← All reviews

Case 22 · Graduation Day

Pramaana

"The safest database is when you have no database" is a thesis-grade line and the Bob story makes the privacy pitch felt — but the walkthrough then registers identities on-chain and shows a dashboard where "all your records are stored" without resolving the contradiction, offers no business model, buries its own founder, and ends asking another student team to collaborate instead of asking the room for anything.

Vaishnavi Rasika6:40 pitchreviewed 2026-08-17
4.5

01 Scorecard · 12 dimensions

Problem7The Bob arc stacks four real fears — tracking, passport handover, "a random Dan can open a credit card on Bob's name while Bob dies in credit debt," quantum breakage — and they land. But four problems in 2:40 is a buffet, not a statement.
Why now5"Then comes the quantum computer in the picture" plus the quantum-emergency / privacy-regulation tracks gesture at a why-now — but no timeline, no harvest-now-decrypt-later argument, no regulation named.
Insight / secret7"The safest database is when you have no database" + "A national ID number is not a secret" — two genuinely sharp reframes. The pitch's best asset.
Solution clarity5The one-liner exists ("the authentication system that stores nothing") but the middle is a term dump — thirty-six bit secret, Argon2id, blinded OPRF, PALC, fee key — faster than any non-cryptographer can absorb.
Product / demo5A real walkthrough (enrollment → on-chain → pseudonyms → dashboard → wallet check), but narrated features only — and "Don't be worried here" asserts privacy exactly where it should prove it.
Moat2"Completely open source… anybody can take it and use it," built on a public IC3 paper. Nothing spoken says what Pramaana keeps or why this instance wins.
Market4Population-scale India, DPI, fintech KYC, age checks — huge surface, zero math, and the sizing contradicts itself ("three million" vs "a billion").
Business model1Absent. Open source, government collaborations, and not one word on who pays, ever.
Competition4Two implicit contrasts — ID systems "centralized by design" and "more robust than what Apple ID uses right now" — but no landscape, and the Apple claim is asserted, not evidenced.
Traction4One named collaborator — Amravati Quantum Valley, a state-government initiative — is real signal. But "trying for a collaboration with the central government" is aspiration, and neither has stated scope, users, or a pilot.
Team2"This is about me. You can check me on LinkedIn… I won't go more on this." The founder slide, skipped by the founder.
The ask2The only ask goes to PyGrids — another team in the cohort. The room of investors is given nothing to do.

02 Klaff 5-lens audit · 2.5/5

Croc brainPARTIAL"Bob knows he's being watched" is a strong hook — but it arrives after a label-first open, and the Bob arc runs ~2:40 before the product appears.
FramePARTIAL"Be like Bob. And you are Bob. There are many Bobs in this room" commands the room — but holds it in the victim frame and never flips to the prize frame.
STRONG arcPARTIALStory → tech → product → use cases → collaborators → bio → thank-you. A narrative and a close, but no proof beat with numbers and no ask beat.
Hot cognitionPASSThe credit card opened in Bob's name, the CVS cashier, the Paris hotel passport — visceral and specific.
StatusFAILThe founder skips her own credibility slide, and the closing energy goes to courting PyGrids — supplicant posture in the final 30 seconds.

THE ONE REORDER — Open on the thesis line ("The safest database is no database"), compress Bob to 45 seconds, spend the reclaimed two minutes on proof and a real ask.

03 Delivery

Length
6:40
Pace
180 wpm
Fillers / 100 words
2

24 fillers ("uh" ×10, "like" ×6, "actually" ×4) — genuinely clean for a live stage. The delivery discipline is real; the problem is the material, not the mouth. 6:40 is twice a demo-day slot.

04 Beat-by-beat

TimeBeatWhat worksThe fix
0:00Label open"Prove that you're real without storing anything" is a strong core sentenceLead with it as the hook, not as ideology-after-the-label
0:15Bob is watchedEveryone recognizes the shoe adsThis beat alone runs ~45s. Cut to 15.
1:30Identity theft"A random Dan can open a credit card on Bob's name… while Bob dies in credit debt" — the pitch's most visceral lineKeep verbatim, move it earlier
2:00Quantum turn + thesis"The safest database is when you have no database"This is the open. It arrives at the 30% mark.
2:40Problem framing"Centralized by design," "biometrics become a liability," "a national ID number is not a secret" — crisp"Used by three million people" contradicts the later "a billion people" — fix the number before anyone Googles it
3:20Crypto pipelineArgon2id + blinded OPRF + rate limiting over a PALC commitment — real technical contentUnexplained, "thirty-six bit secret" sounds weak to a technical ear; say in one sentence why stretching makes it safe, or cut the bit-count
3:50Product flowEnrollment → hash → on-chain → master key → pseudonyms; Sybil resistance via biometric verification"Register it on the chain" + "all your records are stored in a dashboard" sit unresolved against "stores nothing" — say what is and isn't stored, in one sentence
4:40Biometric systemCompare-to-document instead of store-your-face is a clean differentiator"More robust than what Apple ID uses" needs evidence or deletion; "assumes the document is legit" is a trust-anchor admission a DD will press
5:20Use cases"Verify you without any surveillance," "fits into existing systems and not rips out the existing systems" — good adoption languageFour use cases, none chosen. Pick the wedge.
5:50CollaboratorsAmravati Quantum Valley, named, state-government initiative — best proof in the pitchAttach scope. "Trying for" the central government is not a beat — cut or convert to a date
6:10Bio + PyGrids + close"A billion people shouldn't have to be stored just so that they can be counted" — a great closing lineIt follows a skipped bio and a collaboration plea to another team. End on the line + the ask, nothing after.

05 The gaps investors block on

  1. No ask. Six minutes forty, and the room is never told what Pramaana wants — the only ask goes to PyGrids, a fellow team.
  2. No business model. "Completely open source… anybody can take it" with no revenue word spoken reads as project, not company.
  3. The numbers contradict: Aadhaar "used by three million people right now" vs "a billion people shouldn't have to be stored." One skeptical listener catches this live.
  4. The stores-nothing / on-chain / dashboard contradiction — the pitch's central claim, undermined by its own product tour.
  5. "More robust than what Apple ID uses right now" is unsourced against one of the most attacked biometric systems on earth.
  6. "Trying for a collaboration with the central government" is a hope narrated as traction.
The one strategic decision

Own the government DPI wedge. The pitch names four use cases and two postures (open protocol vs product). Lean entirely on the one real asset: the named state-government collaborator (Amravati Quantum Valley). Government DPI is the one buyer for whom "no database" is a procurement feature, not a curiosity — and the pitch already owns the Aadhaar critique, the billion-people framing, and the fits-into-existing-systems adoption line that wedge needs.

“Pramaana is proof you're a real, unique person that stores nothing — a post-quantum personhood layer a government can adopt without ever building a biometric database, live first in [Indian state — confirm Amravati collaboration scope].”

06 The rewritten pitch

The safest database is the one that doesn't exist. Every ID system you're enrolled in — your national ID, your bank's KYC, your face on some server — is a breach waiting to be monetized, and a random stranger can open a credit card in your name while you die in the debt. And the cryptography guarding all of it breaks the day a good-enough quantum computer shows up.

Pramaana proves you're a real, unique person while storing nothing. Your ID details are used once, as entropy — stretched through Argon2id and a blinded OPRF into a key only you can reproduce — and what goes on-chain is a one-way, quantum-safe commitment. Not your name, not your face, not your number. To stop duplicates, we match your live face against the photo on your own document, and then throw both away. [confirm: exact data retention — this sentence must be literally true]

This isn't a paper. [confirm: working demo status] And it isn't unproven: we're building it with the Amravati Quantum Valley, a [confirm: state] state-government initiative — the first government asking how to count a billion people without storing them.

I'm Vaishnavi Rasika. I'm asking for [the ask: pilot scope with one state department / $X / N design partners] to take Pramaana from one state initiative to the identity check that never has to be breached — because there's nothing to steal.

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

  1. Pick the wedge — government DPI via the Amravati collaboration — and cut the use-case buffet from the pitch.
  2. Write the ask — pilot scope, amount, or design partners; its absence is the difference between a project and a company.
  3. Resolve the stores-nothing paradox in one spoken sentence: what's on-chain, what's discarded, what the dashboard actually holds.
  4. Fix the Aadhaar number — "three million" vs "a billion" cannot both be said; verify and standardize.
  5. Source or delete "more robust than Apple ID."
  6. Cut the pitch to 3:00 or less — Bob to 45 seconds, crypto to two named primitives, bio restored to one proud sentence instead of "I won't go more on this."