35 items tagged with this topic
Microsoft researchers share advances in building and operating large-scale distributed systems, spanning datacenters, networking, and the growing intersection with AI during NSDI ’26. The post Microsoft at NSDI 2026: Advances in large-scal…
Together AI and Adaption partner to bring Together Fine-Tuning natively into Adaptive Data, helping teams optimize datasets, run fine-tuning, evaluate results, and deploy stronger open models.
a quiet day.
3 weeks ago GBrain and Mempalace looked kind of similar 3 weeks later, it's pretty clear GBrain is its own category that is ideal for OpenClaw/Hermes personal AI scenarios, not trying to solve needle-in-a-haystack retri…
Deploying large language models (LLMs) in real-world, high-stakes settings is harder than it should be. In high-stakes settings like law, medicine, and cloud incident response, performance and reliability can quickly break down because ada…
Note: This episode was recorded just after AIE Europe, but before the Cursor-xAI deal.
Wanted a truly local storage for my tweets so built birdclaw. Imoorts your archive, backs it up on github, has jobs so you can import your x bookmarks daily (since they are not fully accessible via the api). https://t.c…
Releaseing GBrain v0.22 - lots of fixes to search and retireval and a new eval system (gbrain-evals is now a separate repo to prevent checkout bloat) https://t.co/qVAQjQeReV
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Huawei’s HiFloat4 training format beats Western-developed MXFP4 in Asce…
GPT-5.5 is live. We’ve been testing the model over the last couple of weeks at Box on our most complex knowledge work evals, and the model saw a 10 percentage point jump on accuracy of these enterprise content tasks vs.…
Next, I finally created new evals for GBrain which show how much more awesome GBrain is when you have graph AND vector search on top of grep on knowledge wikis https://t.co/eiVHZWvStL
Speaker 1 | 00:00 - 00:03 Isn't that crazy? That number is just mind boggling. Speaker 2 | 00:03 - 00:06 What is the state of the AI coding wars today? Speaker 1 | 00:06 - 00:16 We're in a phase of sort of, like, capability exploration. Th…
My Favorite Person Needs a Cofounder
Kudos to the folks from Tencent for working with us and providing evals to improve OpenClaw's harness performance! We're also working with them to bring fixes/improvements back to the open source repo. Great option for…
When you ask Claude to evaluate a potential investment and it replies “cute, just not as good as me” lol https://t.co/HL1aByBAoj
Agentic coding benchmarks like SWE-bench and Terminal-Bench are commonly used to compare the software engineering capabilities of frontier models—with top spots on leaderboards often separated by just a few percentage points. These scores…
Today is my last day at OpenAI, as OpenAI for Science is being decentralized into other research teams. It’s been a mind-expanding two years, from Chief Product Officer to joining the research team and starting OpenAI f…
When I build skills now I always ask AI to spin up a separate eval agent to do yes/no checks to grade the first agent's output. And if the output isn't "yes" across the board it asks the first agent to keep working. Bui…
The roadmap for Claude Cowork is… one month long. Things are moving too fast in AI to plan any further. Ship, evaluate, iterate. @felixrieseberg, who heads up Cowork at @AnthropicAI https://t.co/8YYpbM2yXU https://t.co/…
GBrain v0.9.3 now available. Search tuning, search evals, CJK queries, and better health checks! Plus lots of security hotfixes. Ask your claw or hermes to upgrade. https://t.co/ywYaRruXEZ
The more I meet enterprise CIOs and AI leaders outside of tech, the more it’s obvious that if you’re building software that doesn’t have a great headless mode, you’re going to be at risk in the coming years. Asked a gro…
Sometimes VCs will ask questions just to test your depth (market, technical, your obsession). The best founders (imo) are able to argue both sides of an argument with extreme clarity. There’s rarely something VCs ask th…
In evals, Sonnet with an Opus advisor scored 2.7 percentage points higher on SWE-bench Multilingual than Sonnet alone, while costing 11.9% less per task. https://t.co/pV4nftN1wz
I'm working on character evals and noticed that Claude would constantly pick itself as #1, so I removed the model names from the judge and changed things. https://t.co/Y9SqqJSYRc
@jackbutcher @thisiskp_ @danlovesproofs @bfirsh @akshat_b here is @simonw on the difference in self evident naming between “prompt injection” and “lethal trifecta” https://t.co/wSSVgEZeWM https://t.co/ovDqwQRvq5
We’re introducing a framework to measure progress toward AGI, and launching a Kaggle hackathon to build the relevant evaluations.
Imagine a robot tasked with cleaning a kitchen. It needs to observe its environment, decide what to do, and adjust when things don’t go as expected, for example, when the mug it was tasked to wash is already clean, or the sink is ful…
A welcome update from Google!
Speaker 1 | 00:00 - 00:20 I think this whole space is extremely difficult as things are emerging now. And, I mean, it's not only for world models. I think it's for everything, including text based models. Right? Because, you know, in the e…
Why it matters: Speaker 1 | 00:00 - 00:20 I think this whole space is extremely difficult as things are emerging now. And, I mean, it's not only for world models. I think it's for everything, including text based models. Right? Because, you know, in the e…