Artificial Intelligence

An Anthropic research model solved a 26-year-old cryptographic mystery in under half an hour: the origin of the NIST elliptic curve seeds. It wasn’t a secret hash, but the output of a 1990s pseudorandom number generator seeded with the system clock.

A Mercor report cross-references the O*NET taxonomy with 29 AI benchmarks and finds that only 208 of the 923 US occupations have at least one task with real coverage. 90% of the country’s paid work hours fall into tasks these tests still don’t measure.

Krea AI unveiled Krea 3, an image editing model that promises to finish an edit in under 8 seconds and charge $0.06 per image. The company compares it directly to Google’s Nano Banana Pro and OpenAI’s GPT Image 2, targeting production workflows where cost and speed matter as much as output quality.

Google DeepMind reports that Gemini 4 Argon sped up the quantum chemistry software xTB by 2.3x, helped migrate 6 million lines of legacy code in Chrome, and cut session preparation time by 40% in the Agent Development Kit. The results, published by Google, show the model applied to real software engineering, not just benchmarks.

Anthropic migrated its entire monorepo from Git to Scion, an in-house version control system, after daily code merges rose from 2,500 to over 30,000 between May and September 2026. The migration took five months and addresses the limitations of Git’s decentralized design in an environment where nearly every client is an always-connected AI agent.

Anthropic introduced Claude Haiku 5.5, its cheapest and fastest model to date, with 75% lower average cost than Haiku 4.5. The launch also brings configurable effort tuning and a price cut on Sonnet 5.5’s cache.