PARTLY

Partly true: every data source in the 'vibe code your way into buying a plumber' reel is real and public, but the AI run it shows matched only a fraction of the list, and the hard parts (headcount, owner status, the deal) are still guesswork and legwork

Facebook · Oct 2026

The claim'Apparently anybody can just vibe code their way into acquiring a boring business.' Prompt an AI to download the SBA PPP files, filter for plumbing businesses, find each one's state registration and keep the ones registered 30+ years ago, check current headcount 'by pulling this form' (Form 5500 on screen), have a small decision model read every review to judge whether the owner still works there, then email or mail every owner offering to buy. Opens on Codie Sanchez: 'Codie Sanchez is a $10k workflow.'

Every source on screen is real and free. The SBA publishes loan-level PPP data at data.sba.gov (the ppp-foia dataset, files dated 240930), and since a December 2020 court order it includes borrower names, addresses, NAICS codes and jobs reported for loans of every size. Texas lets anyone search franchise tax accounts and charter dates, with a public API. The Labor Department publishes Form 5500 filings. TypeSafe's Jev is a real decision model on OpenRouter. So an agent really can chain these. The catch is in the reel's own frames. The AI session admits name matching is 'approximate', that Form 5500 only exists for firms with a benefit plan, and that 'only 85 rows got a headcount' out of 1,219 Texas plumbing loans. PPP data is a 2020 to 2021 snapshot, many family plumbers never file a 5500, and reviews rarely say whether the owner is retiring. What you really get is a rough prospect list, not acquisition targets. Writing to owners is legal: business mail is fine, and cold email must follow CAN-SPAM (honest headers, a real postal address, a working opt-out). The poster is selling nothing but follows, says outright the method is old and only the AI glue is his, and the opening dig at Codie Sanchez (Contrarian Thinking, BizScout) is a joke about paid acquisition communities, not a factual claim.

What holds up

Contact sheet read frame by frame: data.sba.gov 'Paycheck Protection Program (PPP) FOIA' page with public_150k_plus_240930.csv and up_to_150k files, a chat session ('ran 16 commands') reporting 1,219 Texas NAICS 238220 loans narrowed to 271 loans across 215 companies, 'Only 85 rows got a headcount', 'Matching is approximate', then the Texas Comptroller Franchise Tax Account Status Search, the DOL EBSA Form 5500 dataset page, TypeSafe Jev 1.13, review screenshots and a drafted letter.
data.sba.gov/dataset/ppp-foia lists 13 CSVs plus a data dictionary, last modified Oct 21, 2024; the Dec 1, 2020 release under the Washington Post and Center for Public Integrity lawsuit added names and addresses for loans under $150k.
TypeSafe's Jev decision model (announced Sep 15, 2026) is listed on OpenRouter as typesafe/jev-1.13 and returns probabilities, not prose; the 'owner still works there' YES/NO cards match that format.
Form 5500 is filed only by employers with a retirement or welfare plan (5500-SF for plans under 100 participants), so most small trade shops have no headcount there; the reel's own output says the same.

What doesn’t

'This prompt will do all of that' hides the yield: the on-screen run matched a small share of the list and called its own matching approximate.
Headcount and owner status are inferred from 2020 to 2021 loan data, benefit plan filings and review text, which is a lead score, not due diligence.
Mass email to owners is legal only inside CAN-SPAM rules; repeated letters are legal but the reel treats volume as the strategy.

The catch

The data pipeline is real and cheap to build; the reel undersells how much falls out at each join and skips the part where an acquisition actually happens: conversations, financials, SBA 7(a) or seller financing, and a quality of earnings check. AI gets you a better list faster, nothing more.

How to actually do it

Download the PPP FOIA files from data.sba.gov, filter NAICS 238220 (plumbing and HVAC) in your state, and dedupe borrowers by name and address.
Join to your state's business registry (Texas: Comptroller franchise tax search or its open data API) on legal name, keep charters 25+ years old, and hand check every fuzzy match.
Treat headcount and owner signals as a score: PPP jobs reported, Form 5500 participants where filed, and review mentions; then send a short, honest letter by mail and CAN-SPAM compliant email (real address, opt-out) to the top 50.
When an owner replies, switch to real diligence: three years of tax returns, an SBA 7(a) lender or seller note, and a broker or attorney before any LOI.

PPP loan data, state charter searches, Form 5500 filings and the Jev decision model are all real and public, so the AI chain shown can be built. Its own output shows thin match rates and proxy headcounts, so the result is a prospect list, not an acquisition, and outreach must follow CAN-SPAM.

Confidence: High
Posted by: a tech and business comedy creator (this reel ~117k views); reel sent to Buddy Mon Oct 5, 2026, ~12:41 PM

See the original claim →

We test hype for free. We build the real thing for a living.

Thirty minutes, no pitch — and you'll leave with something useful either way.

Book a call with Todd or start with the free Business Checkup →

The Verdict Weekly

Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.