Introducing Claude Opus 5.5
Anthropic claims Opus 5.5 matches Fable 5.1 on most work at 40% lower cost than Opus 5, with external evaluators including METR.
Daily edition
Anthropic launches Claude Opus 5.5 and OpenAI ships GPT-6 Sol/Luna, while Xiaomi’s MiMo-V2.6-Pro tops open-weight charts.
Daily edition
97 of 97 feeds responded
413 distinct stories
77 previously issued, held back
25 worth your time
Collection needs attention
The separate collector was overdue when this page was built.
375 feed entries found this sweep; 413 candidate events; 56 shortlisted stories; 44 with article text available; 25 selected. Counts describe different stages; clustering, validation and refills can change the shortlist.
11,349 article records retained; 5,486 feed entries observed by collection this sweep; 78 retained candidates recovered for this ranking. Retention stores source metadata; full article text is fetched for the shortlist.
Collection snapshot: 2026-09-22T05:00:03.968256+00:00. This snapshot updates with the published briefing.
Active filters: none. Sort: Ranked.
All 25 stories, in ranked orderMore filters
Introducing GPT-6 Sol and Luna
OpenAI positions Sol and Luna as cost-efficient GPT-6 variants trained with similar methods as Astra, targeting professional work and coding.
Overreliance on AI contributed to missile strike on Iran school – Pentagon
The Pentagon reportedly found that overreliance on AI contributed to a missile strike on an Iranian school, per community discussion.
Xiaomi MiMo-V2.6-Pro 1T-A42B: the new top Open Weights model, trained for $3M
Xiaomi’s natively omnimodal MiMo-V2.6-Pro was trained for $3M and includes an UltraSpeed variant with up to 20x faster output.
Also covered byVentureBeat AIChina Probes DeepSeek, Moonshot Over Potential Data Leaks to Anthropic
China's internet regulator is investigating DeepSeek and Moonshot AI after Anthropic alleged user data was routed to Claude models.
China eyes chip-industry foothold with a tiny Han Xin code to challenge US standard
China is promoting a miniaturised ‘Micro Han Xin’ code for semiconductor labelling, entering the formal international standardisation process.
Jev introduces a new shape of LLM - System One, aka Decision Models
TypeSafe AI's Jev model returns typed probabilistic decisions instead of text, with free output and a lower input price than standard LLMs.
UN science panel says there is "no assurance humans will keep control" over AI agents
The UN's AI science panel warns that control over AI agents is not assured, citing the OpenAI Hugging Face incident as a real-world example of misaligned goals.
Meta patches Muse exploit that let attackers control the AI agent
Meta patched a zero-day in Muse for macOS that let local attackers redirect transcription to their own endpoint via an undocumented setting.
Also covered byVentureBeat AISimplifying Model Serving Across Multiple GPUs with NVIDIA TensorRT Multi-Device Integration in NVIDIA Dynamo-Triton
NVIDIA TensorRT multi-device inference now lets a single network execute across multiple GPUs via NCCL collectives, supported in TensorRT 11.0.
Meta's AI agent Muse tops app charts, driving AMD's market cap past US$1 trillion
Muse's app-chart popularity is cited as a factor pushing AMD's market cap past $1 trillion by highlighting CPU demand for agentic AI.
Xiaomi open-sources MiMo-V2.6 models after scaling reinforcement learning
Xiaomi says MiMo-V2.6-Pro scored 46 on the Artificial Analysis Intelligence Index, ahead of Kimi K3 and GLM-5.3.
Also covered byDigiTimesDeepSeek's Liang: Next Models Must Train on Huawei and Domestic Chips
DeepSeek’s CEO told investors training on Huawei chips is a priority, with deliveries eyed for Q4 2026–Q1 2027.
Building standards for the next phase of AI
OpenAI outlines a path to shared global AI standards, prioritizing coordinated evaluation, reporting, and governance to improve safety.
OpenAI, Microsoft ask for 2 week extension to opposition to New York Times’ summary judgment briefs due to many late “errata” changes & failure to…
OpenAI and Microsoft requested a two-week extension in the NYT lawsuit, citing late errata and 60 unserved exhibits from plaintiffs.
NVIDIA Introduces SoL-Pi: Auto-Research Loops That Cut Coding Agent Token Traffic by Up to 49%
NVIDIA's SoL-Pi harness mechanisms cut Pi coding agent token traffic by up to 49% on EdgeBench and API cost by roughly 33%, per NVIDIA's measurements.
OpenAI forms math advisory group as its AI resolves more than 100 open problems
OpenAI claims an internal model resolved over 100 open math problems and formed an advisory group after publishing a Navier-Stokes solution.
AWS Strands Agents Team Releases Strands Harness: An Open-Source Agent Harness With 28% Lower Token Cost at Comparable Accuracy
AWS's Strands Agents team claims Strands harness achieves 28% lower token cost than other harnesses across six benchmarks.
Ant Group consolidates Alipay operations in big bet on AI ‘agentic commerce’
Ant Group consolidated Alipay operations into a single business group, betting on ‘agentic commerce’ for growth.
Rabbit Is Back, This Time With an AI Agent App
Rabbit’s OS3 is a cross-platform agent app for desktop, Telegram, and iMessage, two years after the R1 hardware’s poor reception.
What’s New for Game Developers: DLSS 5 with 3D-Guided Neural Rendering, NVIDIA ACE Updates, and New RTX Kit Capabilities
NVIDIA DLSS 5 adds 3D-Guided Neural Rendering as a final neural stage, with NBA 2K27 cited as an early adopter.
Sovereign AI Overcomes Compliance Challenges And Feeds Innovation
Sovereign AI gives enterprises full control over data and infrastructure to meet compliance mandates within specific geographic boundaries.
ByteDance launches Dramagic, a full-pipeline AI platform for producing short dramas from script to screen
ByteDance launched Dramagic, an AI platform that handles the full pipeline for producing short dramas, from script to video preview.
How UK AISI and EvalEval Are Making Benchmark Results Reproducible
UK AISI and EvalEval are building a shared schema and Evaluation Cards platform to make benchmark results reproducible and interpretable.
Study: Young users (9 to 18Y) ditch Google for AI, with unknown consequences
A review of 173 studies found no clear evidence that GenAI strengthens or weakens schoolchildren’s critical thinking.
No stories match. Active filters.
Follow this briefing: Daily digest RSS.