The Probe Kit — the ×1.2 September-14 Check, the Model-Routing Probe Commands, the HY4 Free-Trial Setup, and the Rumor Scoreboard

The companion to the video. Four stories broke in one week — and this kit is the part you can act on, with every claim verified or labeled. PAGE 2 — THE ×1.2 CHECK: Anthropic permanently raises standard weekly Claude Code limits +25% vs the OLD baseline on September 14 while the temporary +50% boost ends the same day — their own follow-up post calls it 'a 17% reduction' vs today. The 60-second check (/usage weekly bar × 1.2) plus the route table: under ~80 stay, 80–100 optimize, over 100 route the overflow — and the precision most coverage misses (it's the weekly ceiling, the 5-hour session cap didn't move). PAGE 3 — THE PROBE: claude-marshmallow-eap and claude-melon-eap surfaced Aug 24 and vanished in hours. The exact copy-paste commands to test whether a hidden newer checkpoint is routing to YOUR account (the Opus 4.6 knowledge probe + the modelUsage routing check), the read-the-result table, and our own live run (routed claude-opus-5, honest don't-know = not in the test group). PAGE 4 — HY4 SETUP: Tencent's 770B/49B-active, 1M-context, Apache-2.0 open model — both free routes (CodeBuddy/WorkBuddy 2 weeks; OpenRouter at $0.834/$2.501 per 1M with the minimal curl call), plus our on-camera receipt: a planted rate-limiter bug fixed in 16.9s for $0.0026. PAGE 5 — THE RUMOR SCOREBOARD: every claim of the week tagged VERIFIED / LEAK / VENDOR with clickable sources — including OpenAI's own Aug 7 and Aug 18 Astra statements that the hype videos skip. Independent kit from Hyperautomation Labs — not affiliated with Anthropic, OpenAI, or Tencent.

By submitting, you'll be subscribed to the Hyperautomation AI Report newsletter — one practical AI email per week. Unsubscribe anytime.

Free. No spam. Unsubscribe anytime.