On February 18th, 2026, a startup called Dreamer came out of stealth with a no-code platform for building AI agents. On March 23rd — thirty-five days later — Meta hired the entire team. We compete in this category, so we studied those five weeks closely. Here's what Dreamer got brilliantly right, what the acqui-hire actually signals, and the half of the business that's still unclaimed.
What Dreamer was
The team was the story before the product was: founders from Stripe, Google, and Meta, a $56M seed at a half-billion valuation, and angels whose names alone generate press. The product deserved the attention too — describe what you want in plain English and Sidekick, an agent that builds agents, assembles it in minutes. Users could build, discover, remix, and publish agents across desktop, mobile, browser extension, and email. It was consumer-grade polish in a category that mostly ships developer tools.
Five launch moves worth stealing
Strip away the celebrity angels and the mechanics underneath were genuinely excellent. Five stand out — we're openly borrowing all of them:
- Share-out without an account. Anything you built could be used by friends who hadn't signed up. Value first, signup second — every shared agent was a growth surface.
- Remix over blank page. The default first action wasn't 'create from scratch,' it was 'customize an existing agent.' Lower activation energy, and every remix minted a new creator.
- Credits for publishing. Publishing an agent to the gallery earned platform credits; being featured earned more. Supply-side flywheel, no cash required.
- 'An extended free trial of a paid product.' Their framing for the beta, verbatim. It set the paid expectation on day one without a fake countdown timer.
- The single-day drop. Founder manifesto, long-form podcast deep-dive, and a wall of coordinated coverage — all in 24 hours, so every aggregator fired at once. No dribbled announcements.
What the acqui-hire actually signals
Read the deal precisely: Meta didn't buy the product. It hired the people and took a non-exclusive license. The product that thousands of users had started building on effectively lost its makers five weeks in. Two conclusions follow.
First, the category is white-hot. When the largest companies on earth are hiring entire teams within weeks of a launch — Dreamer was Meta's third AI team grab in about four months — “normal people building AI agents” has stopped being a niche thesis. The giants just voted on it with money.
Second, and less comfortably for users: platform risk is real, and it lands on the builders. If you'd spent February building your business on Dreamer, March was a bad month. It's a reminder that the durable platforms in this space will be the ones whose builders' incomedepends on them staying — because that's the obligation that keeps a platform honest.
Five weeks was long enough to prove people want this. It wasn't long enough to prove anyone could make money from it.
The half Dreamer never built
Look at the five weeks and notice what's missing: nobody got paid. Dreamer shipped the joy of building — beautifully — but not the storefront, the billing, the payouts, or the machinery that turns a clever agent into someone's income. To be fair, almost nobody has: the big canvas platforms still have no native creator payments, and the open-source stacks bolt monetization on through fragmented third-party marketplaces.
Our version of the bet
We're taking the launch mechanics Dreamer validated — remix-first, credits for publishing, share-out without an account — and pointing them at the outcome Dreamer never reached: builders with revenue. No celebrity angel wall here; our launch asset will be receipts. Twenty-five founding builders, taken hands-on from “here's my workflow” to first money earned, in public.
If Dreamer's five weeks proved the demand, the next proof the category needs is a cohort of ordinary experts getting paid. That's the one we intend to publish.