Sergei Arbo

A month ago I shipped Verdict (https://verdict.tools), a registry that tracks the life-status of AI and SaaS tools: alive, price-hiked, degrading, sunset-announced, or dead. Every status carries a date and a source, reachability is re-checked twice a day from two servers, and a "dead" verdict is only ever set by a human.

The registry now holds 18 confirmed deaths and sunsets, plus a set of dated decline signals on tools that are still alive. Cataloguing them turned up patterns I did not expect. Sharing the five that surprised me most.

1. The two-day goodbye is real

Xembly, a Seattle "AI chief of staff" that raised about 20M USD and counted Salesforce and Twilio among customers, told users on May 30, 2024 that the service would end June 1, 2024. Two days. Anything not downloaded by May 31 was gone.

The epilogue is grimmer: the lapsed xembly.com domain now hosts an Indian cricket betting site. If your workflow depends on a venture-backed tool, your real SLA is however fast you can export.

2. An acqui-hire is usually a death sentence for the product

The pattern repeats with minor variations:

  • Rephrase.ai -> Adobe (Nov 2023): signups closed, wind-down announced Feb 2024, site now fully offline
  • Play.ht -> Meta (Jul 2025): product shut down Dec 31, 2025
  • Papercup -> RWS (Jun 2026): RWS bought only the IP; the standalone product is gone
  • Clockwise -> Salesforce (2026): product unavailable since March 27, 2026

The press release always says "we're joining X to continue the mission." The product's mission ends anyway. When your vendor gets acqui-hired, start the migration that week.

3. The consumer app dies first; the API gets a stay of execution

OpenAI shut down the Sora app on April 26, 2026, but the Sora API keeps running until September 24, 2026. That five-month gap is the migration window for developers, and it is closing.

The same staging shows up elsewhere: user-facing surfaces are expensive to run and easy to kill; contracts and integrations buy the API extra months. If you build on APIs, the app's death is your early warning.

4. Deaths come in clusters

Last week alone: BitMEX (an 11-year-old exchange!) announced on July 23 it will close permanently on September 23, 2026, and BitMart followed on July 26 with trading ending August 26. Back in 2023 it was the newsletter wave: Revue, TinyLetter, Mad Mimi within months of each other.

When one notable player in a category folds, check the rest of the category. The forces (margins, regulation, platform shifts) rarely stop at one company.

5. "Best tools" listicles rot silently

This whole project started because listicles do not get updated. Half the "best AI tools 2024" articles still recommend products that no longer exist. A recommendation without a verification date is folklore.

That is the one design rule I care about most: every status on Verdict shows when it was last verified, and the checks run twice a day whether anyone is watching or not.

The data

The full dated census is here: https://verdict.tools/mortality - with migration guides for the dead tools and sourced signal pages for the shaky ones. There is also a free status API (https://verdict.tools/developers) and an RSS feed of status changes if you want the deaths as they land.

It is my own project and still early. If you have been burned by a tool dying on you, I would genuinely like to hear the story - and if you know a shutdown I have missed, tell me and I will verify and add it.