Имбаша-ла-ша-ла-ла-ла

We’re all zombies and we should be using AI to reclaim our agency “Everyone, when they're kids, has big ideas and big dreams. And when they imagine their life and play it out, they imagine this life where they can be an astronaut, or save the planet, or change the world, or whatever it might be. They think big. And then, for a variety of reasons, by the time people finish school and enter the workforce — by the time you've been in a job for a while — all that hope and ambition and ability to dream has been sucked out of them. It's been beaten out of them. Life has been beaten out of them. I don't think we, as adults, appreciate the degree to which we're all zombies. We've lost our agency. And ambition, I think, is really a good word for it. One of the promises of AI, broadly speaking — and I think through products like Muse — is to keep that going from when you're a kid. You're a kid, you have big dreams. Use AI to help make those things happen, then dream bigger, make those things happen, dream bigger, make those things happen, and experience this escalator of agency — an increase in agency throughout your life — versus having lots of agency and then having it crushed.” — Alexandr Wang

I burned a shitload of compute so you don't have to. Releasing the Harness Leaderboard. Which harness should you run your open model in? I ran 28,849 trials to find out. The leaderboard compares nine coding-agent harnesses on open models. In each comparison only the harness changes: the model, provider and reasoning effort stay fixed, and every task gets three attempts. The trials ran GLM-5.3-Flash, DeepSeek-V4.1-Flash and Flint on Terminal-Bench 2.1 and DeepSWE, for $4,746 of inference. Averaged over the three models, mini-swe-agent (@KLieret), a bash-only agent, solves 51.4% of tasks and @openclaw 27.3%. Four others are statistically tied with mini-swe-agent. One of them is Pi (@pidotdev), at 49.4% for $0.29 per solved task, about 45% less than mini-swe-agent's $0.53. The best harness depends on the model: mini-swe-agent on GLM-5.3-Flash, Hermes Agent (@NousResearch) on DeepSeek-V4.1-Flash and Kimi Code (@Kimi_Moonshot) on Flint. @opencode costs $3.35 per solved task, 11.7 times as much as Pi. See Harness Hub's Harness Lens for what the nine harnesses do differently. It documents the prompts, tools and compaction of 34 coding-agent harnesses and links every claim to its source. Leaderboard: huggingface.co/spaces/joelni… Harness Lens: huggingface.co/spaces/harnes…

*sshole is out if you run sshd on alpine, openwrt or any musl box with password auth on: update now official writeup: sshhole.com

Плюс 15, а по ощущениям и все 20 на солнышке в Алмате сегодня 🍂🍁🌞

Фонд Almaty Air Initiative предложил установить проверяемый годовой лимит использования угля на алматинской ТЭЦ-2 после ее перехода на газ и обеспечить открытый доступ к данным о выбросах станции

Пацан сказал, пацан сделал 🫡

Тем кто не успел купить RTX5090 или DGX Spark и аналоги: только начали появляться в продаже RTX Spark ноутбуки, и их еще можно успеть урвать по примерно рекомендованным ценам 🤓

⚡Iterating at lightning speed! Gemini 4 Carbon is SOTA across key benchmarks and combines frontier capabilities with deep safety & security research.