Three models on June 26 - per OpenAI's announcement: Sol at $5/$30 per M tokens, Terra $2.50/$15, Luna $1/$6 - but access is limited to roughly 20 US-cleared orgs under a White House executive order. The 30-day review window closes July 2, the real date to watch before broader API access opens. Read our take →

An 880 trillion won, 10-year plan for memory, logic, OLED, and AI data centers inside South Korea, the largest single company-level infrastructure bet on record. Supply for the chips your training runs depend on just got a new floor. Read our take →

Meade spent seven years on Apple's spatial-computing and smart-glasses hardware and now joins Jony Ive's device team, OpenAI's fourth senior Apple hardware hire. The AI-device race is being staffed by the people who built the last one. Read our take →

A June 10 letter describes an April-to-June model-distillation campaign Anthropic calls the largest known, routing China-origin calls through proxy services in permitted jurisdictions. If you sell API access, this is the threat model your abuse team now defends against. Read our take →

GA for all paid ChatGPT plans on June 25, with one-to-one QR pairing and a plugin that provisions a Droplet, sets up SSH, and attaches it as a persistent workspace so long jobs survive a laptop sleep. If you run long agent tasks, this removes the babysitting. Read our take → Try DigitalOcean →

The Ad Network and Boosts unlock at $43/mo Scale, and the free Launch plan caps you at 2,500 subscribers, so the switch decision is really about when your list crosses that line. Rated 4.4. Read it → Try it →

One Advanced AI step with two tool calls now bills 9 tasks, so the per-run math changed. Bring your own model key to keep steps at 1x. Rated 4.1. Read it →

With AI Agents GA, Make is now the cost-efficient pick if you want agent loops without Zapier's bill. Rating climbed from 4.2 to 4.4. Read it → Try it →

Copilot switched to AI credits on June 1, and Pro's budget does not survive a million-token session. Pro+ at $39/mo is the daily-driver tier. Rated 4.2. Read it →

Gartner's warning via The Register: heavy agentic use can push a solo dev's tooling bill past $2,000/mo. We ran the numbers per team size - solo dev to Cursor Pro, 5-person startup to Cursor Business, 50-person org caps token spend first. Read it →