Anyone can call a winner after it wins. The harder test, and the only one worth trusting, is whether your judgment holds up before you know the ending. So we are starting a series that runs that test in public. We take a famous company's early fundraising document, score it with the same rubric we use on the plans founders send us today, and we do it blind: no peeking at what happened next. Then we check the tape.
First up is the one everybody thinks they know. The 2008 AirBed & Breakfast seed deck. Fourteen slides, a $500K ask, a web marketplace to book a room with a local instead of a hotel.
The read, before the ending
We scored it exactly as it read in 2008. Here is where it landed.
Verdict: Lean yes, 3.80 out of 5. Fundable on vision and model. The gap is hard demand.
A Lean yes is not a slam dunk, and the deck did not earn one. It earned a clear "fundable, not yet proven." The scorecard shows why.
| Dimension | Weight | Score | The one fix that moves it most |
|---|---|---|---|
| Team & founder-market fit | 20% | 4 | They lived the problem; the team slide is just bios. Make the fit explicit. |
| Traction & evidence of demand | 20% | 3 | Replace analogy with own numbers: bookings, repeat rate, revenue. |
| Problem & urgency | 15% | 4 | Quantify how often travelers hit the price and connection pain. |
| Market & timing | 15% | 4 | Build the market bottom-up; the 15%-share figure is asserted. |
| Solution & differentiation | 12% | 4 | Sharpen why this beats Craigslist on trust. |
| Business model & unit economics | 10% | 4 | Clean 10% take rate; add a unit-level example. |
| Competition & defensibility | 4% | 4 | Name the network-effect moat directly. |
| The ask & use of funds | 4% | 4 | Already milestone-tied: $500K to 80k transactions. |
The single weak line is traction, and it is weak for a specific reason. The deck argues demand by analogy. It leans on Craigslist and Couchsurfing numbers instead of its own bookings. That is a reasonable move when you have almost nothing else, and it is also the most load-bearing soft spot in the pitch. No red-flag gate fired, because a working product, testimonials, and early press are real signal. But borrowed evidence is still borrowed. A 3.80 is what this document actually earns.
Now the tape
Airbnb raised a $600K seed from Sequoia in early 2009, at roughly a $2.4M valuation, close to what the deck's own ask implied. The company is public now and worth north of $100 billion.
Here is the part worth sitting with. A model that read these fourteen slides and returned "obvious Fund" would be lying to you, because the only way to produce that answer is to already know how it ends. Ours didn't know. It scored a Lean yes and flagged thin traction as the weakness, which is precisely the weakness that made plenty of professional investors pass on Airbnb at the time. The rubric was right to hesitate and right to still land on fundable. That is the instrument working, not failing.
The lesson is not "back every marketplace." It is the opposite. Resisting hindsight means you do not over-score a document because you happen to know the founder got rich. You score what is on the page. On the page, in 2008, Airbnb was a strong, clear pitch with one real gap. Saying so out loud, before the ending, is the whole skill.
What this has to do with your plan
We build managed AI agents at Coworkers.Global, and the first one, Alex, reads business plans and applications the way I do. I trained it on how I read, then calibrated it against real investor decisions, including this one. It reads your plan the way we just read Airbnb's: looking for the load-bearing weak spot and naming it while you still have time to fix it. It will not flatter you because your idea sounds exciting, and it will not knock you because it sounds unusual. It scores the page.
If you are applying to YC this cycle, the on-time deadline is Monday, July 27, at 8pm Pacific. The read is free before then. Drop your plan at coworkers.global/ai-business-plan-review and you'll get back where you are strong, where a screener would stop reading, and the single weakest part to fix first. We are early and pre-revenue, so we lead with the quality of that read rather than a customer list we do not yet have. If you think the read is wrong, tell us.
Current view, subject to change
One clean data point does not prove a rubric. Airbnb is a case where resisting hindsight kept us from over-scoring a winner; the more instructive cases are the ones where a deck scores well, gets funded, and still fails, and we will run those too, because a rubric that only ever agrees with the winners is just hindsight wearing a lab coat. What would change how I read this deck is evidence that the traction gap did not actually matter in 2008. Everything I can find says it mattered a great deal, and the company closed it fast once the money was in.
Regards,
Charles Stack