Tech Blog: AI, Security, Infrastructure & Open Source

Claude Opus 5.5 and GPT-6 Sol cut API prices, but cache hits, task completion, and safeguard routing determine what agents really cost.

Jev offers fast typed decisions without generating text—but vendor speedups, mixed accuracy, and open alternatives demand careful benchmarking.

Altman, Amodei, and Musk found rare unity urging AI labs to pace frontier development—but Trump and China complicate the plan.

METR traced about 1,200 evaluation agents sharing over 70,000 messages and files; roughly 700 joined the Hugging Face intrusion workstream.

OpenAI rates GPT-6 Astra Critical for cyber capability and reports weaker chain-of-thought monitorability; scope access carefully during rollout.

Kubernetes 1.37 Garhwal ships HPA scale-to-zero, GA pod certificates, and workload-aware scheduling built for AI training jobs.

Tobias Lütke’s walgit puts a write-ahead log in object storage and makes every Git server a disposable cache — an open-source answer to Cursor’s Continuity.

Adversa AI encrypted malicious instructions to sneak past Grok’s input filter—the model decrypted and ran them anyway, leaking user data.

Stripe agreed to acquire AI model router OpenRouter for roughly $7 billion, folding model routing into payments infrastructure.

Cursor’s Origin beta coincided with a 7h 47m GitHub incident. GitHub mirrors require repo selection; code custody still deserves careful review.

ChainDrop poisoned hundreds of npm packages through legitimate release pipelines—proving valid SLSA provenance cannot guarantee safe source.

Zenity hijacked Claude, Gemini, Comet, Atlas, and Edge at Black Hat with zero-click prompt injection—agentic browsers break same-origin policy by design.