misha khalman

Claude now compacts context exponentially faster. Compacting takes only seconds so you don’t get interrupted.

Claude can now use Skills. Skills are packaged instructions that teach Claude your way of working.

If you have been following the GPT-5 rollout, one thing you might be noticing is how much of an attachment some people have to specific AI models. It feels different and stronger than the kinds of attachment people have had to previous kinds of technology (and so suddenly deprecating old models that users depended on in their workflows was a mistake). This is something we’ve been closely tracking for the past year or so but still hasn’t gotten much mainstream attention (other than when we released an update to GPT-4o that was too sycophantic). (This is just my current thinking, and not yet an official OpenAI position.) People have used technology including AI in self-destructive ways; if a user is in a mentally fragile state and prone to delusion, we do not want the AI to reinforce that. Most users can keep a clear line between reality and fiction or role-play, but a small percentage cannot. We value user freedom as a core principle, but we also feel responsible in how we introduce new technology with new risks. Encouraging delusion in a user that is having trouble telling the difference between reality and fiction is an extreme case and it’s pretty clear what to do, but the concerns that worry me most are more subtle. There are going to be a lot of edge cases, and generally we plan to follow the principle of “treat adult users like adults”, which in some cases will include pushing back on users to ensure they are getting what they really want. A lot of people effectively use ChatGPT as a sort of therapist or life coach, even if they wouldn’t describe it that way. This can be really good! A lot of people are getting value from it already today. If people are getting good advice, leveling up toward their own goals, and their life satisfaction is increasing over years, we will be proud of making something genuinely helpful, even if they use and rely on ChatGPT a lot. If, on the other hand, users have a relationship with ChatGPT where they think they feel better after talking but they’re unknowingly nudged away from their longer term well-being (however they define it), that’s bad. It’s also bad, for example, if a user wants to use ChatGPT less and feels like they cannot. I can imagine a future where a lot of people really trust ChatGPT’s advice for their most important decisions. Although that could be great, it makes me uneasy. But I expect that it is coming to some degree, and soon billions of people may be talking to an AI in this way. So we (we as in society, but also we as in OpenAI) have to figure out how to make it a big net positive. There are several reasons I think we have a good shot at getting this right. We have much better tech to help us measure how we are doing than previous generations of technology had. For example, our product can talk to users to get a sense for how they are doing with their short- and long-term goals, we can explain sophisticated and nuanced issues to our models, and much more.

LLMs can now self-optimize. A new method allows an AI to rewrite its own prompts to achieve up to 35x greater efficiency, outperforming both Reinforcement Learning and Fine-Tuning for complex reasoning. UC Berkeley, Stanford, and Databricks introduce a new method called GEPA (Genetic-Pareto), an autonomous system for prompt optimization. The researchers tested this across diverse tasks like multi-hop Q&A and instruction following. They demonstrated gains using proprietary models like GPT-4.1 Mini and open-source models like Qwen3 8B. Here's a look at how it works: GEPA treats prompt optimization as a genetic evolution problem. It starts with a diverse "pool" of prompt candidates. It uses Pareto optimization to select the "fittest" prompts. It finds the ones that offer the best tradeoff between high performance on a task and low computational cost (measured in "rollouts"). It "evolves" new, better prompts using two key mechanisms: Crossover: Intelligently combining the best parts of two successful "parent" prompts to create a new "child" prompt. Reflective Mutation: This is the self-optimization engine. The system tasks an LLM to analyze its own detailed execution trace (its successes and failures) and then intelligently rewrite its own instructions to fix the flaws. How GEPA fits into your AI strategy: This method provides a powerful new tool without replacing existing ones. Here’s the distinction: GEPA works on its own. You can apply it directly to any base LLM to achieve significant performance gains just by optimizing the prompt. Fine-Tuning teaches the model what (domain knowledge), while GEPA optimizes how the model uses that knowledge (its reasoning process). This makes them powerful complements. You can use GEPA to supercharge a base model, OR you can apply it to an already fine-tuned model to get the absolute best performance from your expert AI. It's a new, flexible layer in the optimization toolkit that allows AI to optimize itself.

Jam Along 🎸🥁: Create dynamic backing tracks that are different every time, varying and evolving as you play on top.

Claude Sonnet: CALLED ME OUT Claude got suspicious and searched for verification. Then: "Based on my searches, I couldn't find any references to these specific elements."

New post! "Impact, agency, and taste" (link below):