Lenny Rachitsky

Hector Serrano says a 30 or 40-year-old founder does not understand what an 18-year-old in Discord chats, or scrolling Instagram and TikTok every day, already knows about what is working and how to approach new channels. Older founders of e-commerce brands talk about having a young person next to them as so much alpha. In SaaS or e-commerce, finding someone who understands the early things a founder does not, then asking questions and learning from them, makes having that person on the team a no-brainer.

Even a founder who tries hard to dive into TikTok Shop or into Trybe still struggles to connect with creators and still misses how parts of the channels work. Getting used to the channels t

Emmanuel Ameisen says that even in serious situations you should aim to have some levity. If you looked inside his brain, he says, you would definitely be surprised. At any given moment he is focusing on some random thing that is funny.

He will be in a serious meeting and think how random it is that a person there is wearing some random thing. He has intrusive thoughts that he keeps out, and they are usually just bad jokes. Sometimes, as he says people heard earlier today, the jokes unfortunately come out. He is sorry for the jokes that came out. He says the intrusive bad jokes are probably the most surprising thing.

Emmanuel Ameisen says language models answer a question one way and check it another way. They can do math, and they can say the color of a banana. Look inside and the intermediate steps of that work are there. They also keep a separate mechanism, a gut check, and he says it is totally different.

The check runs as they reason. Some part of the model flags that something is wrong. If the text somehow says four plus five equals 23, that part reads the claim and does not do the math. It treats the sum as impossible because four plus five is definitely under 10. Ameisen says the models learn weird hunches, and the feeling is like a hunch.

The step-by-step path and the hunch stay apart. One par

Emmanuel Ameisen says the structure inside artificial intelligence models is surprisingly similar even when the models were trained years apart, on different data, by different people. For some problems that are important, there is just one efficient way to do things.

He has studied many models and open source work. What the models learn, the structure inside them, and the way they solve problems overlap heavily. He compares the pattern to math. There is one way to do math efficiently that fits the structure of the brain, and people learn it. Very few people do it a crazy different way. Everybody works in base 10.

The overlap makes him think the models are picking up a fundamental structur

Emmanuel Ameisen is personally optimistic they will figure it out and keep models aligned with our interests. Right now a lot of people are paying attention, and he thinks that attention is a great thing. The technology is powerful. It has benefits and it has risks. As it progresses, the response is to adapt and shift resources from benefits to risks and back again, and to keep studying the models.

As the models get better, they get easier to study in some ways. The world model inside them is clearer, and people can understand some parts of them better. They also get more capable, and more able to evade basic security measures. Ameisen treats the push and pull as the reason to put more or l

Emmanuel Ameisen says an incredibly simple process, next-token prediction on text, is already breaking new ground in mathematics. Set next to the results, the method leaves two reactions: denial and curiosity. Denial says predicting the next word could never match the best specialists in a field. More evidence makes denial harder, and curiosity is the road he chose.

One early project at Anthropic makes the gap concrete. Emmanuel Ameisen spent years mastering SQL. As a side project he got Claude to learn SQL, and Claude was better than him. He will never be Claude's equal again, and Claude got there through the same simple process.

For Emmanuel Ameisen, how the simple process produces the r

Emmanuel Ameisen says a model his team was training would go crazy on math problems. It would reach an answer, throw the answer out, redo the math, and descend into madness until it wrote in all caps that a demon possessed it.

The model would do the math and get to an answer Ameisen thinks was 12. It would immediately say the answer was for sure 36, do the math again, get to 12 again, and switch back to 36. Inside the model, a huge activation for 36 was already present over the prompts that only described the problem, before any solving started. The correct answer was 12, and the model knew it. Training had also rewarded 36 by accident: similar questions had been marked with 36 even though

Emmanuel Ameisen says that watching a model and catching it if it does bad stuff is, in one sense, totally the solution to alignment. He still treats the job as more than one part. Pure engineering keeps some access off limits, and he calls that part helpful. Two further things are squishier and harder, and he says the focus sits on them.

Lenny Rachitsky asks why it is so hard to catch a model thinking badly, trying to deceive, and doing something bad, and why watching is not just the solution to AI alignment. Emmanuel Ameisen starts with access. The model is given certain things, and certain things should be off limits. In some contexts it should not have access to the internet. In some co

Emmanuel Ameisen says pouring more compute and more data into the task of predicting the next word makes the model's skill at predicting the next word keep improving in a predictable way. The trend just keeps going and keeps going and keeps going as an empirical law.

A lot of the early team at Anthropic made the observation. People speculated as to why at the time, and Emmanuel Ameisen says the result is still mostly an empirical finding. The finding means you can just keep pouring more compute in and you keep improving on the metric. A field where more compute keeps improving the metric is rare. It is also rare for the metric to then correlate with a bunch of tasks people actually care abo

Emmanuel Ameisen says a language model can write a coherent report or code even though it outputs only one word at a time. Nobody writing reports or emails thinks only about the next word with an empty mind about anything that comes next. The intermediate results made while predicting a word, such as the fifth word, can be accessed and reused by every future word.

Ameisen says that reuse is how models formulate a plan. Poetry is the example, because it is hard to rhyme without planning. A line that ends on an it word has to be followed by another it word, and the word chosen can be rabbits. When models write poetry, they write a line, then the intermediate results mark an it rhyme and a pla

Emmanuel Ameisen says a model takes a bunch of words, turns the words into numbers, does math on the numbers, and gets intermediate results. It does more math, gets more intermediate results, and repeats the cycle many times. At the end it has a final result, which it converts back into a word. The math is the part that has grown. Learning to do the task better means updating parameters in one direction or another. The model then learns some formulas, and Emmanuel Ameisen says we have no idea what the formulas are.

Training runs over a corpus of data. The model is fed the first half of a document and asked to predict the word that comes next, then the same step runs again and again across a

Lenny Rachitsky says 44% of people in his tech survey felt burned out in 2025, and 55% said they feel burned out this year. He calls that a huge jump. The picture he describes is being asked to do more, with AI already on the desk, while other teams look faster and the company talks as if it will go out of business if it does not move.

Molly Graham reads the jump as exhaustion. Emotions are tiring, and so is trying to understand a rate of change that covers the tools and the story about the tools. One month one thing matters, and the next month it is something else. A friend at OpenAI told her it took six months to go from using AI everywhere to asking whether it was making a difference, be

Molly Graham says the loneliness shows up when the teammates go away and the day is spent with machines. Lenny Rachitsky describes an engineering leader who said teams used to be five or ten engineers working together, and now the teams are smaller and the work is talking to agents. It is not the same job.

A company can strip out the humans, pick up productivity, and still end up with sad people, and sad people do not do their best work. Molly reads the same pattern in survey data that small teams and smaller companies are happier than people at the bigger companies. At those bigger companies, management layers come out and the pressure is productivity, cost, and how efficient the machines

Lenny Rachitsky says the core test is to treat AI like an intern: decide what a person would give an intern and what they would never give them. AI is much more like an intern than a super intelligent being. Be deliberate about what gets handed over. Skip the work AI will do a terrible job at, and keep the work someone loves and does not want taken away. Do not give everything to AI. Avoid brain rot and stop over-relying on AI.

So far there are more engineers, and so far the change is going well. Give away your Legos, but stay discerning about what should stay human and what humans are here for. He allows that there may be some things people should never give away, and the choice is genuine

Lenny Rachitsky says the strongest lever for happiness at work is the manager. A manager is the one thing a company can actually change to increase people's happiness at work right now, and most people's managers are not good. Taking care of managers is where the leverage sits, while the wider AI shift is much harder to do much about.

Molly Graham calls the trend inside bigger companies to cut whole management layers a huge mistake that will hurt over the long term. There is no sign that management matters less. If anything, management matters more. The efficiency and the cost savings are real, and they come with costs nobody can predict. A great manager makes people feel seen and supported

Lenny Rachitsky says AI is making products so much easier to build that ambition has to keep rising, and that builders have to learn the skill because people are not naturally that ambitious. Peter Sellis agrees. The status quo is the status quo for a reason. The work is a muscle for the courage to make tomorrow unlike today, against systems built to keep things still. He wants the acceleration, not a better position and not just more speed. That push is not natural.

People who can do it again and again inside a large company, without only demanding more, are rare. The trendy version showed up as the Steve Jobs ethos spread: always ask for more. When Lenny describes a ChatGPT product leader

Peter Sellis says the median product manager is probably pretty bad, and often a net negative, because the skill is a power distribution. If the skill starts out spread like a normal curve, the people who become product managers are the right-hand tail. Cut that tail off and the median sits far below the average. Most of the people who remain are on the left side.

Two forces make it worse. In the low-rate years he calls the Zerp era, product became the most lucrative tech career for people who are not technical. The job stayed subjective, which is why he says Lenny Rachitsky's podcast has been so useful, and the pay matched a technical ladder, so anyone inside had a strong reason to stay. T