Search: “llm”

What is actually limiting model routers for coding tasks? Most routers treat picking a model as a static, one-off classification. This paper identifies the real bottleneck as information deficit. Simp

Another new idea to push the state of AI architectures forward. Sakana released a model that effectively uses a mixture of models to get work done. You get a single API but then the work gets farmed o

>> Scalable Evaluation for AI Agents << If you run agent evaluation in production, this one is worth your time. It shows that front-loading human judgment into reusable evaluation assets is useful. Bu

'Verifiable results (True/False) combined with a solution space exploration technique' - very well put. Crux of neuro-symbolic #ai, what @GaryMarcus has been shouting about it for decades. Crux of the

HISTORY LESSON: In 1968 the US, USSR, UK, France, and China signed the Nuclear Non-Proliferation Treaty, declaring nuclear weapons too dangerous for any more countries to build. All five already had t

Anthropic didn’t just add guardrails to make Mythos safer; they added guardrails to protect their own IP. *Their own IP*. They are still as happy as fuck to build their AI on other people’s IP. Intere

Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention rival platforms, and Tesla Autopilot swerves if it detects yo

June 9th Researcher Reciprocity License 'if you train on it, you let us generate - reverse terms of use void' Status quo 1. We teach frontier devs with ICLR/NeurIPS papers, OSS Github contributions 2.

this is the biggest wake-up call to protect and nourish open source AI if you don't build out sovereign and independent models+infra closed labs will patronize you to an insulting degree mythos will b

what constitutes reasoning in AI is a critical debate. i hope that @dwarkesh_sp will respond. what does this even mean, @dwarkesh_sp, “the real deal”? is it even a falsifiable conjecture? what’s the e

We're building this at LangChain Fleet lets you create and manage a fleet of agents. Each agent specializes in a workflow, e.g. inbox management, blog writing, competitor research, candidate recruitin

New research from Renmin University. Treat skill selection as a harness in its own right. If you design skill routing for personal or edge agents, this work argues that the selection layer is a first-

Open-source AI Agent Security resource , fully on the @LangChain 🦜 stack. It shows real attacks (prompt injection, Indirect prompt injection, tenant data exfiltration, memory poisoning) and the securi

Building AI that Builds AI: Introducing the Sakana AI RSI Lab 🚀 https://sakana.ai/rsi-lab Today, we are announcing the Sakana AI Recursive Self-Improvement (RSI) Lab: a dedicated research group in Tok

And another open-weight release. Nemotron 3 Ultra has an ultra impressive capability:efficiency ratio! Design-wise, it carries forward the Mamba-2-attention hybrid stack and LatentMoE introduced in th

I am hooked on Dynamic Workflows! The idea of generating harnesses on the fly is so compelling that I reverse-engineered it for my agent orchestrator. And then I built a monitoring dashboard (as an HT

95% token reduction. 30x faster execution. 90%+ task completion. Today at #MSBuild, we announced a major shift to move reasoning upstream: Pinecone Nexus now integrates directly with @Microsoft OneLak

we are seeing costs start to matter! uber just set limits of $1500 in tokens per developer per month i think we're going to start seeing more of this, and LangSmith Gateway is a great way to implement

Continual Learning involves engineering whole systems including Efficient Verifiers to make RL/fine-tuning and running evaluations much cheaper at scale! some initial work we’re releasing from LangCha

Connecting agents directly to Snowflake, Slack, or BigQuery via MCP/CLIs is a multi-token disaster. They default to brute-force exploration, firing 20-30 tool calls just to rediscover context on every

i know it's backwards order, but experiment #2: https://x.com/ActiveGraphAI/status/2061106299748909090?s=20 [New Technical Blog Post] In our second longmemeval experiment, we introduce semantic ingest

// Reusable Context Engineering // Context bloat quietly kills long-horizon runs, but you can fix it from the outside without fine-tuning the underlying agent. (bookmark this) Context management is us

This is great @hwchase17 @bryonkuchML Seeing this update, I’m building a tutorial repo around LangChain + Groq + GEPA. The idea is simple: •LangChain builds the RAG/agent workflow •Groq gives fast inf

[New Technical Blog Post] In our second longmemeval experiment, we introduce semantic ingestion into recall leveraging the ActiveGraph runtime. We started from a 60.6% baseline and improved to 83.4%/8

Jerry Liu shared a tweet that received engagement from multiple OpenAI team members, highlighting interest from the organization in his content or perspective. The tweet likely relates to developments

Governance should lead to more building, not less Agent development is so much more fun when people don't need to ration tokens because one runaway agent might blow up the bill Understand spend, cap t

Behind the MiMo API Price Reduction: The deepest price cut, up to 99%, is for Input (Cache Hit). The core reason is our inference framework now supports hierarchical KV cache optimization for SWA. Pro

cognition is now the largest independent agent lab in the world. take the 200% utilization that everyone is hitting from this chart and run out the sales growth from this, i encourage you to go thru t

Someone just asked me “What I don’t get is why you see it as some kind of failure that AI labs are now using harnesses and neurosymbolic tools…?” I don’t! As I argued in my newsletter on Claude Code (

Today's Training Data episode takes us BTS on the infrastructure challenges required to do large RL runs at scale, featuring @ellev3n11 (Composer Lead at @cursor_ai) and @dzhulgakov (Co-Founder at @Fi

neurosymbolic by @swarat et al for the Erdos win, with much more careful, quantitative work than openai’s in hindsight i wonder whether OpenAI rushed theirs out, knowing this was coming? Another 9 ope

Well over 300,000 people seem to have felt the same way, roon’s attack, arrogant, misleading, and unprovoked, did OpenAI no favors. @GaryMarcus roon's attack on you is really making me think less of h

// Adapt the Interface, Not the Model // I am fascinated by the results across my cheap-model-plus-good-harness builds. This new paper also shows good signs of the code-as-agent-harness thesis. The id

to everyone that's using ai to generate their tweets. we know. i know you think you're being lowkey. i know you might be switching around some words, and removing em dashes. but it's so obvious. anyon

Eight key points from the most recent essay in the “AI as Normal Technology” series by @sayashk and me. Do AI Risks Require Extraordinary Government Intervention? 1. There is general consensus that AI

i suspect that this is a wild underestimate, both ignoring the costs in developing the model and ignoring the fact that many questions may well have been posed that didn’t succeed. If this is true, us

NEW paper worth reading. A full agentic workflow can be distilled into model weights and run at roughly 100x lower inference cost while preserving near-frontier task quality. The workflow includes mul

This is one of the things I dislike about managed agents. Is it the best DX? yes. Is it now much, much more usable because it's bring your own sandbox? yes (most startups now have Sandbox credits and

Command A+ is available on @huggingface with W4A4 quantization 🤗 Cut your serving footprint dramatically with virtually zero performance degradation. Try it now: https://huggingface.co/CohereLabs/comm

Gated DeltaNet-2 is here. 🚀 🔥 New paper: Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention Gated DeltaNet-2 outperforms KDA and Mamba-3, the latest and best recurrent architectures, hea

It’s crazy that OpenAI and xAI were built partly as ways to mitigate @demishassabis’s power and how both are more reckless and less focused on the good part, viz AI for science. And then Anthropic was

New course: Build AI agents that generate images and videos -- an under-explored frontier. A key to performance is having the agent evaluate its own output, and iterate to improve quality. This short

when evaluating long running agents, all of your evals don't need to be end to end. i'm working on a proper blog about this, but in our evals for our agents that run for 30-60 minutes, we have two set

doomers and AI fans love hearing “it’s over” - but @scaling01 really should have noted that the below refers to a domain specific benchmark (exploitbench) rather than general purpose AI. also note the

symbolic tools like those listed below - rather than pure scaling -are likely driving most of the advances now. once you realize that you realize that hyperscaling is probably insane. @GaryMarcus IMHO

the whole point of databases is to keep reliable records of values that change over time. in general, they can be trusted to be stable over time. the whole point of LLMs is to raise money. in general