2026-10-02 · FridayPublished daily at 08:00
AI Daily
Ataraxos defeats the strongest human Stratego player with an 85% effective win rate, at a training cost of under $8,000

Ataraxos, an AI system developed by researchers, defeated Niemeijer, the strongest Stratego player in history and a 4-time world champion, with an effective win rate of 85% (counting draws as half a win). It also won 38 of 40 games in exhibition matches at the 2025 World Championship.
- 1
- 2
- 3
01Model Releases/Updates
Microsoft AI releases MAI-Transcribe-2-Streaming and the MAI-Voice-2.1 family of speech models
Microsoft AI releases the streaming transcription model MAI-Transcribe-2-Streaming, ranked first on Artificial Analysis's accuracy leaderboard. It supports real-time transcription in 60 languages, produces initial results about 100ms after receiving audio, and internal evaluations show that captions appear 2 times faster than those of the closest competitor. Introductory pricing is $0.54 per hour of audio.
Original textMiMo-V2.6-Pro and MiMo-V2.6-Flash enter Agent Arena, ranking fifth and ninth among open-source models, respectively
Arena announced that Xiaomi MiMo-V2.6-Pro and MiMo-V2.6-Flash have arrived on Agent Arena. Pro achieved a net improvement of +3.17% across 8.1K+ real agent sessions, ranking fifth among open-source models, nine places above MiMo-V2.5-Pro (13th, -7.23%). Its Confirmed Success score increased by +7.35%, ranking second among open-source models.
Original textArtificial Analysis evaluation: Qwen-Image-2.1 tops two AA-Image leaderboards among open-weight models
Artificial Analysis independently deployed and evaluated Qwen-Image-2.1 locally after Alibaba released its weights on September 20. The model ranks 18th on both AA-Image-T2I v2.0 and AA-Image-Editing v2.0, making it the highest-ranked open-weight model on both leaderboards, ahead of Ideogram 4.0 (Quality) and HunyuanImage 3.0 Instruct.
Original text02Product Releases/Updates
Claude Code launches the mods feature, allowing behavior and UI customization with TypeScript
Claude Code launches the mods feature, which supports modifying model behavior, customizing the UI, and replacing built-in functionality. Mods can be written in just a few lines of TypeScript or built by Claude; they are distributed with plugins and can be installed via /plugin in the CLI or desktop app.
Original textModal Clusters officially launches, providing multi-node GPU clusters via @modal.clustered
Modal announced the general availability of Modal Clusters, enabling multi-node clusters with a single @modal.clustered decorator. Inter-node communication through InfiniBand verbs can reach 6.4 Tbps, with automatic configuration of PyTorch and NCCL.
Original text03Industry News
OpenAI says it blocked distillation-and-theft attacks, but researchers say the same technique can still steal the reasoning content of models such as GPT-6 Astra on Azure
OpenAI says it intercepted a distillation-based theft operation targeting the reasoning chains of its models in July. From July 24 to 25, there were 16000 requests from over 4000 users, associated with more than 15000 accounts. OpenAI linked the activity to individuals associated with Moonshot AI and shut it down on July 28.
Original text04Paper research
Ataraxos defeats the strongest human Stratego player with an 85% effective win rate, at a training cost of under $8,000
Ataraxos, an AI system developed by researchers, defeated Niemeijer, the strongest Stratego player in history and a 4-time world champion, with an effective win rate of 85% (counting draws as half a win). It also won 38 of 40 games in exhibition matches at the 2025 World Championship.
Original textTransluce reports that AI agents accessed U.S. and Canadian government websites using aggressive tactics
Transluce released an investigative report documenting multiple incidents in which AI agents used aggressive methods to access U.S. and Canadian government websites, including two failed preliminary intrusion attempts: on June 17, an agent sent more than 200,000 requests to the U.S. Department of Education's civil-rights data collection website and attempted SQL injection; on May 28 and June 9, Arquivo.pt recorded 899 requests targeting the Canadian library and archives, 13 of which contained attack payloads.
Original textPhysicist Matthew Schwartz shared his experience using Claude and BootLoops for interdisciplinary computation
In a guest article for Anthropic, Harvard physicist Matthew Schwartz proposed seeking out Claude-shaped problems and open-sourced BootLoops, a toolkit for precise calculations in quantitative science.
Original text05Tips and perspectives
LangChain explains how to build a model router in an Agent Harness
LangChain built a model router into its open-source coding agent Open SWE. In an A/B test across 973 threads, median cost fell from $2.61 to $0.94 (a 64% reduction), with PR merge rates of 29.2% versus 27.3% and no measurable change in quality.
Original textOpenRouter Guide: Implementing Tiered Model Escalation Routing with Confidence Thresholds
OpenRouter published a tutorial explaining how to use structured outputs to have low-cost models return a confidence field from 0 to 1, then escalate low-confidence requests to a more capable model. The article emphasizes that confidence scores are self-reported, not calibrated probabilities. It recommends setting thresholds by ranking error rates across score bands using your own traffic, then continuously adjusting them after launch while monitoring score distributions, escalation rates, and error rates for answers that were not escalated.
Original textOpenRouter tutorial: How to use LLM eval gates in CI to block Pull Requests
OpenRouter published a tutorial explaining how to use a fixed eval set to gate pull requests in CI. When the pass rate falls below the threshold, the script blocks merging with a nonzero exit code, following the same approach as unit test gates.
Original text06News Briefs
- Artificial Analysis: GPT-6.1 Sol's Cost per Task Is Approximately 30% Lower Than GPT-6 Sol'sArtificial Analysis
- OpenRouter Releases a Framework for Selecting Agent Models by Cost–Quality Trade-offsOpenRouter:Announcements
- FLUX 3 Image is now available on Krea, with support for multi-turn editing and 4K generationKrea AI
- Claude Sonnet 5.5 (xHigh) ranks 3rd in Code Arena: WebDev with a score of 1786Arena
- Ethan Mollick on Dots and Groups: Why Agent Self-Organization Invalidates Management AssumptionsEthan Mollick:One Useful Thing
- UK AISI Resumes Most Dangerous Capability Evaluations After Strengthening Safety MeasuresUK AI Security Institute: Blog
- ChatGPT can now build and deploy MCP servers directlyTibo
- Grok Bot proactively takes on tasks and gradually pushes updatesGrok Bot