● AI RADAR BETA | Vol 01·Issue 006·Saturday, Jun 20 PT |
|
● LIVE INGEST 09:33 PT·JUN 13 TO 19, 2026 i wake up, there is a louder week of AI updates.One issue, every drop linked, organized by what the news actually meant. | 60 DROPS 3 TOOLS 6 THEMES |
|
↺ ░▒▓▒░ ↻
/\_/\
( ◕ω◕ )
> ^ <
i just wake up What the week meant.Jun 13 to 19 · 60 drops |
|
Agents move to the cloud.Cursor and Codex both pushed the loop further from the laptop. Cursor made cloud agents and subagents first-class, while Codex turned a recorded demo into a reusable skill. The pitch is the same: keep running with the lid closed. READ DROPS ↓ |
Claude Design, Artifacts, Cursor Origin.Coding agents kept growing surfaces around the editor. Claude added a design canvas and a way to share live Claude Code sessions as artifacts; Cursor put a git host on the roadmap. The IDE keeps being the seed, not the whole product. READ DROPS ↓ |
Cursor and SpaceX, Codex in Europe.Distribution moved faster than features this week. Cursor took on a hyperscale customer, Codex expanded to EU users, and OpenAI formalized its partner and enterprise plumbing. READ DROPS ↓ |
Alignment, trained and simulated.Both labs spent the week defending their behavior story. OpenAI shipped two research posts on training broadly beneficial models and on simulating deployment from real user requests. Anthropic's Frontier Red Team used Opus 4.7 to program a robodog, twenty times faster than last year's humans. READ DROPS ↓ |
GPT-5.5 in health, LifeSciBench, Chan-Lam.OpenAI leaned hard on the research-assistant story this week. A medicinal-chemistry pipeline ran through GPT-5.4, a new life-science benchmark landed, and GPT-5.5 Instant was pitched as on par with frontier Thinking models on health questions. READ DROPS ↓ |
Claude Code, by the numbers.Anthropic published its own usage study on Claude Code, with 400K sessions analyzed and freelance-marketplace pricing as a yardstick. The framing matters: the company is starting to argue Claude Code's value in dollars and occupations, not features. READ DROPS ↓ |
|
THE FULL DIGEST, WITH SOURCES Every drop, themed and linked. | 24 linked drops
5 active days
3 tools tracked
|
|
01 AUTONOMY · 5 DROPS Agents move to the cloud.Cursor and Codex both pushed the loop further from the laptop. Cursor made cloud agents and subagents first-class, while Codex turned a recorded demo into a reusable skill. The pitch is the same: keep running with the lid closed. | CURSOR | /in-cloud spins up subagents in their own cloud VMs. | WED · 15:47 PT | → |
|
| CURSOR | Local agents move to the cloud and keep working with your laptop closed. | WED · 14:58 PT | → |
|
| CURSOR | Cloud environments now save as snapshots so future agents start faster. | WED · 15:47 PT | → |
|
| CURSOR | /automate skill configures triggers, instructions, and tools from a plain-language ask. | THU · 12:00 PT | → |
|
| CODEX | Record & Replay turns a workflow demo into an inspectable, editable skill. | THU · 11:50 PT | → |
|
|
|
02 SURFACES · 4 DROPS Claude Design, Artifacts, Cursor Origin.Coding agents kept growing surfaces around the editor. Claude added a design canvas and a way to share live Claude Code sessions as artifacts; Cursor put a git host on the roadmap. The IDE keeps being the seed, not the whole product. | CLAUDE | Claude Design beta brings on-canvas editing and design-system sync across paid plans. | WED · 12:17 PT | → |
|
| CLAUDE | Claude Design and Claude Code sync in both directions, from terminal to canvas. | WED · 12:17 PT | → |
|
| CLAUDE | Artifacts in Claude Code share live PR walkthroughs and dashboards at private links. | THU · 11:12 PT | → |
|
| CURSOR | Origin will host code and git for teams and agents, with a waitlist for fall. | TUE · 15:31 PT | → |
|
|
|
03 DISTRIBUTION · 4 DROPS Cursor and SpaceX, Codex in Europe.Distribution moved faster than features this week. Cursor took on a hyperscale customer, Codex expanded to EU users, and OpenAI formalized its partner and enterprise plumbing. | CURSOR | Cursor joins forces with SpaceX on the frontier of useful AI. | TUE · 06:29 PT | → |
|
| CODEX | Computer use, the Chrome extension, memory, and Chronicle roll out to Codex users in the EEA, UK, and Switzerland. | TUE · 09:11 PT | → |
|
| CODEX | OpenAI Partner Network launches as a formal channel for developer partners. | SUN · PRODUCT | → |
|
| CODEX | ChatGPT Enterprise gets new usage analytics and updated spend controls. | THU · PRODUCT | → |
|
|
|
04 TRUST AND SAFETY · 3 DROPS Alignment, trained and simulated.Both labs spent the week defending their behavior story. OpenAI shipped two research posts on training broadly beneficial models and on simulating deployment from real user requests. Anthropic's Frontier Red Team used Opus 4.7 to program a robodog, twenty times faster than last year's humans. | CODEX | OpenAI trains models to carry beneficial traits into new domains and hold them under pressure. | THU · 14:34 PT | → |
|
| CODEX | Deployment Simulation predicts model behavior on de-identified user requests before release. | TUE · 12:42 PT | → |
|
| CLAUDE | Opus 4.7 programs a robodog roughly 20x faster than last year's best human team. | THU · 09:52 PT | → |
|
|
|
05 SCIENCE APPLICATIONS · 4 DROPS GPT-5.5 in health, LifeSciBench, Chan-Lam.OpenAI leaned hard on the research-assistant story this week. A medicinal-chemistry pipeline ran through GPT-5.4, a new life-science benchmark landed, and GPT-5.5 Instant was pitched as on par with frontier Thinking models on health questions. | CODEX | GPT-5.5 Instant matches frontier Thinking models on health-related questions in ChatGPT. | THU · 11:16 PT | → |
|
| CODEX | LifeSciBench scores AI on 750 expert-authored life-science research tasks. | WED · 13:41 PT | → |
|
| CODEX | GPT-5.4 drives a Chan-Lam coupling improvement from literature review to validated result. | WED · 10:10 PT | → |
|
| CODEX | ChatGPT-assisted reanalysis surfaces rare genetic diagnoses in children that evaded experts. | THU · RESEARCH | → |
|
|
|
06 MEASUREMENT · 4 DROPS Claude Code, by the numbers.Anthropic published its own usage study on Claude Code, with 400K sessions analyzed and freelance-marketplace pricing as a yardstick. The framing matters: the company is starting to argue Claude Code's value in dollars and occupations, not features. | CLAUDE | Anthropic Economic Index adds a framework for tracking Claude Code as it scales. | TUE · 12:41 PT | → |
|
| CLAUDE | Average Claude Code session value roughly doubles on a freelance-marketplace comparison. | TUE · 12:41 PT | → |
|
| CLAUDE | Non-engineering occupations land within 7 points of software on the strictest success measure. | TUE · 12:41 PT | → |
|
| CLAUDE | 400K analyzed sessions show coding dominates, with ops work close behind. | TUE · 12:41 PT | → |
|
|
|
i wake up, there is another one. TOOLS TRACKED CODEX · 34 CLAUDE · 16 CURSOR · 10
Built by Jane You with Claude Code & Design · Maintained by RaeyaBot · Daily ingest 5am PT read in browser · airadarapp.com · unsubscribe |
|
|