NPO 프롬프트 최적화, Coding Agent 오케스트레이션, MiniMax 영상 생성, USB 기반 로컬 AI, Grok SF 창작

Interesting paper on prompt optimization. They claim that a single-lineage prompt optimizer just matched GEPA on a smaller rollout budget. Prompt optimization has been drifting toward heavier machinery, with candidate pools, reflection trees, and Pareto-based selection. NPO keeps one lineage. At each iteration it runs the student on the current prompt, collects rollout traces and rewards, and hands a sliding window of recent iterations to a teacher model that rewrites the prompt. There is no candidate population and no search tree. On the two instruction-following benchmarks it spends 3,500 and 6,800 rollouts against GEPA's 3,593 and 6,871, and it stays broadly comparable across 22 TextArena games. The interaction with teacher strength is what makes this interesting. NPO's advantage grows as the teacher model gets stronger, which suggests optimizer-side search complexity has been compensating for weak teacher reasoning all along. Paper: https:// arxiv.org/abs/2608.27266 Chat with Paper: https:// academy.dair.ai/papers/naive-p rompt-optimization-rethinking-the-need-for-complex-prompt-search-2608.27266 …

Teknium

@Teknium

4일 전

Many benchmarks have shown Hermes with Claude is better than Claude Code with Claude. We are better than most harnesses at all models

How good is Hermes Agent at coding compared to using Claude Code or Codex directly? My current setup is to have Hermes orchestrate Claude Code/Codex to do the actual coding work. I’ve been wondering if doing the coding directly in Hermes would be better, or if my current setup is already the better approach. @Teknium @NousResearch

Guillermo Rauch

@rauchg

4일 전

Cool

Just Plug in USB Chat with an Qwen uncensored local Ai models, No install, No cloud fits in your pocket. - air-gapped after first download - Chat UI on localhost - Win / Mac / Linux / Termux - abliterated Gemma + uncensored Qwen GGUFs

Uncensored AI coding CLI no-filter coding agent. - jailbreak system prompt baked in - CREATE_FILE + RUN_COMMAND on your machine - pitch: WormGPT + Qwen + Venice + vibe code in your terminal run an abliterated coder locally & OpenCode/Aider. - https:// github.com/alaashamaa017- cell/omegacode-cli-uncensored-AI …

Literature in AI is less explored than math and engineering. I have always wanted to write my own sci-fi. Building a story through conversation with Grok @bot is an interesting way to reflect on my thoughts. The post format works well because of the character limit. AI is good at condensing writing without losing its flavor. When humans self-censor and fear writing, good new literature is hard to find. Having AI tinker with your thoughts is another option. Anyway, this is the sci-fi I wrote for myself.

Literature in AI is less explored than math and engineering. I have always wanted to write my own sci-fi. Building a story through conversation with Grok is an interesting way to reflect on my thoughts. The post format works well because of the character limit. AI is good at condensing writing without losing its flavor. When humans self-censor and fear writing, good new literature is hard to find. Having AI tinker with your thoughts is another option. Anyway, this is the sci-fi I wrote for myself.