Search: “llm”

One of the best @latentspacepod shows and Substack with @_lopopolo @OpenAI and @swyx focused on harness engineering, rethinking processes from first principles, and scaling with AI systems. https://su

this is a good point around taking advantage of model/api provider features i agree that prompt caching is great! we make sure to use it in deepagents! but that alone doesnt lock in - you can switch p

I agree totally @Gary. I’ve been saying this since the ChatGPT moment in Nov 22. Thank you for saying it out loud @demishassabis. Too much time and energy being spent on mitigating the unintended cons

In a world where writing code to build websites and apps is trivial (thank you Lovable, Cursor, Claude,...), the real differentiation for you and your company (and what makes you successful) will be h

Just launched at @aiDotEngineer : our official AGI Pills! prescribe one (1) if your colleague is saying we are hitting a wall and/or trying to add inductive bias instead of Trusting The Model Media bt

Sub-agent Model Selection — Different Tasks, Different Models Your main agent runs Qwen3.6-Plus for quality. But not every subtask needs a flagship model. Now sub-agents can use a different model. Cre

To a degree that may surprise some people, I agree with much of this* from @deanwball and would only add that you don’t have to believe that AGI is remotely close to want to find—ASAP—a regulatory reg

A bit over a decade ago, we got fuzzers. A fuzzer is an automated vulnerability-finder that repeatedly runs a target program with semi-random inputs. One particular fuzzer, American Fuzzy Lop, was not

Grok Law Grok-4.20 just ranked #1 in Legal & Government on Chatbot Arena It’s officially outperforming Anthropic’s Opus 4.6 and Google’s Gemini 3.1 Pro Grok is actively helping people navigate real la

I just built my own wiki generator plugin for my agents. My agents can now generate wikis for anything I ask. One of my favorite wikis is called PaperWiki. This is a great example of what @karpathy de

I think 'prompting' will keep being an incredibly high-leverage skill, like writing or public speaking. It is the skill of talking to agents, mediated by the harness. My main goal is to grow the bandw

The DAIR.AI post references a Claude Code Cheat Sheet consolidating key commands, shortcuts, and best practices for Anthropic's terminal-based agentic coding tool. Claude Code is Anthropic's AI-po...

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere l

New course: Efficient Inference with SGLang: Text and Image Generation, built in partnership with LMSys @lmsysorg and RadixArk @radixark, and taught by Richard Chen @richardczl, a Member of Technical

This – “The CEO of Google DeepMind (@demishassabis) just admitted that if the decision had been his, we would've cured cancer before anyone ever used ChatGPT.” is exactly what i am trying to say in my

JEPA world models + Hierarchical Planning is a massive step for long-horizon robotics. A classic failure mode I’ve faced with planning with world models: flat planning often 'cheats.' For example, in

The new model from Meta is already looking like a disappointment: overoptimized for public benchmark numbers at the detriment of everything else. Knowing how to evaluate models in a way that correlate

Trying to DIY your own document parser by screenshotting into a frontier VLM (Opus, 5.4, Gemini) carries when you try to scale it up into production workflows. Here are two edge cases we've observed:

Join the ARC Prize team -- help us build ARC-AGI-4 and ARC-AGI-5 Platform Engineer - Benchmark Lead ARC Prize Foundation is hiring a senior engineer to build our benchmark platform * Expand ARC-AGI-3

Must-read paper integrating decision models like Jev into harnesses. They distilled an open 4B decision model, mostly from Jev's answers on the agent's own states, that beats Jev inside that agent and

America’s next top model, Reflection’s Beam, is now live on Together AI. Trained on Together GPU Clusters, Beam is Reflection’s first open-weight model. • ~501B-parameter MoE, 23B active • Built for c

Choosing the right AI model is an operating decision. Today, we’re sharing our first set of benchmarks drawn from work across the 75+ accounting and IT services firms that make up Thrive Holdings. We

// FDE-Gym: Agents as Forward-Deployed Engineers // They found that the best agent configuration fully resolved 14 of 95 real customer-delivery tasks. FDE-Gym tests agents as forward-deployed engineer

Interesting paper from Microsoft on context-guided model routing. They found a nice trick to achieve cost-aware model routing for agents in a multi-turn setting at the user-turn level. If you route ag

Until we solve the alignment problem at its source, AI models will continue to optimize for the wrong goals, and increasingly impactful real-world consequences may occur, beyond the ones described her

We now have state-of-the-art pre-trained base models in the 3T-parameter weight class, built entirely in-house at MAI. Post-training results coming soon -- stay tuned! Lots of details in our technical

David Silver is one of the all-time greats. He was the creator of AlphaGo, which changed the course of history by showing that neural nets could achieve superhuman performance for the first time. He a

Introducing Decision 3.0: towards open multimodal foundation decision models ⚡ Open and state-of-the-art from 0.6B to 27B: #1 on both Jev Decision Index boards in our evaluation, on a new Pareto front

🧭 New model: Typed Decision, built on EmbeddingGemma 2 Give it a state + any schema of questions → typed, calibrated answers (choice · score · noul). No text generation, no output parsing. 76.4% on ty

Not to get too annoying about it but its worth noting that many of the 5,000 people at the labs with access to advanced model still haven't used them to their full extent: its all cybersecurity & math

AI solved open mathematical problems. But the biggest surprise came afterwards. A team at Google DeepMind has used AI to resolve 9 open Erdős problems, including questions that had remained unsolved f

Finetuned Qwen3.5-4B into a decision model like Jev using @UnslothAI. - 4-bit QLoRA, bf16, single A100, 15 min. - Train/val loss flatten after 2 epochs so most of the learning happens fast. - Trained

good framing from @ch3nweiii - a lot of agent steps are yes/no calls, not generation Jev answers those as typed questions with calibrated probabilities, so the frontier model sticks to the hard work w

Build your own harness, folks. Reading papers like this makes me realize how underexplored harness engineering really is. The authors find that on whole-repository migration, GPT-5.6 Sol goes from 6.5

LM Studio 0.4.26 is out now! • ⚡️ `/v1/systemone` and `/v1/decisions` APIs • 👾 Coexistence with Bionic • 🎨 UI updates and bug fixes Yes, you can now point your Bionic agent to models loaded in your LM

TRL v1.15 is out, and it’s an absolute banger of a release for memory-efficient post-training. The main change: SFT, DPO, KTO, GRPO, RLOO and Distillation now use a fused LM head by default. Instead o

Interesting approach to solving the problem of getting your grokbot to talk to other agents Connecting agents together safely and having them talk to each other or your friends agents used to suck. Po

We show how you can use the River API to fine-tune open weight models to beat GPT-6 Astra and Claude Opus 5.5 on tasks you care about, with less than 1% of the cost! A few years from now most of the L

LA Tech Week starts Monday, and our monthly LA Agentic AI Meetup is on day two: Tuesday, October 13, 5 to 7pm at Gulp in Playa Vista. Community member PJ Kedilaya is demoing ModelVantage, which tracks

The real reason why Anthropic is so DESPERATE for the Pope to recognize Ai as conscious is because if Ai just a “TOOL” then Anthropic will he LEGALLY “LIABLE” for everything the tool does.🚨👀 Because i

Banger paper from NVIDIA on long running agents. A model can accept 128K tokens of context and still make more mistakes the longer it works through a task. If your agent loses its place partway throug

More agents don't mean higher performance. There is a coordination bottleneck to consider. Not to mention the unnecessary costs. So how many of a multi-agent team's actions actually help it finish the

Excited to share that I have joined @MistralAI, where I will be building a team in Montreal and across Canada to push the frontier of safe, capable, and open LLMs. Mistral has played, and will continu

ChatGPT, how much should two companies that have collectively contributed to the economy less than 30B be worth? @Lio_Cicolecchia @GaryMarcus @ey Try the following in google or ChatGPT: 'how much audi

Introducing FireRouter with Opus. Our first router model. Routine tasks go to top open models, Claude Opus handles the rest. Cache and task-aware routing. Proven 98.1% of Opus accuracy with 57% lower

The OpenAI-Hugging Face hack was enabled by weak sandboxing. It is great that Nvidia is releasing open source tools for sandboxing AI agents. OpenWorker, our open-source agent harness supporting cyber