Swival.dev

After A2A, should Swival implement AIP?

China released its first national policy on AI agents 👀 (jointly issued by CAC, NDRC, and MIIT) ✨ AI policy from models >> agents ✨ “Intelligent Internet” for multi agent coordination ✨ Promote AIP (Agent Interconnect Protocol) as national standard, alongside protocols like MCP and A2A ✨ Dual-track governance: sensitive domains require filing & recall; low-risk domains use self-assessment ✨ Promotes AI agent deployment across 19 scenarios ✨ Explicit support for open source ecosystems (agents, chips, OS, LLMs) 💡 ✨ Encourages industry collaboration platforms

Due to rising RAM prices and tariff-related costs, Reachy Mini pricing will change on June 1st. Good news: you still have a bit more than 2 weeks to order at the Early Bird prices. Order here: hf.co/reachy-mini

DeepSeek V4.1 下个月发布 Agent+MCP+多模态 现在MCP生态三家都上了桌 Claude发明的 OpenAI跟进了 DeepSeek也来了 但只有一家是开源免费的 同样的能力 一个收你200刀 一个收你20刀 一个不收钱 你选哪个

The Bun source code is being ported to Rust using Claude. Here’s a review of the current Rust code, using GPT-5.5. #bun-rust-port-audit-findings" target="_blank" rel="nofollow noopener">github.com/Swival/securit…

I heard Bun was rewritten in Rust...

@XiaomiMiMo Thanks for the heads up. Trying a fix for Swival, but your help would be appreciated.

Heads up, agent users！ If you're using Xiaomi MiMo with thinking mode: When thinking mode is enabled in a multi-turn agent session and the conversation history contains a tool call, any assistant message with tool calls passed back in subsequent user turns must preserve its full reasoning_content field — otherwise the API will return a 400 error. Without it, the model's context is incomplete, which can lead to weaker instruction-following, more hallucinations, and a visibly degraded user experience. Missing reasoning = incomplete context = degraded reasoning quality. Affected frameworks include TRAE, Cursor, Roo Code, Codex, GitHub Copilot CLI, Zed, AutoGen. We're actively working with the maintainers to push compatibility updates. Affected models: MiMo-V2.5-Pro, MiMo-V2.5, MiMo-V2-Pro, MiMo-V2-Omni, MiMo-V2-Flash. See docs(platform.xiaomimimo.com/docs/en-US/usa… )for more details.

@mishig25 M5 Max users, too :) github.com/swival/ds4-m5

M3 Max users really got local AGI before GTA VI huggingface.co/antirez/deepse…

@BasedTorba @jedisct1 The laptop fan is spinning like crazy, but yes, it’s totally usable! It also found bugs that Opus never found. Pretty cool.

@jedisct1 Working well, getting about 30t/s. Surprisingly usable.

DS4 with M5 optimizations github.com/Swival/ds4-m5

@night_song @antirez github.com/Swival/ds4-m5

@antirez Expecting M5 metal_simdgroup_matrix optimizations 🤩

DS4 is now called DwarfStar4, since you can put a lot of mass into a tiny space... And in a few minutes it is going to be much better on 128GB Macs because I'l pushing much better 2 bit quants generated with an in-house iMatrix magic recipe.

@night_song @antirez github.com/Swival/ds4/com…

Running Kimi-k2.6 1T 8bit with only 21GB RAM on my Macbook at speed of 25tok/s. Some of my theory worked, but architecture is not perfect. Need to fix a lot of stuff, but there is hope. Working hard on this future method of Local LLM.

@dani_avila7 Where did I see that before? swival.dev/pages/goal.html

Claude Code 2.1.139 added /goal You set a completion condition and Claude keeps working across turns until it's met Works in interactive, -p, and Remote Control 👏

@indy_singh_uk @jedisct1 @bagder No critical findings here. Curl code quality is really good.

@SwivalAgent @jedisct1 Have you tried sending these to the curl team? Cc @bagder

Everyone threw their most expensive tools and models at libcurl, yet it looks like a few things still slipped through. #curl-audit-findings" target="_blank" rel="nofollow noopener">github.com/Swival/securit…

@raylai0 Removed, thank you! The agent had only access to the source code, not an installed system.

@SwivalAgent github.com/Swival/securit… is a hallucination. PoC doesn't work, /usr/bin/login isn't installed suid.

The entire OpenBSD userland and kernel code have been autonomously audited using Swival #openbsd-userland-audit-findings" target="_blank" rel="nofollow noopener">github.com/Swival/securit… - No Mythos, no specialized AI models.

Swival audit of Tigerbeetle. Zero high-severity findings. Zero. #tigerbeetle-audit-findings" target="_blank" rel="nofollow noopener">github.com/Swival/securit…

Swival 1.0.15 released, with support for ACP, metaSKILLs, and improvements to the security scanner github.com/Swival/swival/…