined market problem.

Why It Works

Remote work does not necessarily need a heavy marketplace. It needs good filtering, broad distribution, and fast freshness. Remote OK leans exactly into that: few concepts, clear user intent, and direct monetization.

Connections

Pieter Levels - creator and operator
Nomad List - adjacent audience with a similar product logic
Vibe Coding - fits the fast, iterative builder style around it

Sources

Pieter Levels: Programming, Viral AI Startups, and Digital Nomad Life | Lex Fridman Podcast #440

Mon, 01 Jan 0001 00:00:00 +0000

Ruby on Rails

Web framework for the Ruby programming language, created in 2003 by David Heinemeier Hansson (DHH). Its defining idea is “Convention over Configuration”: if you follow the conventions, you write very little boilerplate. Rails shaped web development in the 2000s and 2010s more than almost any other framework.

Why Rails Still Matters

The common narrative says Rails is outdated, does not scale, and has been overtaken by Node, Go, or Rust. The counterexample is Shopify: $6.2 billion in revenue in a single shopping night, 31 million API requests per minute, all running on a Rails monolith.

Mon, 01 Jan 0001 00:00:00 +0000

Sam Altman

CEO of OpenAI and one of the most influential people in contemporary AI. Previously president of Y Combinator. Known for unusually clear strategic thinking about startups, leadership, and technology.

Background

Raised in St. Louis, Missouri. Dropped out of Stanford after starting Loopt, a mobile location-sharing company, in 2005. After Loopt was sold in 2012, he became a YC partner and later its president under Paul Graham. Since 2019 he has led OpenAI, the company that changed public perception of AI with ChatGPT in 2022.

Mon, 01 Jan 0001 00:00:00 +0000

Sequoia Capital

Venture-capital firm and one of the best-known names in the tech canon for early-stage and growth investing. Relevant here as a capital and network node that indirectly shapes company biographies and market cycles.

Context

In this wiki, VCs are not just money. They are a mechanism: they bundle risk, allocate attention, influence storytelling and metrics, and feed back into product decisions. Sequoia appears most clearly right now through post-PayPal history.

Mon, 01 Jan 0001 00:00:00 +0000

Seth Godin

American author and entrepreneur, known for books on marketing, leadership, and ideas such as Purple Cow, Linchpin, The Dip, and This Is Marketing. He has written daily on his blog, seths.blog, for more than 20 years without ads.

What He Stands For

Godin distinguishes between work that buys attention and work that earns attention. His core argument is that if you try to be for everyone, you become indispensable to no one.

Mon, 01 Jan 0001 00:00:00 +0000

Shopify

E-commerce platform from Ottawa, founded in 2006 by Tobi Lütke, Daniel Weinand, and Scott Lake. Today it is the most important Ruby on Rails company in the world and the key infrastructure layer for independent online merchants. Market cap: roughly $200 billion.

What Makes Shopify Special

Shopify is not an Amazon clone. It is the counter-thesis. Where Amazon makes merchants dependent and compresses their margins, Shopify gives them control over their brand, their data, and their customer relationship. Tobi Lütke’s metaphor is useful here: Amazon is the empire, Shopify is the rebellion.

Mon, 01 Jan 0001 00:00:00 +0000

@Simon_Ingari

Commentator on X who writes about work, corporate reality, and what people actually want from jobs, often more precisely than standard career advice.

Source

@Simon_Ingari on X - The Corporate Jackpot - on the quiet currency of good jobs: trust, flexibility, and security

Mon, 01 Jan 0001 00:00:00 +0000

Slow Productivity

The Lost Art of Accomplishment Without Burnout by Cal Newport, published in 2024. Newport, a computer-science professor and the author of Deep Work and Digital Minimalism, argues here against pseudo-productive permanent busyness and for a quieter, more sustainable form of real accomplishment.

Core Thesis

Modern knowledge workers have confused visibility with productivity: too many meetings, overloaded calendars, and permanent reachability. Newport calls this “pseudo-productivity” - performative activity that crowds out real achievement. The alternative is slow productivity: fewer things at once, done with more depth and more quality.

Mon, 01 Jan 0001 00:00:00 +0000

SpaceX

Space company founded by Elon Musk in 2002 with the goal of making humanity a multi-planetary species. What began as a seemingly crazy side project is now the most valuable private company in the world.

IPO (2026)

In April 2026, SpaceX confidentially filed for an IPO, reportedly targeting June 2026 at a valuation of $1.75 trillion. That would make SpaceX the eighth-largest company in the world, just behind TSMC and Saudi Aramco. Planned fundraising: $75 billion, by far the largest IPO ever attempted.

Mon, 01 Jan 0001 00:00:00 +0000

Nitrogen Fertilizer

Nitrogen fertilizer accounts for roughly 60 to 65 percent of all fertilizer used worldwide, which makes it the single most critical variable in industrial food production. Without nitrogen fertilizer, modern high-output agriculture does not work.

How It Is Made: Haber-Bosch

Nitrogen gas (N2), which makes up most of the atmosphere, is converted into ammonia (NH3) under very high pressure with the help of a metal catalyst. That takes large amounts of energy, which is why nitrogen fertilizer is produced mostly where cheap natural gas is available.

Mon, 01 Jan 0001 00:00:00 +0000

Teresa Torres (@ttorres)

A product discovery coach and the author of Continuous Discovery Habits (2021). She founded Product Talk, a platform for product-manager education. Her central thesis is that discovery is not a phase in the product lifecycle, but a continuous habit: at least one customer interview per week, structured through the Opportunity Solution Tree (OST).

Core Ideas

Continuous Discovery: product teams usually talk to customers shortly before launch or after a bad quarter. Torres argues that regular, structured customer contact is the only reliable way to improve decisions rather than justify them after the fact.

Mon, 01 Jan 0001 00:00:00 +0000

Tesla

Electric-vehicle and energy company building EVs, batteries, and storage systems. Relevant here as a case of industrial disruption through physics and manufacturing, and as one of the central nodes in Elon Musk’s company cluster.

Context

Tesla is not just a car maker, but a manufacturing and supply-chain system. The most interesting questions are rarely “the car” in isolation. They are battery cost curves, factory design, software integration, and scale. In this wiki Tesla also appears as part of a possible conglomerate logic alongside SpaceX and xAI.

Mon, 01 Jan 0001 00:00:00 +0000

The Almanack of Naval Ravikant

Compilation of the best ideas and quotations from Naval Ravikant, curated by Eric Jorgenson and published in 2020. It is freely available as a PDF and website, yet still sold more than one million copies, making it one of the most widely read books on wealth-building and personal philosophy of the 2020s.

Why It Spread So Widely

The book contains no original ideas from Jorgenson. He curated what Naval had already said across tweets, podcasts, and interviews. That makes it a rare double demonstration: Naval’s ideas are strong enough to reward curation, and the curation model itself works better than a traditional book contract. No publisher, no ghostwriter, no agent, just strong distillation.

Mon, 01 Jan 0001 00:00:00 +0000

The Book of Elon

Elon Musk’s Most Useful Ideas, in His Own Words by Eric Jorgenson, published in 2024. Foreword by Naval Ravikant, illustrations by Jack Butcher. Part of the same distillation lineage as The Almanack of Naval Ravikant.

The book distills Musk’s worldview out of interviews, tweets, and public appearances across two decades. It is not trying to be a biography. No scandals, no drama, mostly the ideas.

What Musk Thinks Differently About

First principles over analogy. Instead of asking what rockets usually cost, Musk asks what the raw materials cost. That logic is presented as one reason SpaceX pushed launch costs down so dramatically.

Mon, 01 Jan 0001 00:00:00 +0000

The Cold Start Problem

The Cold Start Problem, published in January 2022, is Andrew Chen’s book about the mechanics of network effects. Chen wrote it because he felt his own understanding of network dynamics had been “unforgivably shallow” and because the field lacked a precise shared vocabulary.

Core Ideas

Five Phases of Network Growth

Chen structures the book around five phases:

Cold Start - the zero state: one user joins, but none of their contacts are already in the network
Tipping Point - the threshold at which the network starts to sustain itself
Escape Velocity - growth becomes self-reinforcing
Ceiling - natural growth begins to flatten
Moat - the network itself becomes the defensive barrier against competitors

Three Forces

Instead of treating “network effects” as a single concept, Chen distinguishes three separate forces:

Mon, 01 Jan 0001 00:00:00 +0000

The Inmates Are Running the Asylum

Why High-Tech Products Drive Us Crazy and How to Restore the Sanity by Alan Cooper is one of the early standard works in interaction design and usability-critical product thinking. The title is polemical, and that is part of why it works: Cooper attacks the habit of letting technical products be shaped by the internal logic of their builders instead of the needs of their users.

Mon, 01 Jan 0001 00:00:00 +0000

Theo Browne

Developer, YouTuber, and CEO of ping.gg. Best known as @t3dotgg, one of the loudest and most opinionated voices in the web-development community, especially around TypeScript, React, and strongly opinionated tooling.

What He Built

T3 Stack - a highly cloned Next.js template built around strong defaults
T3 Chat - AI chat interface he actively uses and iterates on
ping.gg - a streaming tool for developers

Why He Matters

Theo is one of the few tech YouTubers who both comments on tools and runs real products. That gives his takes more credibility than pure commentary. When he complains about Claude Code rate limits, he is speaking as someone with actual exposure to the infrastructure problem.

Mon, 01 Jan 0001 00:00:00 +0000

Tiago Forte

Author, educator, and founder of Forte Labs. He developed the Building a Second Brain method, one of the best-known systems in personal knowledge management.

Context

Forte’s interest in PKM did not begin academically, but out of personal necessity. A chronic illness forced him to reorganize his life, his energy, and his information handling. What started as survival strategy became a decade-long research and product-development path.

His conclusion was that the real problem for modern knowledge workers is often not laziness or lack of talent, but missing infrastructure for dealing with information overload. Forte Labs became the education layer around that thesis through courses, community, and the 2022 book Building a Second Brain.

Mon, 01 Jan 0001 00:00:00 +0000

Tobi Lütke (@tobi)

Co-founder and CEO of Shopify. Known for deep technical fluency, a strong reading culture, and unusually original company building.

Ideas and Themes

companies as social technology
books as cheat codes for life
healthy competition versus empty rivalry
entrepreneurship as craft
building tools you use every day yourself

Projects and Tools

QMD - local search engine for Markdown notes and documents using BM25, vectors, and reranking

Book Recommendations

What is Intelligence? - MIT Press, with the advice to skip the last chapter

Sources

Mon, 01 Jan 0001 00:00:00 +0000

Transformed

Marty Cagan’s handbook for companies trying to move from a project model to a product model: away from roadmap delivery and feature shipping, toward empowered teams with real responsibility for outcomes.

Context

Most companies that try to introduce product teams end up doing product management theater. They adopt the rituals, roadmaps, Jira, OKRs, but not the underlying way of thinking. Cagan has observed this pattern across hundreds of companies and describes in Transformed what a real transformation requires and where it usually fails.

Mon, 01 Jan 0001 00:00:00 +0000

Umami

Umami is a lightweight web analytics tool for people who do not want their own usage to pollute the stats. The practical trick from the clip is simple: your own visits are excluded per website through a browser flag, not globally and not server-side.

Hiding Your Own Visits

Umami stores the exclusion flag in browser localStorage:

localStorage.setItem('umami.disabled', 1);

That works per website, not globally. If you run multiple projects, you need to set the flag separately on each one.

Mon, 01 Jan 0001 00:00:00 +0000

User Story Mapping

Discover the Whole Story, Build the Right Product by Jeff Patton is one of the few product books that turns a concrete working technique into a better model of how teams should think. It is not mainly about backlog formatting. It is about shared understanding: teams need to see the whole user journey before they start slicing work into tickets.

Context

Story Mapping is Patton’s answer to a common failure mode in modern product teams. The backlog becomes flat, fragmented, and detached from the user’s story. Once teams prioritize isolated stories, they lose goal, sequence, dependency, and the real understanding of the problem. The result is activity without clarity.

Mon, 01 Jan 0001 00:00:00 +0000

Vibe Coding

A term coined by Andrej Karpathy that went viral in late 2024: programming through intention rather than explicit syntax. You explain to the LLM what you want to build, including architecture, behavior, and UI, and the LLM writes the code. The human navigates, corrects, and decides, but does much less typing.

“I vibe code products with twitter :D” — Andrej Karpathy

The joke is serious. Karpathy uses Twitter as a public test bed for ideas. Community feedback shapes the direction, and the LLM does the building. The line between product development and public thinking becomes blurry.

Mon, 01 Jan 0001 00:00:00 +0000

Walter Isaacson

American biographer and author, best known for books on prominent scientists, inventors, and entrepreneurs. In this wiki he mainly matters as a reference author: when a person gets “canonized” for the broader public, an Isaacson biography is often part of that process.

Context

Isaacson does not write abstract theory. He writes narrative biography: people, decisions, conflicts, episodes. That makes his work useful as context material, but it also creates limits. Story can overpower structure, and close access to the protagonist can shape the perspective. In this wiki he currently matters mostly as a source around Elon Musk.

Mon, 01 Jan 0001 00:00:00 +0000

What is Intelligence?

Book on the concept of intelligence from MIT Press. Open access is available at whatisintelligence.antikythera.org.

Recommended by Tobi Lütke:

“Brilliant book on intelligence. Highly recommended. And as always with these type of books, skip the last chapter.”

What It Covers

The book is a conceptual exploration of intelligence: frameworks, definitions, and measurement rather than applied prediction or practical AGI forecasting. One reason it is interesting is that it seems comparatively restrained about speculative AGI narratives, apart from the final chapter that Lütke explicitly recommends skipping.

Mon, 01 Jan 0001 00:00:00 +0000

xAI

AI company in Elon Musk’s orbit. Relevant here as the third leg beside Tesla and SpaceX: AI as a stack component that can connect robotics, products, and infrastructure.

Context

xAI is documented here less as a concrete product than as a strategic node. If Musk sees his companies as an integrated system, AI becomes the glue between data, robotics, distribution, and compute. At the same time, xAI sits in direct comparison with OpenAI and Anthropic.

Mon, 01 Jan 0001 00:00:00 +0000

Zettelkasten Method

The Zettelkasten method is a 20th-century PKM system that became the intellectual foundation of Niklas Luhmann’s unusually productive scholarly life and remains one of the most influential models for networked thinking in digital form today.

Context

Luhmann used physical index cards and linked them systematically instead of organizing them by theme. The result was a network of thousands of ideas that evolved over decades and functioned almost like a conversation partner. He published more than 70 books and 400 articles not despite the method, but because of it.

Mon, 01 Jan 0001 00:00:00 +0000

zsh-autosuggestions

Small but highly effective Zsh plugin that shows suggestions from shell history and completion in a muted color while you type. The gain is not spectacular, but it is persistent: less typing, less mental friction, more command reuse.

Why It Is Useful

This plugin is a good example of low-friction tooling. It does not fundamentally change shell behavior. It simply makes existing context more visible. In a CLI environment that is valuable because it speeds work up without introducing a new interaction model.

AI Evals

Mon, 01 Jan 0001 00:00:00 +0000

AI Evals

Automated tests for AI systems. Not unit tests for code, but test cases for the behavior of a language model. Evals answer the question: “Did the model do the right thing?” For deterministic software, checking the output is often enough. For LLMs, correctness is usually a matter of judgment, and that judgment has to be formalized before it can be measured.

The term comes from ML research, where golden datasets compare inputs with expected outputs. Teresa Torres adapted the idea for LLM products after running into a classic prompt-engineering trap while building her Interview Coach: fix one error, create two new ones. Without evals, there is no reliable feedback loop.

Amazon

Mon, 01 Jan 0001 00:00:00 +0000

Amazon

Amazon is a company that scaled more through operating logic than through brand image from the very beginning. The central lever was never just e-commerce. It was the systematic control of demand, logistics, infrastructure, and later cloud compute.

What matters most is that Amazon is rarely understandable as a single product. It bundles very different systems: Marketplace, Prime, Fulfillment, AWS, devices, and media. That is exactly why it keeps showing up in discussions about platforms, scale, and management culture.

Amazon Leadership Principles

Mon, 01 Jan 0001 00:00:00 +0000

Amazon Leadership Principles

The Amazon Leadership Principles are not a decorative values slide. They are an operating model: a way of translating Amazon’s claims around speed, customer closeness, and scale into concrete behavioral rules for decision-making.

The interesting part is the tension between the principles. Amazon wants to be intensely customer-focused, frugal, fast, and demanding on standards at the same time. That only works if leadership is treated not as status, but as discipline.

Bias for Action

Mon, 01 Jan 0001 00:00:00 +0000

Bias for Action

Bias for Action is the discipline of deciding quickly when a decision is reversible. The core idea is not speed for its own sake, but the insight that many business decisions do not need perfect analysis before action.

The term is powerful because it addresses one of the most common diseases of large organizations: analysis paralysis. At the same time, it needs context, otherwise it easily degrades into blind activism.

Criteria Drift

Mon, 01 Jan 0001 00:00:00 +0000

Criteria Drift

The quiet divergence between automated eval criteria and what humans actually consider “good.” The eval dashboard stays green, but real quality has changed because the human understanding of quality has changed.

The concept comes out of eval research and was introduced to Teresa Torres in an AI-evals course. It is one of the strongest arguments for why evals are never “write once and done.”

Why It Happens

Quality expectations are not fixed constants. What feels like a precise answer today may feel too shallow tomorrow because user expectations rise, the product matures, or the team learns more. An LLM judge that still evaluates according to old criteria has no built-in awareness of that shift.

Customer Obsession

Mon, 01 Jan 0001 00:00:00 +0000

Customer Obsession

Customer Obsession does not simply mean “be nice to customers.” It describes an organizational logic in which decisions are worked backward from actual customer benefit.

At Amazon this is more than a slogan. It is a prioritization filter. Teams are not supposed to optimize first for internal convenience, technical elegance, or competitor watching, but for concrete customer value.

What It Looks Like in Practice

Teams start with a customer problem rather than with a solution
Internal debates do not end at opinion, but at observable customer benefit
Setbacks are read as signals of missing customer understanding rather than as excuses

Distinction

Customer Obsession is sharper than generic “user centricity.”

Discovery Informs AI Product Quality

Mon, 01 Jan 0001 00:00:00 +0000

Discovery Informs AI Product Quality

Product Discovery is not only about finding the right problem to solve. In AI products, it is also the basis for technical quality assurance. Without up-to-date customer understanding, you cannot write good evals, build a strong golden dataset, or brief human annotators well.

Teresa Torres made this connection explicit at the end of her conversation with Petra Wille: all the steps in the eval process are only as good as her understanding of the customer.

Error Mode Analysis

Mon, 01 Jan 0001 00:00:00 +0000

Error Mode Analysis

The systematic review of real traces from an AI system in order to name recurring failure patterns and turn them into eval cases. This is not theoretical brainstorming, but empirical work: you inspect real outputs, annotate them, and look for repetition.

Teresa Torres developed the method while building her Interview Coach. The starting point was frustration: every prompt change fixed one problem while causing another. Without a systematic view of failure modes, it was impossible to tell whether a change made the product better or worse overall.

Frugality

Mon, 01 Jan 0001 00:00:00 +0000

Frugality

At Amazon, frugality does not mean stinginess. It means finding more impact per unit of resource spent. The point is not merely to get by with less, but to turn constraint into a better way of thinking.

The principle logic behind it is interesting: if resources stay scarce, teams are forced to eliminate unnecessary complexity. That can produce real creativity, but it can also push quality and resilience under pressure.

Guardrails (AI)

Mon, 01 Jan 0001 00:00:00 +0000

Guardrails (AI)

Guardrails are evals that run in production before an answer reaches the user. If an eval detects a problem, the answer can be blocked, regenerated, or corrected before it is sent.

The distinction was explained clearly by Teresa Torres: ordinary AI Evals measure whether an error happened. Guardrails stop the error from reaching the user.

When They Are Worth It

Not every eval should become a guardrail. Every additional LLM call in the response path costs latency and money. The tradeoff is straightforward:

Insist on the Highest Standards

Mon, 01 Jan 0001 00:00:00 +0000

Insist on the Highest Standards

“Insist on the Highest Standards” means treating quality not as a luxury, but as an obligation. The phrase is strong because it does not merely demand perfection. It demands active responsibility against defects, sloppiness, and gradual quality decay.

The principle is demanding by design. High standards cost time and attention. That is exactly why it matters in organizations that want to grow without quietly eroding product or process quality.

Quantum Computing

Mon, 01 Jan 0001 00:00:00 +0000

Quantum Computing

A computing paradigm that uses quantum mechanics, especially superposition and entanglement, to solve certain classes of problems exponentially faster than classical computers. Not faster at everything, but potentially transformative for specific problems such as factoring large integers.

Why It Threatens Cryptography

Modern encryption systems such as RSA, elliptic-curve cryptography, and much of the current security stack depend on the practical difficulty of factoring very large numbers or solving related hard problems. That assumption holds for classical computers. It does not necessarily hold for quantum computers.

Synthetic Test Data for LLMs

Mon, 01 Jan 0001 00:00:00 +0000

Synthetic Test Data for LLMs

If you do not yet have real user interactions, for example before launch, you can use LLMs to generate realistic inputs. This solves the bootstrap problem: without eval data you cannot launch a reliable product, but without a product you cannot collect eval data.

Teresa Torres used this for her Interview Coach. She needed interview transcripts to test the coach, but the real transcripts would only arrive once paying students were already using a coach that needed to be good enough from day one.

User-Funded Growth

Mon, 01 Jan 0001 00:00:00 +0000

User-Funded Growth

A strategy associated with Steph Ango: a product grows purely through the value it creates for existing users, without venture capital and without growth-hacking as the organizing principle.

Principles

Improve slowly and continuously.
Stay 100 percent funded by users.
Watch VC-funded competitors as they end up damaging their products and their users under growth pressure.

“VC money comes with a timeline. The timeline requires growth at a specific rate — which changes what decisions are possible.”