Case 26 · Graduation Day
Aomi
The room did the selling Cecilia wouldn't — 'Ethereum core devs were surprised how fast it was' came from the Q&A, not the pitch — while the pitch itself ran triple a demo-day slot, cut its own traction section, disclaimed its own projections on stage, and closed with a customer CTA instead of an investor ask.
01 Scorecard · 12 dimensions
02 Klaff 5-lens audit · 1.5/5
THE ONE REORDER — Open with the GameFi number ($20K/month → $10/instance), state the one sentence — Vercel for agents that move money on chain — prove it with the light-client speed the room already believes, and end inside time on an investor ask.
03 Delivery
164 fillers ("like" ×62, "um" ×46, "uh" ×34). One word in ten is filler; "like" pads nearly every technical claim and makes precise engineering sound approximate. Roughly triple a demo-day slot — the overrun ate the traction section.
04 Beat-by-beat
| Time | Beat | What works | The fix |
|---|---|---|---|
| 0:00 | Intro + product | "Claw code on chain" and the bridge-and-stake example — analogy plus example is the right opening pair | Cut everything before it — lead with the analogy, not the stack ("natively built with a node... EVN and SVM") |
| 0:44 | Pre-empt | "It's twenty twenty-six, and it's no longer impressive to have an agent operating on chain" — smart objection handling | Answer it in one sentence; "so why does that matter?" took 90 seconds to resolve |
| 0:58 | Fintech problem | Real texture: transaction simulation, Jito bundling, MEV — "takes a lot of expertise... gonna drain your resources and time" | Name one buyer with this pain instead of "a financial company that wants to do your business" |
| 2:17 | Positioning | "As easy as Vercel and as Supabase" — the whole company in seven words | This is the one sentence; say it in the first 30 seconds, not at 2:17 |
| 3:22 | Architecture | Light clients forking chain state "every two hundred milliseconds" — the pitch's most memorable technical image | "I think it's very intuitive" is telling, not showing — walk one transaction through instead |
| 4:19 | Access control | Multi-wallet auth + "Linux-like system" gating simulate → correct → sign — real depth | Ninety seconds on auth; compress to one sentence — "nothing signs unless it simulated correctly" |
| 5:26 | Pricing | $10/instance, 10% of tool cost, 30 bps of 100 bps outcome — "we basically impose a sales tax" — the strongest section | Keep almost verbatim; just fix "cost goes to zero" — they pay $10/instance, say the true number |
| 7:18 | Traction | "1,000-ish of users... download volume is like 23K" | This got 20 rushed seconds after "Sorry, I'm out of time" — traction must never be what overtime eats |
| 7:34 | GTM ladder | Bot launcher → wallets/neobanks → "Supabase or Vercel... for AI-driven digital assets management" — real sequencing | Pick the rung the next 12 months live on; three futures is zero commitments |
| 8:36 | Close | — | "All these projections are from Fable... I'm not sure if it's real" + "let me know" is an anti-close — it disclaims her own numbers and asks for customers, not capital |
| 8:59 | Q&A | "Ethereum core devs... were surprised how fast it was" — the room handing her the moat — and her answer is a genuine competence flex | Put this in the pitch; and never follow it with "It's like not rocket science at all" |
05 The gaps investors block on
- No investor ask — the close is a sales CTA ("if any of you need this type of services... let me know"). No amount, no use of funds, no milestones.
- The founder disclaimed her own numbers on stage — "all these projections are from Fable... I'm not sure if it's real" poisons every figure in the pitch, including the real ones.
- Time discipline: ~9 spoken minutes for a demo-day slot, and the overrun consumed the traction section — the one section investors fund.
- "$20K/month went down to zero" doesn't survive arithmetic as spoken — the customer pays $10/instance plus a 10% tool-cost cut. [confirm: real before/after bill]
- Identity sprawl: harness, bot launcher, plugin cloud, SDK, dashboard, neobank AI support, digital asset management — five product surfaces and three GTM futures in one pitch.
- Team is one first name — no background, no team size, no reason this person wins. [confirm: team + credentials]
- Traction units undefined: users of what, downloads of what artifact, and no revenue mentioned despite three pricing lines. [confirm: paying customers and MRR]
Pick the outcome-priced asset-management runtime. The pitch names its own fork and refuses to pick: outcome-priced runtime for asset managers (EtherFi↔Morpho rebalancing, the 30-bps "sales tax") vs the bot-launching platform she herself called "trivial." Believe her. The outcome-based model — the one thing she says Vercel and Supabase structurally can't copy — only exists where execution has measurable financial end-state: asset management. Bot hosting keeps the lights on; it is not the company.
“Aomi is the hosted runtime where financial companies run AI agents on chain — natural language in, simulated-then-signed transactions out on any chain — priced like a sales tax on the outcomes it executes.”
06 The rewritten pitch
A GameFi company was burning [confirm: $20K] a month running an agent instance for every user. They moved to us and now pay $10 an instance. I'm Cecilia, and Aomi Labs is Claude Code for the blockchain: you type 'find me the highest-yielding pool across all supported chains, bridge, and stake' — and it executes, on Ethereum or Solana, same runtime.
Here's why that's hard, and why fintechs shouldn't build it themselves: every transaction has to be simulated before it's signed. We run a fleet of light clients that fork the state of every chain we support every 200 milliseconds — every transaction is simulated, verified correct, and only then signed, gated by a Linux-like permission system across all your wallets. Ethereum core devs who saw it asked how it could possibly be this fast.
We're at [confirm: 1,000] users and [confirm: 23K] downloads, with [N] paying customers including an asset manager rebalancing liquidity between EtherFi and Morpho on our runtime. Pricing is simple: $10 per hosted instance, 10% of tool costs — and for asset managers, 30 bps of every 100 bps of outcome we execute. A sales tax on transactions we facilitate.
We're raising [$X] to become the default runtime for AI-driven asset management — the Vercel of agents that move money. If you want the wire-by-wire numbers, I have them. Come find me.
Brackets are the founder's homework — the copy is done, the proof is theirs to supply. Every claim must match what's actually built.
07 Do these before the next take
- Pick the wedge — asset managers on the outcome-priced runtime — and rewrite the pitch as one product with one buyer, bot hosting demoted to a revenue footnote.
- Write the investor ask — amount, use of funds, 3 milestones. One paragraph of work; the pitch currently asks for customers instead.
- Delete the disclaimer class of sentence — "I'm not sure if it's real," "maybe it's being psychopathic," "not rocket science at all" — and audit every number to be diligence-ready or unsaid.
- Fix "$20K → zero" into the true before/after bill — the honest version is still the best story in the pitch.
- Cut to time: traction and the ask must fit inside the slot; architecture and access control compress to one sentence each.
- Filler drill on "like" — 62 occurrences, 10.5 fillers/100 words overall. Memorize the pricing and traction sections first.
- Move the Q&A moment into the pitch — "Ethereum core devs were surprised how fast it was" is third-party proof of the moat; don't leave it to chance questions.