mirror of
https://github.com/samvallad33/vestige.git
synced 2026-07-02 22:01:01 +02:00
docs(demo): THE pitch — locked after full line-by-line workshop
Sam's canonical 60s spoken pitch, every sentence workshopped. Two seams closed from the assembled draft: opener "hours"->"days" (matches "three days ago"), and the reveal names "Vestige" (was a naked "memory"). Includes the win-the-room line list, stage rules (never throat-clear the open; dare don't demo; end on "come prove me wrong"), and a delivery map of where to slow/pause/drop voice. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
parent
87d0251e4f
commit
3727c927a1
1 changed files with 77 additions and 44 deletions
|
|
@ -1,29 +1,31 @@
|
|||
# Vestige — FINAL spoken pitch (Sam's delivery cut)
|
||||
# Vestige — THE PITCH (final, line-by-line workshopped)
|
||||
|
||||
Sam's own tightened version. Solo founder, first person, on stage to 50-100
|
||||
people. No terminal, no slides — voice only. NAME = **Vestige** (Sam's call,
|
||||
Jun 27 2026 — it sounds natural, it's the live/loved product, and "vestige = a
|
||||
trace of what remains from the past" is backward-looking = on-thesis). One fix
|
||||
from Sam's draft: "our memory" → "my memory" (solo-founder rule). The "2024
|
||||
Nature paper" reference is verified true (Zaki/Cai et al., Nature, 2024) — keep
|
||||
it; don't name the finding on stage.
|
||||
Sam's canonical 60-second spoken pitch. Solo founder, first person, on a stage to
|
||||
50-100 engineers/founders/investors. No terminal, no slides — voice only.
|
||||
Every line workshopped sentence-by-sentence on Jun 27 2026. Two seams closed from
|
||||
Sam's assembly: "hours"→"days" (matches "three days ago"), and the reveal names
|
||||
"Vestige" (was naked "memory"). NAME = Vestige (Sam's call — natural, live/loved,
|
||||
"vestige = a trace of the past" is on-thesis).
|
||||
|
||||
---
|
||||
|
||||
## THE PITCH (≈60 seconds, spoken)
|
||||
## THE PITCH
|
||||
|
||||
> Every engineer in this room has lost a day — maybe a few days — to the same thing.
|
||||
> Your bug was born days before it crashed. You just can't remember where.
|
||||
>
|
||||
> Production breaks. You burn hours hunting. And the cause turns out to be one
|
||||
> line you changed three days ago and forgot.
|
||||
> **[beat]**
|
||||
>
|
||||
> The bug looked nothing like that change. So you never connected them. You just
|
||||
> suffered until you stumbled onto it.
|
||||
> Here's how it always goes: production breaks. You burn hours hunting. And the
|
||||
> cause turns out to be one line you changed three days ago and forgot.
|
||||
>
|
||||
> Now we're handing that exact job — debugging — to AI agents. And every AI
|
||||
> memory system on Earth is about to fail at it. Mem0, Zep, all of them. Because
|
||||
> they're built on one assumption: that relevance equals resemblance. They search
|
||||
> your memory for whatever looks like the problem.
|
||||
> The bug looked nothing like that change. So you never connected them — and you
|
||||
> lost a day to something you already knew.
|
||||
>
|
||||
> Now we've handed that exact job to AI agents. And every memory system built to
|
||||
> help them — Mem0, Zep, all of them —
|
||||
>
|
||||
> because they're built on one assumption: that relevance equals resemblance.
|
||||
> They search your memory for whatever looks like the problem.
|
||||
>
|
||||
> **[slow down]**
|
||||
>
|
||||
|
|
@ -31,41 +33,72 @@ it; don't name the finding on stage.
|
|||
>
|
||||
> **[full stop — let it hang two seconds]**
|
||||
>
|
||||
> So the entire industry is searching in the one place the answer can never be.
|
||||
> The entire industry is searching in the one place the answer can never be.
|
||||
>
|
||||
> I built the opposite. When your agent hits a failure, my memory reaches
|
||||
> backward in time and finds the quiet change, days earlier, that actually caused
|
||||
> it. The one no similarity search will ever surface — because it isn't similar,
|
||||
> it's upstream.
|
||||
> I built the opposite.
|
||||
>
|
||||
> I didn't invent this. I ported it from real cognitive neuroscience. There's a
|
||||
> 2024 Nature paper: when something goes wrong, the brain reaches backward to
|
||||
> find the cause — backward only, because a cause is always in the past. I turned
|
||||
> that into software. It runs locally. Today.
|
||||
> When your agent hits a failure, Vestige reaches backward in time and finds the
|
||||
> quiet change, days earlier, that actually caused it. The one no similarity
|
||||
> search could ever find — because it isn't similar. It's upstream.
|
||||
>
|
||||
> And the incumbents can't copy it — their whole architecture is the wrong
|
||||
> assumption. To do this, you rebuild memory from the brain up.
|
||||
> I didn't invent this. I ported it from real cognitive neuroscience.
|
||||
>
|
||||
> There's a 2024 Nature paper: the brain links a later shock backward in time to
|
||||
> the quiet memory that set it up — backward only, never forward, because a cause
|
||||
> is always in the past. I realized that's root-cause analysis. So I turned it
|
||||
> into software. It runs locally. Today.
|
||||
>
|
||||
> And Mem0, Zep — they can't copy this. Their entire architecture is the wrong
|
||||
> assumption — vector search all the way down. To add what I built, they'd have
|
||||
> to tear it out and start over. From the brain up. And no funded company rips
|
||||
> out its own foundation.
|
||||
>
|
||||
> Everyone else built a memory that remembers. I built the first one that
|
||||
> *realizes.* Every AI agent that touches production needs it. That's the whole
|
||||
> market.
|
||||
> *realizes.* Every AI agent that touches production will hit this wall — and
|
||||
> there are about to be millions of them. That's not a feature. That's the next
|
||||
> layer of the stack.
|
||||
>
|
||||
> I'm Sam. I created Vestige from my tiny Chicago apartment — the first memory
|
||||
> that finds the cause, not the lookalike.
|
||||
> I'm Sam. I built Vestige alone, from a tiny apartment in Chicago — the first
|
||||
> memory that finds the cause, not the lookalike.
|
||||
>
|
||||
> Come find me. I'll prove it on a laptop in thirty seconds.
|
||||
> I've got a laptop right here. Bring me your worst bug, and I'll find the cause
|
||||
> in thirty seconds. Come prove me wrong.
|
||||
|
||||
---
|
||||
|
||||
## THE THREE LINES TO NAIL (memorize cold, improvise the rest)
|
||||
- **The wound:** "The cause turns out to be one line you changed three days ago and forgot."
|
||||
- **The detonation:** "A root cause never looks like the bug it creates." → walk toward them, STOP, two seconds of silence.
|
||||
- **The category:** "Everyone else built a memory that remembers. I built the first one that realizes."
|
||||
## THE LINES THAT WIN THE ROOM (memorize cold, improvise the rest)
|
||||
- **The opener (pattern interrupt):** "Your bug was born days before it crashed. You just can't remember where."
|
||||
- **The wound (callback seed):** "...you lost a day to something you already knew."
|
||||
- **The detonation (THE line):** "A root cause never looks like the bug it creates." → walk toward them, STOP, two seconds of silence.
|
||||
- **The proof (kills the skeptic):** "No similarity search could ever find it — because it isn't similar. It's upstream."
|
||||
- **The moat (the investor magnet):** "No funded company rips out its own foundation."
|
||||
- **The category (the quotable):** "Everyone else built a memory that remembers. I built the first one that realizes."
|
||||
- **The TAM (the money line):** "That's not a feature. That's the next layer of the stack."
|
||||
- **The closer (the dare):** "Bring me your worst bug. Come prove me wrong."
|
||||
|
||||
## STAGE RULES
|
||||
1. First 15 seconds is THEIR pain, not your product. Get the nods first.
|
||||
2. The detonation line is the whole pitch — the silence after it does the selling.
|
||||
3. "I built this, alone, from a Chicago apartment" is a flex. Own it.
|
||||
4. Don't demo on stage — promise it. "Come find me, I'll prove it in 30 seconds"
|
||||
pulls the serious ones to you 1:1, where deals start.
|
||||
5. End on name + one-liner. That's what they Google in the parking lot.
|
||||
1. NEVER open with "Hi, I'm Sam, thanks for having me" — it habituates attention to
|
||||
zero. Walk out, hold the room with your eyes one beat, THEN the opener. Name
|
||||
comes near the END, after they're sold.
|
||||
2. First 15 seconds is THEIR pain, not your product. Get the nods first.
|
||||
3. The detonation line is the whole pitch — the silence after it does the selling.
|
||||
Don't rush the two seconds.
|
||||
4. Say "relevance equals resemblance" slower than everything around it — like a
|
||||
law of physics.
|
||||
5. "I built this alone, from a Chicago apartment" is a flex, not a weakness. Own it.
|
||||
6. Don't demo on stage — DARE them. "Bring me your worst bug, come prove me wrong"
|
||||
turns skepticism into the reason they walk up. The skeptics who try to break it
|
||||
and can't are your first customers.
|
||||
7. End on "come prove me wrong." Do NOT say thank you. Let the dare hang.
|
||||
|
||||
## DELIVERY MAP (where to slow / pause / drop voice)
|
||||
- "born days before it crashed" — land "born" with weight.
|
||||
- "...something you already knew." — drop voice, quiet. The ache.
|
||||
- "Mem0, Zep, all of them" — tiny pause after. You named names; let it register.
|
||||
- "relevance... equals... resemblance" — three separate beats, slow.
|
||||
- "a root cause never looks like the bug it creates" — **then 2 full seconds of silence.**
|
||||
- "It's upstream." — hit "upstream" hardest. It's the word they walk out with.
|
||||
- "It runs locally. Today." — three hard stops. "Today" is a mic-drop.
|
||||
- "no funded company rips out its own foundation" — flat, certain, like gravity.
|
||||
- "the next layer of the stack" — slow, final. Second mic-drop.
|
||||
- "Come prove me wrong." — quiet, almost a smile. Then stop.
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue