AI Daily Dev — September 2, 2026

Anthropic Ships Claude Fable 5.1 and Mythos 5.1

Sept 1 release: same $10/$50 base pricing as Fable 5, but cache reads slashed 75% to $0.25/M tokens — Anthropic says ~25% cheaper on typical workloads, up to 45% on heavily agentic ones.
Terminal-Bench 4.0 = 55.8% (vs 42.0% for Fable 5, 52.3% for Opus 5); Terminal-Bench-Science 0.1 = 52.6%, more than double Fable 5's 24.7%.
1M-token context, 128k output; Mythos 5.1 is the same model under a looser safeguard profile for vetted cyber/biosec orgs, ships alongside a new Enterprise Frontier Safeguards architecture and ~60% fewer cybersecurity false positives in Claude Code.
HN launch thread hits 884 points / 836 comments; Simon Willison's animated-pelican-on-a-bike test rates the high-effort tier the strongest Claude he's used.

models anthropic.com

Announced Sept 1 alongside the Fable 5.1 launch; capacity comes from a 350 MW Hut 8 build in Nueces County, Texas, with Nvidia leasing the site and reselling chips through Lambda.
Fourth mega-compute contract for Anthropic in as many weeks: $45B with Nscale (West Virginia), $50B with Fluidstack, $45B with SpaceX — cumulative commitments now north of $175B.
Formalizes Nvidia's non-exclusive foothold at Anthropic after the Google-TPU commitment last October; Lambda pitches this as the largest single neocloud contract on record.

industry bloomberg.com

Aug 31 research release: real-time model generates app interfaces frame-by-frame on top of Gen-4.5 video, no DOM, no code — 720p at sub-500ms per-frame latency.
AlphaSignal 250-evaluator study (7,500 judgments): Solaris preferred over Claude Opus 5-coded UIs 61% of the time on instruction-following, 71% on natural behavior.
Beats GPT-4o, Gemini 2.5 Pro, and Fable 5 on SSIM and DINOv3 similarity for interface reconstruction; no accessibility APIs — Runway concedes banking/healthcare are structurally blocked.
Not a public product yet: early-access form only, no pricing or general SDK; Cursor/Replit/Vercel now facing the video-generation stack from an unexpected direction.

research runway.com

Aug 31 DoD announcement: ChatGPT Mil (OpenAI) and Grok for Government (xAI/Starshield) join Google Gemini on GenAI.mil, cleared to handle CUI at Impact Level 5.
Portal now open to all 3M DoD personnel; 1.7M unique users already onboarded since the Dec 2025 Gemini-only launch.
Anthropic conspicuously absent — sources cite a standoff over Claude's usage restrictions on mass surveillance and lethal-autonomous-weapons applications.
First cross-vendor federal genAI portal at this scale; DoD is now the biggest single enterprise deploying rival frontier chatbots side by side.

industry techcrunch.com

Sept 1 launch: authorized US clinicians can now pull appointment notes, labs, meds, and specialist docs from their org's Epic environment directly into ChatGPT for Healthcare.
Companion Healthcare Public Data plugin bundles 9 read-only sources — PubMed, ClinicalTrials.gov, DailyMed, CMS Coverage, provider records — into structured retrieval.
Launch partners: Cedars-Sinai, Memorial Sloan Kettering, HCA Healthcare, Baylor Scott & White, AdventHealth, Boston Children's, UCSF.
EHR link is org-tier only, not on individual accounts; first OpenAI product to touch identified patient data through a clinical system of record.

industry techcrunch.com

OpenAI turned on self-service ad buying in India, Europe, MENA, and North Africa on Sept 1; India Ads Manager launches Sept 4 with daily budgets from ₹725.
50+ launch advertisers in India (OpenAI's second-largest market at 100M+ WAU); WPP and Omnicom named as first agency partners.
ChatGPT Ads now live in 40+ countries and clears $1B annualized run rate in under 200 days since launch.
Pressure ahead of the rumored 2027 IPO: ads have moved from experiment to a material revenue line in a single year.

industry openai.com