Cognition ties extra Devin usage to an efficiency score
The Information says Cognition divides estimated human hours by a Devin session's cost. Docs raise limits when the score is Healthy or Satisfactory.
Wildness 3/5 · Partly tamed
Topic · 115 stories · page 2 of 5
agents
115 stories tagged agents on AI Tamer, newest first.
OpenBMB's README announces PilotDeck 2.0 on 15 October
On 15 October the OpenBMB README announced PilotDeck 2.0, an AGPL agent OS with per-project memory. Its task scores are the project's own.
Wildness 3/5 · Partly tamedStepFun prices the STEPX Neo early-access phone at 3,999 yuan
StepFun priced the STEPX Neo early-access phone at 3,999 yuan, with Step AOS 1.0 and agent Amoo. Reports set the first sale on 16 October.
Wildness 3/5 · Partly tamedLexmount's Moli, a Rust browser kernel for agents, publishes a 17 October fix release
Lexmount's Moli is a Rust headless browser for agents, public since August. The 17 October v1.1.18 tag is a compatibility pass. Scores are Lexmount's.
Wildness 3/5 · Partly tamedOne honesty sentence changes what models admit in work reports
A 28 September preprint says GPT-5.5 flagged a planted negative result in 2 of 200 reports, and in 190 of 200 after one honesty sentence.
Wildness 3/5 · Partly tamedAtria Dawn is an official release, and the full weights are still coming soon
Shanghai AI Laboratory's site calls Atria Dawn an official release. Full weights are marked coming soon, and its pages split on 256K and 512K context.
Wildness 3/5 · Partly tamedDeepSeek points deepseek-flash at V4.1-Flash-1017 and posts the weights
DeepSeek published DeepSeek-V4.1-Flash-1017 under MIT on Hugging Face. Callers still use deepseek-flash, and DeepSeek says the price is unchanged.
Wildness 3/5 · Partly tamedTencent Hunyuan open-sources Hy Memory 2.0 and a 30B memory model
Hy Memory 2.0 is on GitHub and PyPI as hy-memory 2.0.2 under Apache 2.0. MemHub imports authorized tool chats into a local store. The scores are Tencent's.
Wildness 3/5 · Partly tamedDeepSeek documents standalone Search and Fetch APIs
DeepSeek's API guide documents Search and Fetch at 5 requests a second. The Chinese pages price each at 5 yuan per 1,000 successes; the English pages say $0.75.
Wildness 3/5 · Partly tamedOpenUI Lang 1.0 is on npm; Vue, Svelte and Angular stay on 0.x
Thesys published OpenUI Lang 1.0 on 12 October for React. Vue, Svelte and Angular packages stay on 0.x. The LICENSE file is MIT.
Wildness 3/5 · Partly tamedChatGPT can use a browser tab you mention, and Work can act in five browsers
OpenAI's docs say ChatGPT can use a tab you mention, and Work can act in five browsers after a site prompt. Treat page content as untrusted.
Wildness 3/5 · Partly tamedA Grok Bot user posts an Amazon block screen, and Musk says buy elsewhere
Patane posted an Amazon screen calling a Grok Bot an unauthorized agent. Musk said to buy elsewhere, then wrote maybe they should start a competitor.
Wildness 3/5 · Partly tamedGoogle starts a US rollout of Universal Cart
Google's 15 October post says Universal Cart is rolling out in the US on Search and in the Gemini app, with price alerts and merchant checkout.
Wildness 3/5 · Partly tamedNVIDIA releases open-weight Nemotron 3.5 Super VL and Bolt
NVIDIA's 15 October posts release open-weight Nemotron 3.5 Super VL and Bolt. The 5x throughput line and partner figures are NVIDIA's.
Wildness 4/5 · Still wildTheoPhysEval scores five agents on 50 unpublished physics problems
TheoPhysEval tests five agents on 50 physics problems. In one realistic run, the authors report GPT-6 Astra at 49 of 50.
Wildness 3/5 · Partly tamedNous tags Hermes Agent v0.22.0, the Threshold Release
Nous published Hermes Agent v0.22.0, the Threshold Release: a signed desktop app, a keyless first run, and local models. The line counts are Nous's.
Wildness 3/5 · Partly tamedNiko Matsakis on what coding agents change about Rust
Niko Matsakis's 14 October essay says Rust fits agentic coding and that types need proof tools. A Meta growth line in it is Tolnay's.
Wildness 2/5 · Mostly tamedLiquid AI releases LFM2.5-24B-A2B for on-device agents
Liquid AI released LFM2.5-24B-A2B on 13 October, a 24B on-device model it says uses 2.45B active parameters. Commercial use is license-limited.
Wildness 4/5 · Still wildDigitalOcean previews a learning loop, with automatic improvement in routing first
DigitalOcean previewed agent learning in public and a data catalog in private. Automatic improvement today is model routing; workflow automation is next.
Wildness 4/5 · Still wildNVIDIA and AMD each propose a rack for running AI agents at OCP
At the OCP summit in San Jose, NVIDIA and AMD each described a rack for AI agents, apart from GPU inference. Both are designs and plans, not shipping tests.
Wildness 3/5 · Partly tamedMeta releases Muse Wonder and gates the larger model behind paid tiers
Meta released Muse Wonder and Muse Wonder Full on 12 October. Wonder is on the API and Muse Code; Wonder Full is for Muse Code subscribers. Scores are Meta's.
Wildness 4/5 · Still wildMemento 3 authors report a ceiling score on public ARC-AGI-3 games
A UCL and Huawei Noah's Ark preprint says a frozen Claude Opus 5 agent cleared all 25 public ARC-AGI-3 games at the authors' efficiency ceiling. ARC Prize has not listed the run, and there is no author code release.
Wildness 4/5 · Still wildTencent retires Yuanqi on 9 November and points builders to its cloud agent platform
Tencent will shut the Yuanqi agent platform on 9 November 2026. New agents stop on 29 October, export runs until 9 January 2027, and Tencent's own guide points builders to its cloud ADP.
Wildness 2/5 · Mostly tamedTest Pydantic AI agents with a fake model instead of the live API
Swap TestModel or FunctionModel in for the real model and block real model requests, so Pydantic AI agent tests run fast, free and the same way every time.
Wildness 1/5 · Tamed
