What KRTR saw at the seed stage, and what happened next.
We ran six now-famous early decks (LinkedIn, Dropbox, Facebook, Airbnb, Uber, and WeWork) through the exact pipeline that scores any inbound pitch. The point isn't the number. It's that the gaps KRTR flagged at seed became the same issues that defined each company's real history.
How KRTR reads a deck
The score is the last thing KRTR produces. Before it, the engine does the work an analyst would do in the first week of diligence, except it finishes in minutes and shows its sources.
It separates what's missing from what to ask
Information gaps are things the deck failed to evidence, and they pull the score down. Diligence questions are what you raise in the room. KRTR keeps them apart, so you know which is a scoring penalty and which is a conversation.
On the Dropbox deck: 20 information gaps logged, and a separate list of 11 diligence questions, 4 of them marked critical.
It fills the gaps it can, from live evidence
When the deck omits something checkable, KRTR goes and finds it instead of just penalizing the blank, then folds the answer back into the analysis with its source.
The Facebook and Uber decks named no founders. KRTR surfaced Zuckerberg and Moskovitz, and Kalanick and Camp, from public evidence, and adjusted the team read once the gap was filled.
It builds the market size itself
When TAM is missing or inflated, KRTR estimates its own from outside sources rather than taking the slide at face value, and calls out the difference.
Airbnb's deck conflated gross bookings with revenue; KRTR grounded a real figure. WeWork's flexible-workspace TAM was validated at $52.3B against independent research.
It finds the competitors the slides left out
Founders rarely list the people who threaten them. KRTR names them from the live landscape, usually where a 'first to market' claim quietly falls apart.
Airbnb's “first to market” met Couchsurfing, which already had 670k users. Dropbox's “unclaimed territory” met Carbonite, Mozy, and box.net.
It grounds every claim and corrects overclaims
An AI reviewer checks each claim against evidence and, when sources disagree, shows the disagreement instead of smoothing it into one confident answer.
Facebook's “$85B purchasing power” was flagged as conflating total student spend with the actual addressable ad market.
It's built not to drift
A single name search can invite a confident, tidy backstory assembled from things that were never true. KRTR gates every statement and every web search to the specific company and its named founders, so the model can't wander to a similarly-named entity. Each report keeps its claims as a ledger: what the deck asserted, separated from what the reviewer verified.
The Dropbox report tracked 29 claims (26 from the pitch, 3 added by the reviewer), so you can always see where a statement came from.
It estimates a base-case valuation
When the deck states no number, KRTR produces its own base-case valuation from comparable transactions and market data, so you start the conversation with a grounded figure instead of a blank.
Across all six, every report carried an AI-estimated valuation, from Dropbox at seed to WeWork at growth, each derived from the evidence, not the pitch.
KRTR isn't built on one AI model. It benchmarks and routes across a range of models under the hood, matching each pipeline task to whichever one reads it best, and falling back automatically if one is unavailable.
That's also why KRTR compounds: every time a foundational model improves, KRTR gets better for free. The frontier moving is a tailwind, not a migration.
Two reads: the universal one, and yours
The scores on this page are the universal read. They measure evidence, gaps, and risk the same way for everyone. That's why they tell you what's shaky, but they aren't a verdict on the company. A 51 doesn't mean "pass"; it means the deck left specific things unproven.
Universal read
What's evidenced, what's missing, where the risk sits, the same for every investor. Tells you what's true about the deck.
Your investor lens
KRTR re-scores each deck against your own thesis. The same company ranks differently for a biotech seed fund than a fintech growth fund. Tells you what's a fit for you.
The universal read is for spotting gaps and risks. The lens is what turns KRTR into thesis-match screening at the top of your funnel, so a pile of decks sorts itself against what you actually invest in.
From gut feel to evidence
Deal volume is exploding and every deck now reads beautifully. Telling real signal from generated polish is exactly where hours disappear. KRTR's job is to let you walk into the first meeting already knowing the gaps, the risks, the real market, and the fit, so the meeting is for judgment, and the decision can follow right after it.
How to read this. These scores aren't a verdict on the companies. They're a read on what each deck evidenced at a single moment. What the exercise shows is consistency: across six decks spanning consumer, marketplace, SaaS, and real estate, KRTR surfaced the specific, auditable risk that later became each company's defining chapter, and it correctly named team and network effects as the strengths where they existed.
A note on this historical showcase. KRTR grounds every analysis against the live web, so the competitor and market data in the full reports reflects today's landscape rather than each deck's original era. That's exactly the behavior you want on a live deal; on these vintage decks it simply means a few named competitors postdate the pitch.
Every report here was generated fresh by the current pipeline. Forward any deck to pitch@analyze.krtr.ai and read your own in minutes.