Case 08 · Graduation Day
Pitch Rotator
The recursion was the whole opportunity — a pitch-practice app pitched live is the one demo that must never fail, and it failed on stage ("Okay, I failed. One second.") while the sharpest thinking (one-good-one-bad, the 18-session self-experiment) only surfaced in Q&A; a real number (600 users in five days) is buried before the product is named, and there is no ask, no model, no market.
01 Scorecard · 12 dimensions
02 Klaff 5-lens audit · 1.5/5
THE ONE REORDER — Make the recursion structural: run the actual 90-second loop as the pitch, let the product show its one-good-one-bad card on this very pitch, and put the ask on that screen.
03 Delivery
36 fillers ("like" ×18, "um" ×9, "uh" ×7), clustered in the unrehearsed Q&A answers. The product measures exactly this — the number belongs on the screen next time.
04 Beat-by-beat
| Time | Beat | What works | The fix |
|---|---|---|---|
| 0:00 | Meta open | "I'm gonna pitch for my pitch app" — the room is instantly on her side | Commit to the recursion: start a live loop instead of joking about one. Cut "I lost 15 seconds" — never narrate the clock |
| 0:35 | Traction flash | "After 50 different videos… 600 users in five days" — the best number in the pitch | It lands before the room knows what the product is, and it's unanchored — 600 users of what? Move it after the product beat [confirm] |
| 0:55 | Problem | The words that convince a user or investor — the right problem, plainly said | Add one concrete moment: the pitch that failed and what it cost |
| 1:15 | Context claim | Feedback connected to your latest build is the differentiator-shaped ambition | "Moved to an… through an MCP that is deployed on a TE" is word salad. One rehearsed sentence or cut it |
| 1:45 | Restart | "Hi, I'm Albi" — honest, likable | The founder is introduced at the midpoint as a mic-failure recovery. The name goes at 0:00 or in the close, on purpose |
| 2:00 | The loop | "The whole product is one loop… one screen that is like a camera" — clearest product moment | Attach the one-good-one-bad mechanic here instead of leaving it to Q&A |
| 2:20 | Live demo | The instinct to demo live is right | "Okay, I failed. One second." — on this product a failed demo is self-disproving. Pre-run the loop and play the result card |
| 2:40 | Close | The narrated fallback ("it's gonna find your best thing that you said") is a clean feature description | It's also the close — a description of what the screen would have shown, with no ask |
| 3:00 | Q&A | The best material of the day: one-good-one-bad rationale, 90-second reasoning, "I did 18 sessions in one go" | This is the pitch. Promote all three answers into the body |
05 The gaps investors block on
- No ask. The pitch ends on a feature description and the Q&A ends on "open to feedback and suggestions." The room cannot act.
- The demo failed inside a pitch about pitching under pressure — the room's take-home memory competes between "600 users" and "Okay, I failed," and the failure came later.
- The differentiation claim is unrepeatable as spoken ("through an MCP that is deployed on a TE") — unsayable means nonexistent.
- "600 users in five days" is unanchored — users of what, doing what, retained how long? [confirm]
- Business model, market math, and competition framing are absent outright.
- The pitch's best thinking (one-good-one-bad, 18 sessions, 90-second rationale) was extracted by audience questions instead of being the spine.
Pitch the loop; earn the context layer. Two products were on stage: the 90-second iteration loop (demoed in words, defended twice in Q&A, validated on herself with 18 sessions) and the context-connected private evaluator via MCP (the bigger, moat-shaped claim — spoken once, garbled, never shown). Pitch what survived contact with the audience: the loop. The context layer becomes one clean roadmap sentence until the day it can be demonstrated, at which point it graduates to the headline.
“Pitch Rotator is the 90-second camera loop where early-stage founders rehearse the sentence that sells what they're building — one take, one thing you nailed, one thing to fix, again.”
06 The rewritten pitch
I'm Albi, and yes — I'm pitching my pitch app. So let's use it. (loop starts on the screen behind her) Every founder rewrites their one sentence a hundred times. I made 50 videos in [timeframe], treating each one as a product, and the version that finally worked brought [confirm: which product] 600 users in five days. What changed across those 50 videos was never the code. It was the words.
Pitch Rotator is a 90-second camera loop. You pitch alone, to the lens, and it hands you exactly two things: the best line you said, and the one thing to fix on the next take. Not a wall of feedback — chat models already give you that, and you apply none of it. One thing, because one thing is what you can fix in the next 90 seconds.
I ran 18 loops in one sitting and watched my own pitch sharpen every round — the pitch you're hearing is loop [N]. [If live and demonstrable: It also reads your repo, so the feedback is about your product, not pitching in general.]
Since launch: [N loops run, M founders, retention]. I'm asking for [the ask: $X / N design-partner founders / intros] to [milestone]. (the screen shows the loop's verdict on this very pitch) — And that's my one thing to fix. Your turn.
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 — the loop, with the context layer as roadmap. This gates the rewrite, the demo, and the moat story.
- Write the ask — amount or design-partner count, use, one milestone. The pitch currently has zero.
- Anchor the 600 — users of what, in which five days, retained how long. Unanchored it dies to one follow-up question. [confirm]
- Make the demo un-failable — pre-run result card as primary, live loop only when venue-rehearsed.
- Promote the Q&A into the pitch — one-good-one-bad rationale, the 90-second reasoning, and the 18-session experiment are the spine.
- One clean sentence for the context claim or cut it — nine words or nothing.
- Filler drill on unrehearsed answers — 36 fillers, 4.3/100, "like" ×18 clustered in Q&A. Run the answers through the product.