• Ars Technica AI (rss) · 6 gundur sessiz
• HuggingFace Blog (rss) · 8 gundur sessiz

Introducing ChatGPT Images 2.5

• <p><strong><a href="https://openai.com/index/introducing-chatgpt-images-2-5/">Introducing ChatGPT Images 2.5</a></strong></p> OpenAI's image generation models are apparently used "more than 3 billion images across ChatGPT Images and the GPT‑Image models in the API". This latest release improves thei

Meta's AIRA3 autonomous agent placed 8th out of ~4,000 human teams in an NVIDIA Kaggle reasoning challenge, and Artificial Analysis just made agentic benchmarks harder to game with real workflow tests.

• Meta's AIRA3 (autonomous research agent) won Gold, placing 8th/~4,000 teams in an NVIDIA Kaggle challenge to improve a 30B model's reasoning, using async multi-agent coordination via shared forum/filesystem with no central controller
• AIRA3 generalizes across domains with only task-spec changes: delivered 27% latency reduction on production GPU kernels and Gold-level translation of 4,000-year-old Akkadian tablets
• Artificial Analysis updated its benchmark suite to include harder, real business-style agentic workflow tests, making it tougher to game leaderboard scores
• SuperAstra: a GPT-6-powered tool lets users mod live SNES games via plain-English prompts (open source, ScottStevenson/SuperAstra) - real-time machine code editing driven by natural language
• Flint AI's 'Switch' lets you drop any AI agent into Slack/Teams/Discord in minutes with no migration or platform lock-in
• New retrieval model uses table-of-contents structure to hit 82.6% recall with near-zero hallucination

OpenAI's 10,000-agent swarm cracked a 90-year-old Navier-Stokes problem in 88 hours, while shipping a faster, cheaper image API most studios will actually use tomorrow.

• OpenAI used 10,000 agents on an unreleased model to solve an open Navier-Stokes blow-up question, with the proof formally verified in Lean — but credit/attribution controversy remains unresolved
• OpenAI shipped ChatGPT Images 2.5: Flare model gives higher quality than GPT-Image-2 at 50% lower latency; Sunburst model targets precise local edits; both priced at 2x GPT-Image-2 API rate, live now with sketch input support
• Mistral raised €3B (Series D) at a €21B valuation — the largest European tech equity round ever — earmarked for 1GW of EU compute capacity by 2030 and regional data-residency options for 125+ enterprise clients including Airbus and HSBC
• Sony AI's open-source hypothesis-ranking tool correctly predicted two real aging-related gene discoveries out of 1.5M candidate hypotheses, signaling AI moving from assisting research to generating novel scientific claims autonomously
• Google DeepMind released predicted-impact maps for all 9 billion possible human DNA mutations, and OpenBMB shipped MiniCPM5-2B, a 2B-parameter model built specifically for on-device deployment

OpenAI's unreleased model solved a Millennium Prize math problem using 10,000 agents, while Meta's new Muse agent and IFA's commercial-ready robots show AI physical/agentic products are outpacing trust and cost barriers.

• OpenAI used an internal model 'significantly more capable than GPT-6 Astra' with ~10,000 agents to solve the Navier-Stokes equations, a Millennium Prize Problem, reportedly racing Anthropic who was independently close to solving it
• Meta launched Muse, a free consumer AI agent (up to 100M tokens/week) that handles tasks, browses web, and makes purchases, but requires deep access to calendars, email, and payment platforms amid Meta's history of privacy scandals and an $18B child-safety settlement
• IFA 2026 showcased commercially available humanoid robots (Xiaomi CyberOne at $99,999, Unitree G1 at $13,500) that box, dance, and climb stairs, signaling hardware is closer to mainstream despite steep pricing
• Nvidia's Omniverse/Cosmos and Fei-Fei Li's World Labs are closing robotics' data-training gap via simulated data, while consumers show growing willingness to share real-time data (e.g., smart glasses) for better AI
• New creative tools shipped: OpenAI's ChatGPT Images 2.5 (sharper editing/generation), Krea Realtime Director (real-time video direction), and Runway's new Team Plan for small studios
• Cognition raised $2B at $48B valuation (AI coding), Mistral raised $3.5B Series D, signaling continued heavy capital flow into foundational AI tooling

Apple's decade-long strategy of invisible AI in iPhone cameras is a more instructive product blueprint than its flashy new Siri, even as agentic hackers and robotics standardization reshape the AI infrastructure landscape.

• Apple has shipped ML-powered camera features every year since 2016 (Portrait Mode, Smart HDR, Deep Fusion, Photonic Engine), treating photos as 'moments to enhance' rather than generative remixes—unlike Samsung/Google's more liberal generative editing
• iOS 27 pushes Apple's photo boundary further with Spatial Reframing, Extend (generative edge-filling), and upgraded Clean Up for removing large objects, testing its own 'no fake pixels' philosophy
• Arm launched a Robotics Capability Framework (SAE-style levels 0-5) and expanded Arm Total Design to 80+ companies (AWS, Hugging Face, Unitree) to standardize fragmented physical AI development; market seen growing from $25B toward one of computing's largest TAMs
• Google's Threat Intelligence Group Q3 report: attackers used multi-agent frameworks to autonomously scan infrastructure and execute a full credential-harvesting attack in under 6 hours post-breach—a shift from passive infostealers to 'offensive agentic harvesting'
• Other signals: Thinking Machines Lab in talks to raise $6B at $40B valuation; Zuckerberg lobbied Trump against a national AI regulator; Gemini 3.8 Flash and Meta's Muse Image (image editing, #4 on Artificial Analysis leaderboard) shipped