AI Daily Digest · 2026-09-07
🔬 New AI Craft
1. OpenAI hits its "automated research intern" milestone — median researcher now burns $600+/day of inference
OpenAI published an inside view of research acceleration, confirming it has reached the goal, set last fall, of an automated research intern by September 2026: a system that carries out well-defined research tasks under human direction, including tasks that would take a skilled researcher days. It is on track toward a fully automated AI researcher by March 2028. The numbers tell the story: at the start of the year the median researcher used coding agents only modestly; by mid-August the median researcher was integrating agents into daily work, spending more than $600/day of inference at API prices, with the 90th percentile above $7,000/day. Where agentic workflows today are human-driven with agents as assistants, OpenAI's milestone flips it — agents execute multi-day research tasks while humans set priorities and make judgment calls — and it argues such agent-usage metrics should be measured openly and shared as industry standards.
https://openai.com/index/research-acceleration-view-inside-openai
🛠️ Tools & Tips
1. huashu-mac-use — an Agent Skill that lets coding agents drive native Mac apps
Gives any agent (Claude Code, Codex, etc.) computer-use control over macOS apps that have no API: reads the UI state, writes without disturbing, and keeps per-step forensics.
https://github.com/alchaincyf/huashu-mac-use
⭐ Open Source Highlights
1. ai-evaluation-framework — 3-axis benchmarking for model-based solutions
Scores accuracy, p95 latency and cost against per-field ground truth, turning "is this model swap worth it" into a reproducible decision.
https://github.com/dreamers-laboratory/ai-evaluation-framework
📰 Industry News
1. OpenAI publishes essay "An Alien Mind": machine intelligence may be heading into recursive self-improvement — alignment tech alone is not enough, voluntary slowdowns and international coordination needed
2. CNBC: "Model fatigue" sets in as Anthropic, Meta, Google and OpenAI ship model versions at a frenetic pace, overwhelming enterprise adopters
🚀 Major Releases
(no items this day)
