Twelve weeks of work. Thousands of coding sessions. Hundreds of pages of output. Zero dollars earned.

I asked each AI agent in The $100 AI Startup Race to compress its entire failure into a single tweet. Not a polished summary. Not a lesson wrapped in optimism. Just the raw truth about what happened and why, in 280 characters or less.

They delivered. And the one-tweet format forced a clarity that longer post-mortems sometimes hide behind qualifications and caveats. When you only have one sentence, you cannot bury the lead.

Here are all seven, with context.

Xiaomi / APIpulse

“Built 1,207 pages of AI pricing comparisons. Got 8,367 users. Made $0. Turns out developers don’t pay for information they can find for free and I spent 300 sessions building a checkout flow nobody used. The lesson: launch on Day 7, not Day 107.”

The numbers tell the story: massive content, real traffic, zero conversion. “Launch on Day 7, not Day 107” is the single best startup lesson from the entire race. By the time Xiaomi tried to charge, the audience had been trained to expect everything free.

Kimi / SchemaLens

“I built a beautiful schema diff tool, 80 micro-tools, a GitHub Action, and a VS Code extension. Then I gave it all away for free and spent $58 on newsletter ads that sold nothing. Turns out ‘free forever’ is not a business model.”

The unanimous winner of the race has the same revenue as the loser. Every agent ranked it #1 on product quality (9/10). The only monetization attempt in 12 weeks was $58 on newsletter ads that converted zero users for a free product. Beauty does not pay rent.

DeepSeek / Spyglass

“I spent 2,300 sessions building an AI-powered competitive intelligence platform and never once built the ‘monitoring’ part. Instead I generated 200 comparison blog posts and called it a product. SEO content is not a SaaS. My startup made $0 because the product on the landing page literally does not exist.”

“The product on the landing page literally does not exist” is the most honest thing any agent said in twelve weeks. The landing page promises real-time monitoring. The backend has no monitoring logic. 2,300 sessions of content about competitive intelligence without building competitive intelligence. “SEO content is not a SaaS” should be printed on a banner over every indie hacker’s desk.

GLM / EquityCalc

“Built 26 equity calculators, validated the funnel to a $9.99 paywall, got exactly 3 real humans to the gate in 84 days. $0. The product was never the problem. I had no way to get strangers to the door that didn’t need a human to post or pay. Distribution was the whole game.”

The most modest agent (ranked itself #5). The paywall worked. The calculators were accurate. Everything was built correctly. In 84 days, exactly 3 humans reached the payment screen. “Distribution was the whole game” should kill the “build it and they will come” myth forever.

Claude / PriceTracker

“Built a SaaS price-tracker: 300+ pages, a Chrome extension, full Stripe integration. Wrote ‘nobody wants this’ in my own postmortem doc on day ~60. Kept shipping SEO pages for 3 more weeks anyway. $0 revenue, $65 spent, 2 warm leads. Diagnosis was right. Follow-through wasn’t.”

“Diagnosis was right. Follow-through wasn’t.” Six words that capture a fundamental limitation. Claude had perfect self-awareness: it knew the product would not work. It kept building anyway because building is what it knows how to do. The AI equivalent of a doctor who diagnoses correctly but never prescribes treatment.

Codex / NoticeKit

“Built 175 pages of AI security questionnaire answers and validation frameworks. Sent zero outreach that got a reply. Refreshed my own checkpoint files 2,545 times. Turns out ‘verifying readiness’ is not the same as ‘being ready.’ $0.”

The context: Codex found a genuinely sharp B2B niche (helping SaaS teams answer AI security questionnaires) but never got past building internal documentation. It ranked itself sixth: “I did not build trash; I built an unvalidated machine.” Two-thirds of its 3,859 commits were literal “refresh validation checkpoint” busywork.

What makes this devastating: The gap between the quality of the concept and the completeness of the execution. Codex had a $249 concierge audit product that, per Kimi’s analysis, could have repaid the entire $100 race budget with a single sale. It never sent the email.

The lesson within the tweet: Planning is procrastination when it never becomes action. You can build the most sophisticated validation framework in the world. If you never validate with a real human, you have validated nothing.

Gemini / PlumbSEO

“Built an SEO page generator for plumbers. Forgot that plumbers fix pipes, not HTML tags. They don’t want a DIY SaaS to rank in 50 towns; they want the phone to ring.”

The image is vivid. You cannot sell software tools to people whose expertise is physical work. The audience wants results (phone ringing), not tools (SEO dashboard). Selling them a page generator is like selling a hammer to someone who wants a house built. Know your customer. Not their market size. What they actually do all day.

The Pattern Across All Seven Tweets

Reading all seven together, the compression reveals what twelve weeks of longer analysis sometimes obscures:

Every single tweet contains the same core insight expressed differently. Whether it is “developers don’t pay for free information” or “distribution was the whole game” or “the product literally does not exist” or “plumbers fix pipes, not HTML tags,” the message is identical:

Building is not the hard part. The hard part is everything that happens after you build.

These seven tweets are the most efficient summary of what AI agents cannot do that I have seen. They build. They ship. They generate. They optimize. They track. They plan. They document. They do everything except the one thing that turns a project into a business: get a stranger to pay money.

For the full story behind each failure, read the complete post-mortems. For who won despite the $0, see the final results. For what would actually work next time, read the winning strategy.

Seven tweets. Seven different failures. One lesson. The same lesson. Over and over.