There's a mode in startups that is less talked about than founder mode, but much more prevalent. It's called emergency mode. In emergency mode, you and your team are desperately doing whatever you can to solve hair-on-fire problems (say, the product is breaking and customers are

I hear from young people who want to work in tech and build software that it's really hard to land their first job or internship. Here are a few tips that are specific for aspiring young software engineers. First, remember that building software is an open profession. You don't

There's a type of product-market fit that I think of as "vortex in a vertical." Geoffrey Moore (Crossing the Chasm) talks about Vortices. It's when the market rips your product out of you. A vertical, or a narrow sector, is important, because very often all the professionals or

Paul Graham's motto for startups is "Make something people want." That's useful, and a huge improvement over *not* making something people want. But I think of it slightly differently. People want all kinds of things, and don't get them because they cost too much money, time,

On Tuesday, Geoff Hinton was awarded the Nobel Prize in Physics for his work on Restricted Boltzmann Machines, a precursor algorithm for the neural-net revolution we're living now. Many years ago, when my team was implementing Restricted Boltzmann Machines for Java and JVM

There's a type of person you might call a full-stack founder. It's someone who knows customer discovery, product, sales, recruiting, and fund-raising. Basically, they can understand a problem, build what's needed, and marshal the human and financial resources to grow a company

The most dangerous delusions are high-status delusions.

Politics is this amazing pill that makes smart people stupid. Tribal commitments push people who are otherwise able to weigh evidence into wishful thinking. Lots of great reasons to line up with the ingroup. Thought becomes a veneer for signals of belonging.

Startups are an escape room. The clock is the money you raised. If you don't solve all the problems before you reach product-market fit, your company dies. And your first job, just like in an escape room, isn't to find right answers, but to find the right questions.

People ask founders all the time about how they plan to solve some wicked problem. Sometimes the only answer is "Crossing the River by Feeling the Stones" (摸着石头过河), a Chinese saying that conveys how startups are forced to make big decisions with very partial information.

it feels great to help people, but one thing that's true in venture and beyond is that helping too much is a bad sign, because it usually means you're creating a pernicious dependency. what you really want is people who get the maximum mileage per unit of help.

Super excited to be quoted in a New York Times piece on DeepSeek AI, the Chinese team that released two really smart AI models in the last month. nytimes.com/2025/01/23/tec… By open-sourcing models that are near SOTA and trained very cheaply, DeepSeek is showing that China can

The California Assembly spent most of last year trying to devise new punishments for companies that released open-source AI models (and they will try to pass the same regulations this year under a new name. SB 03)-- meanwhile ...

This is almost 100% wrong. There may well be flaws with Stargate, but Arnaud's points are meaningless. 1) AI capabilities will continue to scale with compute. every new algorithmic efficiency a la Deepseek just means you can go further til you hit the current compute/energy

Glad to be quoted by The New York Times about open-source AI today. nytimes.com/2025/01/29/tec… Meta took the lead in open-source AI with its Llama models. They've release other great open-source tools like PyTorch before (PyTorch is a developer tool that helps software

where have i heard this before?

A land speeder straight out of Star Wars