Case 04 · Graduation Day

TinyCloud

The clearest idea — "super base for sovereignty," backed by three named cohort teams with the same private-data wall — arrives at 2:35 buried under invented vocabulary, the demo video sells consumer apps while the words sell B2B infrastructure, and the only ask spoken is "We're currently raising, um... And yeah."

Sam4:14 pitchreviewed 2026-08-17

4.5

01 Scorecard · 12 dimensions

Problem

5"Best case… locked into an ecosystem you can't escape. Worst case… manipulated in ways you can't comprehend" is a real stakes line — but the whole problem is abstract fear. No named person feels it today.

Why now

3Agentic AI and brain-computer interfaces are gestured at, never argued. Nothing says why this layer is fundable this year.

Insight / secret

6Two real ideas — the "common shape of problem" across three named cohort teams, and the eject mechanism (a user takes their data and becomes your customer) — neither landed as the secret it is.

Solution clarity

4Spaces, signed messages, a "permission information surface," "super operable apps" — four minutes of internal vocabulary. The room cannot repeat back what TinyCloud is.

Product / demo

6The produced video is the clearest 50 seconds of the pitch — one passkey, an agent you own, approval gating every action. But it demos consumer apps while the words pitch infrastructure.

Moat

2"A hard problem that we have a lot of experience with" is the entire moat argument. Asserted, never shown.

Market

3"We focus on people and organizations" is everyone, which is no one. Zero market math anywhere.

Business model

3"Kind of a B2B to C product" plus "infrastructure as a service" names a shape. Who pays, and how much, is never said.

Competition

4OpenAI/Anthropic as threat, Palantir as contrast, Supabase as analogy — but no answer to who else builds private-data infrastructure or why TinyCloud wins.

Traction

3"We have a customer we've been building for," three built apps, one named partner (Synclave). No customer name, no users, no revenue, no numbers.

Team

2"I'm Sam. I'm the founder of Tinycloud" is the complete team slide. No background, no reason to believe this founder ships hard infrastructure.

The ask

2"We're currently raising, um... And yeah." An ask begun and abandoned mid-sentence — no amount, no use of funds, no milestones.

02 Klaff 5-lens audit · 2/5

Croc brainPARTIALThe open rides an earlier talk's image ("Peter Thiel's bathtub") instead of making its own; the best pattern interrupt belongs to the video's "most personal data set on Earth" line.

FramePARTIALFear flips to a genuine prize frame — sovereignty as technology that serves its user — but the flip arrives via a false start and is never held.

STRONG arcPARTIALThreat → solution → stack → cohort validation → video → ask. Right shape; the ask beat collapses mid-sentence and the proof beat has no numbers.

Hot cognitionPASSThe bathtub, handing a brain-computer interface "the keys," "manipulated in ways you can't comprehend," and the video's approval beat all fire.

StatusFAILFalse starts on the thesis sentence, 34 um/uh in the live sections, and a close that trails into "um... And yeah."

THE ONE REORDER — Open on the three cohort teams' shared wall instead of a borrowed bathtub, put "super base for sovereignty" in the first 15 seconds, and finish the ask sentence.

03 Delivery

Length

4:14

Pace

171 wpm

Fillers / 100 words

6.4

46 fillers (um ×18, uh ×16). The produced video voiceover (~133 words) is filler-free, so the live-spoken sections actually run 7.8/100 — the gap between the video's voice and the live voice is the rehearsal gap, exactly.

04 Beat-by-beat

Time	Beat	What works	The fix
0:00	Intro + one-liner	"Infrastructure for private agentic applications on user-owned data" — a real one-liner exists	It's jargon-dense; the plain version ("super base for sovereignty") doesn't arrive until 2:35. Swap them.
0:09	Fear open	Bathtub + brain-computer-interface imagery is hot	It's borrowed ("as we heard earlier") and abstract. Open on your own observation — the three cohort teams.
0:33	Stakes	"Locked into an ecosystem you can't escape… manipulated in ways you can't comprehend" — best stakes line spoken live	Keep verbatim; deliver it without the four "um"s around it.
0:41	Sovereignty flip	The prize frame appears: technology that serves the person using it	The thesis lands via a false start ("you grant control over-- you gra- you grant authority"). Memorize this sentence cold.
1:15	The stack	Protocol layer → hosting choice (own devices / TinyCloud / Synclave) → apps: a real architecture	"Permission information surface" and "super operable apps" are internal vocabulary. One plain sentence: you own the data, apps get revocable permission.
1:53	Cohort validation	The best evidence in the pitch — three named prospective customers in the room: Conclave, Sanctum, Dealproof	Say their status. Is any of them actually building on TinyCloud? Observed demand and signed demand are different pitches. [confirm]
2:35	Positioning	"Super base for sovereignty" + the eject mechanism — the pitch's clearest 20 seconds	Move it to 0:10. Everything before it should serve it.
3:01	Traction	"We have a customer we've been building for. We built three apps."	Name or bracket. One unnamed customer is a rumor, not traction.
3:10	Demo video	Polished, concrete, filler-free: one passkey, "No platform reads it. No provider trains on it. No silent logging."	It demos consumer apps while the words pitch B2B infrastructure — brand them as reference apps or demo a developer building on TinyCloud.
3:59	Close	The Palantir contrast is a good closing frame: sovereignty for governments vs. "people and organizations"	"We're currently raising, um... And yeah" is the anti-ask. Finish the sentence — amount, use, milestone — then "come chat."

05 The gaps investors block on

The ask was abandoned mid-sentence — no amount, no use of funds, no milestones.
Identity crisis: the words pitch a platform, the video sells TinyCloud Chat and a feed. Two products, no bridge sentence.
Traction is unnamed — one customer, three apps, one partner, zero checkable names or numbers.
No moat beyond biography: "a lot of experience" doesn't answer why OpenAI or the next protocol can't do this.
Vocabulary tax — "permission information surface," "super operable apps" — the room can't retell the pitch on Monday.
Team is one first name; nothing spoken supports "we have a lot of experience."

The one strategic decision

Pick infrastructure — and make the cohort the first customers. The pitch straddles infrastructure ("super base for sovereignty," B2B2C) and consumer apps (the video's chat + feed). The only real evidence points to infrastructure: three named teams with "a common shape of problem" is observed developer demand, and the eject mechanism only pays off at platform scale. The apps become reference apps — proof the platform works — instead of the identity crisis.

“TinyCloud is the super base for sovereignty — the private-data backend agentic apps build on, where the user owns the data, grants every permission, and can walk away with everything — starting with [N — confirm] design partners from this cohort.”

06 The rewritten pitch

Three teams pitching on this stage — Conclave, Sanctum, Dealproof — are building on private data. Transcripts, journals, negotiations. And every one of them hits the same wall: the moment that data touches a backend, it isn't private anymore. That's not their bug. That's a missing layer of the stack. I'm Sam, founder of TinyCloud, and we build that layer — think Supabase, for sovereignty.

Here's the model. Your data lives in a space you control — on your own device, hosted with us, or with a partner like Synclave. Apps and agents never own it. You sign a permission; they get exactly what you granted and nothing else; revoke it and they're blind. Best case today, you're locked into an ecosystem you can't escape. Worst case, you're manipulated in ways you can't comprehend. Sovereignty is the third option: powerful technology that serves the person using it.

It's running. [Customer — confirm name] builds on TinyCloud today, and the three apps we shipped on it — you just saw the chat and the agent feed — run entirely on user-owned spaces: no platform reads it, no provider trains on it, only your key. [Confirm every claim is attestable in the current build.] And the model is B2B2C: any app's user can eject, take their data with them — and become our customer.

Palantir just pivoted to sovereignty for governments. We're building it for everyone else. We're raising [$X] to get [N] apps live on the protocol by [date], starting with the teams in this room. If you're building on private data, you're our customer — 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

Finish the ask sentence — amount, use of funds, three milestones. It died mid-sentence on stage; it is a paragraph of work.
Pick infrastructure and re-cut the video's first 10 seconds so the demo demos the product being pitched.
Convert the cohort observation into signed design partners — get at least one of Conclave, Sanctum, or Dealproof sayable as "building on TinyCloud."
Name the customer ("we have a customer we've been building for") or stop mentioning them.
Delete the internal vocabulary and put "super base for sovereignty" at 0:10.
Audit every absolute trust claim in the video ("no silent logging," "no provider trains on it") against the current build before it plays again.
Filler drill + memorize the thesis sentence: live sections run 7.8/100 (um ×18, uh ×16) against the video's 0.