AlphaGenome Atlas Maps Effects of 9 Billion Human DNA VariantsOpenAI Solves Navier-Stokes Problem with Unreleased ModelMIT Researcher Uses GPT-5.6 Sol for Autonomous Quantum Lab WorkOpenAI Releases ChatGPT Images 2.5 with Improved Instruction-FollowingVercel Sandbox Routing Now 18x Faster with Regional ReplicasVercel Introduces Flat Rate CDN Pricing for Predictable CostsOpenAI Launches $5M Grant Fund for AI and Teen Development ResearchTerence Tao: AI Is Mining Finite Pool of Good Open ProblemsOpenAI Explores How More Capable, Affordable AI Expands Human WorkFirst GPT-6 Model Released, Creator Shares What They're BuildingClaude Code v2.1.266 Fixes LLM-Gateway Routing RegressionOpenClaw 2026.9.3 Update Enhances Safety with Isolated RehearsalsOpenAI Codex Releases Version 0.154.0-alpha.8AlphaGenome Atlas Maps Effects of 9 Billion Human DNA Variants

Google DeepMind unveiled AlphaGenome Atlas, a predictive map charting the molecular effects of 9 billion single-letter DNA variants across the human genome. This resource aids researchers in understanding the link between genetic mutations and disease.

OpenAI, using an unreleased model, has reportedly produced a resolution to the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize Problems with a $1,000,000 prize, unsolved since May 24th, 2000. The result is described as impressive.

OpenAI released ChatGPT Images 2.5, its latest image generation model, which has already produced over 3 billion images across ChatGPT Images and the GPT-Image models in the API. The new version improves multi-turn instruction-following, responds faster, and better remembers user preferences. Developers can integrate the updated model via the API for use cases like iterative design edits and brand asset creation that require back-and-forth refinement.

Vercel announced that routing to Sandbox public domains is now 18x faster. Domains created via the sandbox.domain() SDK call are now resolved from the nearest regional replica instead of a single centralized store, significantly reducing the latency for incoming requests to reach the process running in the sandbox.

Vercel has introduced a Flat Rate CDN pricing model. Following a year of Pro pricing adjustments, customers consistently cited unpredictable CDN costs as a pain point, especially due to viral traffic spikes. The new pricing aims to make costs more stable and predictable.

OpenAI launched a $5 million grant program supporting independent research on how generative AI affects teen development, well-being, and safety. The initiative is open to external research teams and aims to provide objective evaluation of AI’s impact on underage users, rather than internal self-assessment. Findings are expected to inform evidence-based guidelines and regulatory recommendations for youth-facing AI products. Applications are open now.

Mathematician Terence Tao recently wrote that the pool of good, fruitful open problems is now being mined in a non-renewable fashion, potentially leading to scarcity. He noted that even the rumor of someone working on a problem can trigger a massive amount of AI activity on it.

OpenAI published an analysis on how more capable, affordable AI expands the scope of work for individuals and businesses while making growth more economical. The piece argues that as AI capabilities rise and costs drop, tasks once deemed too expensive to take on become viable, spawning new workflows and business models. It points to a shift from marginal optimization to scaled expansion of AI applications.

Ben’s Bites reports on the release of the first GPT-6 model, along with insights into what the author is building with it. Specific benchmark data and project details are yet to be disclosed.

Claude Code released version v2.1.266, fixing a regression introduced in v2.1.265 that affected LLM-gateway and proxy setups. The issue caused the undocumented CLAUDECODEUSE_GATEWAY environment variable to force Cloud-gateway sign-in on its own when the other two variables were not set.

OpenClaw released version 2026.9.3. This update introduces safer update mechanisms by rehearsing core and plugin changes in an isolated candidate state before activation. It also supports eligible 2026.9.2 migrations and can recover abandoned update records without stopping a healthy matching Gateway.

OpenAI Codex released version 0.154.0-alpha.8, following closely on the heels of 0.154.0-alpha.7. Specific details on new features have not yet been disclosed.