One layer that watches every model surface — and finds you on any device.
The radar is step one. What we're building: alerts for everything that changes — new models, vanished models, price cuts, quiet nerfs, your own rollout access — delivered wherever you are, with real value every week, not just on launch day. No token reselling, no routing wars: we sell knowing first.
Live now
shippedThe core loop works today: sweep the catalogs every 60 seconds, fire the moment a model answers.
Every model landing next — three carrying live Polymarket odds — swept every 60 seconds.
Email + Chrome push on every tracked drop. SMS on The Signal; the automated phone call on The Call.
154 models on the watchlist — leaked, rumored, announced, just dropped — across 42 labs.
Per-model intel: the leaks, the news wire, the official word, live release odds.
Accounts & the dashboard
liveSign in, pick your models, choose your channels. Every tier gets the dashboard — paid tiers get louder, faster, and more precise.
Pick exactly which models alert you — SOTA-only, every major lab, or a custom list. Free includes it.
Every paid plan fires the moment a drop is confirmed. Chrome push is instant on every plan; free email rides the batch ~15 minutes behind.
"Email me when the odds on GPT-6 cross 80%" — you pick the line: 60, 70, 80 or 90. Get moving before the drop, not after.
Per-token price cuts, context-window bumps, rate-limit changes. A 50% price cut is a money event too.
The full leak-and-rumor feed with sources and confidence grades, across every tracked model.
What moved this week: launches, odds swings, price changes, roadmap leaks — synthesized, five minutes.
Slack, Discord, Telegram, and webhooks — alerts where your team already lives, ready for CI and agents.
The watchers
up nextNobody else watches this layer: the products you actually pay for, not just the APIs. Predictable, daily value between launches.
Alerts when a model appears in — or vanishes from — a subscription lineup, and when plan pricing or limits change. Watching today: the Gemini app and Anthropic’s published model table. claude.ai, Claude Code and ChatGPT sit behind bot walls we’re still clearing — they’re next.
The same prompt suite runs against live models; drift charts answer the eternal question with data. The daily battery has been off since July 22. The personal eval harness answers the same question on your own prompts, at every launch.
Browser extension reads your own usage page locally — nothing leaves your machine — and warns at 80% of your weekly limit.
We diff the ChatGPT / Claude / Gemini web bundles and app binaries for new model strings — scoops from the primary source.
Plug in your own API keys; we probe your account and alert the minute your org gets flagged into a staggered rollout.
Ask us to watch anything — a lab, a Hugging Face org, an App Store string, a keyword in provider catalogs.
Per-model p50/p95 latency and availability, with degradation alerts. "Is it down or is it me," answered.
Everywhere you are
plannedThe alert finds you — on whichever screen you're looking at. Native apps for every platform.
A tiny always-on radar in the menu bar — countdowns, odds, a dot that goes green. Full-screen banner on the big drops.
Native push with time-sensitive delivery, plus lock-screen and home-screen widgets with the next-drop countdown.
The menubar radar, everywhere else — one codebase, every desktop.
You're sitting in claude.ai and a strip appears: "Grok 4.8 is in your picker — 2 minutes ago." The alert at the point of use.
A subscribable calendar feed of expected windows that updates itself as the odds move.
Drop day
plannedWhen it lands, the next question is "is it actually good — for me?" We answer it within hours, not weeks.
Our standard eval suite — coding, agentic, writing, vision — runs within hours of every drop. The scorecard hits your inbox before the press writes a word.
Up to 20 of your own prompts run on every new model and on the model you use now. A judge from a third lab compares each pair blind, and the verdict lands by email with a PDF of both answers side by side.
One composite day-0 number — capability, cost, speed, vibes. The rating people cite.
Per release: what it's good at, quirks, best settings, tool-use notes — and ten prompts to steal. Good with the new thing in 20 minutes.
An auto-generated live page per drop: first responses, evals streaming in, odds settling. The box score of a model launch.
Members vote old-vs-new on blinded outputs in the first hours; "the people's verdict" publishes next day.
Register your machine; we alert only when an open-weights release actually fits it — with the quant and the one-line install.
Store your own API keys; the minute a model drops, one click runs it against your saved prompts.
The network
laterThe data brand: citable, embeddable, everywhere — and the community that feeds the wire.
Queryable history of every launch, price change, and deprecation — for analysts, researchers, and tool-builders.
A three-minute daily audio brief as a private podcast feed. The commute-sized version of the radar.
"State of Anthropic / OpenAI / Google today" — official status in one place now; per-lab what's-next landing next.
Bloggers and newsletters embed the live radar. Every embed is a billboard.
Verified leaks earn subscription credit — a source network for the wire that pays for itself.
Org accounts, shared alert rules, roles, Slack org install.
Alert SLA, priority support, and a quarterly custom eval run on your workload.
A live room that opens for major drops, with our eval results streaming in real time. Once the crowd exists.
What we killed — on purpose.
A roadmap is also what you refuse to build. These came up, and died for good reasons.
The radar is already sweeping.
Everything above lands on top of a loop that works today. Get on the list now — the roadmap comes to you.
Watch free →