Every story across every category, newest first. Each card links to the original publisher; daily-brief posts open as editorial pages.
Artificial Analysis has launched Optima, a platform that lets users build custom AI benchmarks from their own data and workflows. Models can be compared not just…
Title pretty much says it all. I’ve deployed Qwen 3.8 27B using vLLM on an RTX 6000 Pro (tried multiple vLLM releases and launch recipes), but…
Running a 12700k + 5060 Ti 16 gb with 128 gb DDR4 ram and I'm wondering what's the ideal setup to maximize performance. I did some…
A representative survey by Epoch AI found that 20 percent of employed Americans hand off at least one task to AI that a human used to…
I've got a dedicated AI rig sitting here with a RTX 5090 and 96GB RAM and for the past few years have been using Windows 11…
Article URL: https://breck.lol/plMakers.html Comments URL: https://news.ycombinator.com/item?id=49316888 Points: 9 # Comments: 2
TL;DR Google DeepMind's recent model DiffusionGemma (DG) generates text via diffusion, meaning many diffusion steps happen before generating the final output. In particular, these diffusion steps…
RT Xudong HanAgent 自进化，并不会每一轮都变得更强。这篇《Rethinking Self-Evolving Agent Skills》专门研究 Skill 在多轮迭代里到底是怎么进化的。作者做了 42 组 evolution run，一共生成 388 个候选 Skill，真正刷新验证集最好结果的只有 55 个。也就是说，大部分修改其实没有带来有效提升，很多轮次会停滞、退化，最后被回滚。还有一个很有意思的结果：最终被选中的 11 个进化 Skill，全部使用了失败轨迹作为反馈；只看成功经验的方案一次都没有胜出。失败轨迹更像是在告诉 Agent「哪里需要修」，成功轨迹则更适合保留已有能力。这也说明了自进化更像一场带验证
Article URL: https://www.falstad.com/mathphysics.html Comments URL: https://news.ycombinator.com/item?id=49316872 Points: 3 # Comments: 0
RT Xiaoyin QuDear Dario,1. If Claude can cure cancer to save people like your dad, why should we "pace the progress"? Does that mean more people…
Article URL: https://forkast.news/the-model-context-protocol-reaches-a-security-inflection-point/ Comments URL: https://news.ycombinator.com/item?id=49316815 Points: 11 # Comments: 1
Article URL: https://koutian.is-a.dev/benchmark-radar/?view=leaderboard Comments URL: https://news.ycombinator.com/item?id=49316791 Points: 7 # Comments: 7
Does anyone know if the ASUS B860M can handle two large BAR GPUs? I am considering a couple build options with this motherboard, but do not…
Article URL: https://www.science.org/doi/10.1126/sciadv.adx3012 Comments URL: https://news.ycombinator.com/item?id=49316685 Points: 7 # Comments: 0
I was surprised by the fact that the qwen 3.8 27b download count is about 1 million (globally). This means that even on this subreddit, very…
RT Min ChoiOk Grok 4.6 is getting scary good.Games. 3D worlds. Game trailers. Even real-world 3D prints.10 wild examples:
Article URL: https://nntaleb.substack.com/p/have-another-drink Comments URL: https://news.ycombinator.com/item?id=49316611 Points: 10 # Comments: 3
Article URL: https://github.com/yebiguo/ProofRun Comments URL: https://news.ycombinator.com/item?id=49316605 Points: 5 # Comments: 0
Article URL: https://spectrum.ieee.org/stimulated-geologic-hydrogen Comments URL: https://news.ycombinator.com/item?id=49316395 Points: 5 # Comments: 1
RT SpaceXDeployment of @Globalstar 2-R satellites has started
My local inference build, with: ASUS WS C422 PRO/SE 10-core Xeon W-2255 64GB ECC RAM 64GB VRAM Pimped case with TurboLEDz indicating the frequencies of the…
RT SpaceXFalcon 9’s first stage lands on the Of Course I Still Love You droneship
RT Wall Street ApesThis PERFECTLY explains why ‘The Left’ is obsessed with being morally superior“Did you ever ask yourself why it's always the weakest, least competent,…
Yutori's browser-use agents run in tight loops: screenshot, action, repeat, dozens of times per task.On Together AI, their Navigator model beats frontier performance at 2x faster…