Daniel Grittner

@mamagnus00 @browser_use Would be cool

@danielgrittner_ @browser_use Sure I can also open source the file

When to do compaction? I could reduce the cost of my agent by 80% with this.

In era of pretraining, what mattered was internet text. You'd primarily want a large, diverse, high quality collection of internet documents to learn from. In era of supervised finetuning, it was conversations. Contract workers are hired to create answers for questions, a bit like what you'd see on Stack Overflow / Quora, or etc., but geared towards LLM use cases. Neither of the two above are going away (imo), but in this era of reinforcement learning, it is now environments. Unlike the above, they give the LLM an opportunity to actually interact - take actions, see outcomes, etc. This means you can hope to do a lot better than statistical expert imitation. And they can be used both for model training and evaluation. But just like before, the core problem now is needing a large, diverse, high quality set of environments, as exercises for the LLM to practice against. In some ways, I'm reminded of OpenAI's very first project (gym), which was exactly a framework hoping to build a large collection of environments in the same schema, but this was way before LLMs. So the environments were simple academic control tasks of the time, like cartpole, ATARI, etc. The @PrimeIntellect environments hub (and the `verifiers` repo on GitHub) builds the modernized version specifically targeting LLMs, and it's a great effort/idea. I pitched that someone build something like it earlier this year: x.com/karpathy/statu… Environments have the property that once the skeleton of the framework is in place, in principle the community / industry can parallelize across many different domains, which is exciting. Final thought - personally and long-term, I am bullish on environments and agentic interactions but I am bearish on reinforcement learning specifically. I think that reward functions are super sus, and I think humans don't use RL to learn (maybe they do for some motor tasks etc, but not intellectual problem solving tasks). Humans use different learning paradigms that are significantly more powerful and sample efficient and that haven't been properly invented and scaled yet, though early sketches and ideas exist (as just one example, the idea of "system prompt learning", moving the update to tokens/contexts not weights and optionally distilling to weights as a separate process a bit like sleep does).

@davidzh @Hormold @baseten Did you test gpt-oss for conversational voice AI? How is the conversational behavior? wondering because it's trained as a reasoning model

@Hormold it's so fast. have you tried gpt-oss on @baseten? I was getting 120ms TTFT and couldn't believe my eyes

First impressions using GPT-5 in voice agents. It's far too slow even with reasoning_effort set to "minimal". I'm consistently getting 700ms+ inference latency. It's only day one, so it could be GPUs melting. I'll test it again in a couple of days.

👀

how cool is this

@dexhorthy Awesome overview! Only evals are missing imo

Everything is context engineering

+1 for "context engineering" over "prompt engineering". People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window with just the right information for the next step. Science because doing this right involves task descriptions and explanations, few shot examples, RAG, related (possibly multimodal) data, tools, state and history, compacting... Too little or of the wrong form and the LLM doesn't have the right context for optimal performance. Too much or too irrelevant and the LLM costs might go up and performance might come down. Doing this well is highly non-trivial. And art because of the guiding intuition around LLM psychology of people spirits. On top of context engineering itself, an LLM app has to: - break up problems just right into control flows - pack the context windows just right - dispatch calls to LLMs of the right kind and capability - handle generation-verification UIUX flows - a lot more - guardrails, security, evals, parallelism, prefetching, ... So context engineering is just one small piece of an emerging thick layer of non-trivial software that coordinates individual LLM calls (and a lot more) into full LLM apps. The term "ChatGPT wrapper" is tired and really, really wrong.

Goodbye StackOverflow 🥳

So true

Y Combinator is coming to Munich to visit CDTM, TUM & LMU on April 23rd. I'm excited to chat with @MarcKlingen from @langfuse (YC W23). I will also share what I learned from working with over 900 startups at @ycombinator and how to choose the right idea to work on. If you are a student at one of these schools - we'd love to meet you.

Rust programming in German. I think it's time to migrate our codebase 🍺🇩🇪 github.com/michidk/rost

Ghibli style is 🔥

Just a group of people building the future. LFG YC W25 🚀

The last few months were quite a ride. Excited for the next level 🚀

Jensen must feel like Anakin Skywalker - hope he doesn't become Darth Vader at some point

@StdBikeRepair @realEstateTrent @Admyral_ai everyone starts with 87 followers :)

@realEstateTrent With 87 followers! @Admyral_ai They coming for you @realEstateTrent 🤣🤣🤣

AI commercial real estate brokerage?? 👀

@realEstateTrent 👀

Our new Chief Security Officer, Jimmy O. Yang, shared some unconventional security tips at the last company all-hands meeting.