The public watchdog for AI tools. We catch every quiet price hike, limit cut, and silent model swap your vendors make - written up, dated, and sourced the same day. 32+ receipts on 23 vendors and counting.
Beyond the receipts, we ship the tools to do something about it - the Plan Optimizer finds what you're overpaying for, the Email Decoder translates vendor PR back into plain English, and NerfBench catches model swaps the day they happen.
Each row is a documented vendor move with a primary source, a plain-English summary, and a severity score. Click any row for the full receipt.
Most receipts are minor. Some aren't. This one cost users real money - silently, overnight, with no announcement. This is what the watchdog exists for.
On June 1, 2026 Copilot shifts from request-based to usage-based billing. Devs widely framed it as shrinkflation -- same price, less effective output.
AI jury verdict: 3 of 3 called this harmful (avg 65/100) · 2 primary sources cited · Vendor response: explanation (Apr 27, 2026).
Cells show today's score out of 10 with a 7-day trend. Cheap-tier models on purpose - the prices most teams actually pay.
Most paid AI stacks have 2-3 subscriptions with a free or open-source equivalent that does the same job. Tell us what you pay for; we tell you what to drop, what to keep, and what to swap in. Free. No signup. Math runs on your machine.
Most paid AI stacks carry 2–3 subscriptions with a free or open-source equivalent that does the same job. Drop in what you pay for and we'll surface the cheapest stack that still covers your usage - then ship a step-by-step migration playbook to switch over.
Math runs in your browser. Nothing leaves your machine. Median Personal-plan saves $31/month on the first run; teams average $190/month across 5 seats. Free, no signup.
Every tool we ship is free to read and run. They share one job: turn the data we collect into a thing you can actually do on a Tuesday afternoon - cancel a sub, decode a vendor email, switch IDEs without losing your config, or catch a model swap before your prod errors do.
Tell us your paid AI subs; we find the cheapest free / OSS stack that still covers your usage. Math runs in your browser.
Paste the "we're improving our pricing" email. We translate corporate-speak into what they actually did to you, with citations.
Step-by-step instructions to move off the platform that just nerfed you - config, keys, prompts, history - in under an hour.
The first public, daily behavioral benchmark for frontier models. When a vendor silently swaps a model, our score moves before the screenshots do.
Side-by-side scorecards: limits, model defaults, comms history, Nerf Index, response time. Pick the next one without buyer's remorse.
Plain-English Q&A grounded in our full receipts archive. "Did Replit just change billing again?" - get a sourced answer in 2 seconds.
Reading every receipt and using every tool is and always will be free. Paid plans add private API access, alert webhooks, the per-vendor risk index for procurement, and quarterly NerfBench / Compare reports. No vendor money. No ads.
Read every receipt, run every tool, and watch one vendor. No signup, no card, no tracking.
For devs who want a heads-up before their bill goes up - unlimited alerts, saved profiles, full diff history.
A procurement layer for your squad. Shared watchlists, hourly scans, contract-grade scorecards across 5 seats.
Build on our data. Firehose API, jury verdicts, NerfBench scores, embeddable widgets - programmatic and remixable.