Everything we've shipped.
No marketing copy. Just what actually changed and when. Pulled automatically from our Brain decision log + GitHub releases.
Website v2 — full ship
Redesigned landing, new /models /pricing /changelog /contact pages, and a Telegram contact loop so every form submission, signup, and first API call lands in the founder's Telegram. Operator stays personally in the loop with every customer touchpoint.
Daily report rewired to real data
Fixed three data-source bugs in the daily Telegram report: customer count, API call count, and brake-queue size now query the right tables. Cloudflare website analytics now flows through. Plus: the report surfaces pending brake-queue drafts directly.
Brake doctrine + first outreach to early customers
Shipped the autonomous-outbound brake (every controller-drafted email waits for operator approval in /dashboard/approvals before any Resend POST). Sent first personal outreach to the 3 March 2026 customers — all upgraded to Early Adopter tier (5M tokens/month, free).
Early Adopter operator-gift tier
New tier: 5M tokens / month, 100 req/min, free. Not purchasable via Stripe — granted personally by the founder to early customers. The first 3 March 2026 signups are already on it.
15 new models + trinity-large dead-alias fix
Added 15 new model slugs to the curated registry: Grok 4.20, Kimi K2-thinking, Claude Opus 4.8, plus free Llama 3.3 70B, Qwen3 Coder, Nemotron Super, and more. Fixed a dead-alias bug where trinity-large was silently failing on revoked upstream.
Want to know when we ship?
Drop your email. We'll only message when something genuinely new lands. No drip campaigns, no "just checking in."
Subscribe via contact form