GDPx is the new benchmark to watch — but its real story is what it reveals about agent work
How SemiAnalysis’s GDPx works, why Opus 5.5 leads, and what the domain split says about agent benchmarks.
AI & Machine Learning
KernelPost section: AI & Machine Learning
OpenAI’s science push looks bigger in the paper trail than in the proof trail
OpenAI posted a technical preprint and a science-framing post, but the captured pages don’t verify the theorem or proof details.
Anthropic's S-1: The Compute Margin Reversal, $637 Billion in Infrastructure Bets, and the Governance Architecture No Shareholder Can Override
Anthropic's S-1 reveals a Delaware public benefit corporation controlled in part by an independent Long-Term Benefit Trust — an unusual governance structure for a Nasdaq listing.
A 10-year-old CNN beat frontier LLMs on 200 skin lesions — but not on triage safety
A 200-lesion dermatology test found a fine-tuned Inception v3 beat two frontier multimodal models, but the result is narrow.
Nodecode’s persistent Lisp runtime turns one agent into a live program
How Nodecode’s SBCL-based, stateful coding agent works, what the launch demo shows, and what remains unproven.
Hamburg Privacy Regulator to Oversee Public-Sector AI Under EU AI Act from January
Hamburg’s privacy authority is reported to oversee public-sector AI from January 2027, showing the EU AI Act’s enforcement layer is taking shape.
Gauge CLI Asks Whether Agent UX Can Be Engineered
What Gauge CLI claims about optimizing docs, SDKs, CLIs, and agent-facing surfaces — and why launch-page language is not proof of production value.
DeepSeek's V4.1-Flash-1017 Reached Users Through a Live API Push — With Thin Official Documentation
DeepSeek's V4.1-Flash-1017 snapshot reached users through a live API update first caught by trackers, with no official announcement identified in available reporting.