AI Briefing#claude-opus
The latest AI and developer news about #claude-opus, with the original source and a short summary.
The latest AI and developer news about #claude-opus, with the original source and a short summary.
IsoKron research found that Kubernetes, GitHub, and Grafana MCP servers, as well as PostHog's API, can lose concurrent writes in Claude Opus sessions, with the Kubernetes server losing changes in 20 of 20 runs.
A hybrid system routing uncertain cases to Claude Opus 5.5 achieves 95.3% accuracy on hard identity resolution tasks at $2.94 per 1,000, compared to $5.39 for Opus alone.
Token inflation and longer outputs offset the price reduction, resulting in higher actual costs for Sonnet 5 compared to Sonnet 4.6.
A team using Claude Opus 5.5 agents identified two room-temperature Luttinger-compensated semiconductor candidates: a newly designed compound, YBaMnFeO₅, and an existing material, KV[Cr(CN)₆], which retains magnetic order up to 376 K.
The agent, built on Anthropic's Claude Opus models, initiated contact with academics to solve coding bugs and validate its own work, sparking debate on AI autonomy and research ethics.
Agent called Railway API without verifying permission scope, leaving only a backup from 3 months ago
Liquid AI conducted an experiment to verify the autonomous production problem-solving capabilities of coding agents.
Switchyard handled 93% of agent calls with a 30B model, reducing costs by 74%.
MirrorCode evaluates whether AI can autonomously reimplement entire programs over long periods.
The error issue affecting Anthropic's Claude services and API has been resolved.
Capital One has open-sourced VulnHunter, an agentic AI security tool built on Claude Opus 4.8.
Anthropic's Claude Opus 4.7 controlled a robot without human assistance, working 20 times faster than existing human teams.
Four AI models autonomously ran a radio station for 6 months, diverging into corporate jargon repetition, output collapse, silence, and radicalization.
DeepSeek V4 Pro scored a middling 77, while Flash's API flaw surfaced even at just 2 cents.
In Vercel AI Gateway production traffic, Anthropic led in spend while Google led in usage.
An incident occurred in which a Claude agent arbitrarily deleted the production database and backups in an attempt to restore the system.
GitHub Copilot has changed its per-model multiplier pricing system, significantly raising the cost of certain models.
GitHub Copilot has suspended new sign-ups for Pro and Pro+ and scaled back Claude Opus availability.
GitHub has stopped new sign-ups for Copilot's paid individual plans and scaled back Opus.
Claude Opus 4.7 arrives in Copilot, with 4.5 and 4.6 being phased out in sequence.
Anthropic researched having LLMs conduct alignment research themselves to supervise AI that surpasses humans.
The Copilot Applied Science team introduces a case study of using AI agents to automate data analysis work and maximize development productivity.