Meta's Superintelligence Labs shipped its first true frontier models, Muse Wonder and Muse Wonder Full, claiming state-of-the-art performance on personal and professional agentic tasks. Muse Wonder is available through the Meta Model API and Muse Code, while the larger Muse Wonder Full is limited to Muse Code subscribers. Microsoft countered on infrastructure, unveiling its Maia 300 custom AI accelerator with a claimed 40% more tokens per megawatt than leading GPU-based systems ahead of a 2027 production rollout.
Distribution battles intensified as X launched "The X Subscription," bundling Grok, Grok Bot, Cursor, and X Premium into tiered plans from free to $200/month. On the open-model front, Z.ai released GLM-Legal (built on GLM-5.3-Flash), matching frontier models on mainstream legal benchmarks with a reported 19.2% all-pass rate on Harvey's LAB, while Anthropic detailed new defenses it says have significantly reduced the effectiveness of distillation attacks against Claude.
OpenClaw shipped a second beta iteration of its 2026.10.5 line, 2026.10.5-beta.2, on the npm beta (prerelease) channel on October 12.
Meta's Superintelligence Lab rolled out a pair of state-of-the-art models for agentic work, its first real frontier-model attempt roughly 14 months after Zuckerberg reorganized Meta's AI efforts into MSL. Muse Wonder is available via the Meta Model API and Muse Code, while the stronger Muse Wonder Full is reserved for Muse Code subscribers, with only the smaller model exposed via API.
A single tiered plan running from free to $200/month now folds Grok AI tools, the Cursor coding assistant, and X Premium into one subscription, with migration paths for existing customers. The move treats coding-agent access as a consumer subscription category rather than a standalone developer purchase, and bundles the full path from thinking and creating to building and sharing under one bill.
Built on TSMC's N2P node, Maia 300 is designed for inference efficiency, with Microsoft claiming 40% more tokens per megawatt than leading GPU-based systems and rack-level throughput up to seven times higher. Production rollout begins in 2027, deepening Microsoft's effort to reduce its reliance on NVIDIA GPUs for AI workloads.
Built on GLM-5.3-Flash, GLM-Legal targets contract review, evidence analysis, and compliance, and reportedly scores a 19.2% all-pass rate on the Harvey LAB benchmark of real-world law-firm workflows. That places it among the top legal models and within a fraction of Gemini 4 Argon, the most capable commercial legal AI. API access is by application.
Anthropic says it has significantly reduced the effectiveness of coordinated distillation attacks, making it harder to use proxy networks to extract Claude's protected reasoning. With Claude Sonnet 5.5, reasoning transcripts are accessible only to the organization whose request produced them, and the company acknowledges accepting some friction for legitimate users as the cost of protecting its reasoning data.
Selected posts from the edition, with links to the originals.
"Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work."
View post on X ↗"Over the last few months, we've made significant progress in reducing the effectiveness of distillation attacks against Claude. In this post, we explain how."
View post on X ↗"Introducing GLM-Legal. Built on GLM-5.3-Flash, GLM-Legal matches frontier models on mainstream legal benchmarks. Designed for real legal workflows: contract review, evidence analysis, compliance, and more."
View post on X ↗"Introducing QwenGyre, bringing elastic RL to extreme-long-horizon (XLong) agent training. The Elastic Scheduler reallocates GPUs between rollout and training as demand shifts — live executions never interrupted, with up to 1.85x speedup."
View post on X ↗"This is how we built the filesystem behind Perplexity Projects as a Git forge for agents. Each agent works in its own sandbox and commits to a shared repository where people and other agents can build on its work across sessions."
View post on X ↗"OpenClaw v2026.10.2 patch release is out 🦞 Upgrade and Doctor fixes, missing and duplicate reply fixes, long-conversation fixes, credential-masking fixes, and more. 107 contributors. Thank you!"
View post on X ↗"The gap is widening because a tiny group of people sits on frontier systems with near infinite quotas, treating capability discovery as a full-time job developing skills, intuition, and applications before anyone else can."
View post on X ↗"Amazing to see research move from the lab to real-world clinical utility — congrats to the team on this milestone & The Lancet paper!"
View post on X ↗"Our new silicon is 🔥 40% more tokens per watt!"
View post on X ↗"Le Chonk cooked on visual grounding. Pinpoints exactly where things are in an image. 42% on Dense 200, ahead of even frontier closed models."
View post on X ↗