Utility backed infrastructure for the 98% who think blockchain is a scam
A public ledger has no notion of a private object shared with a chosen few. The options are publish it, or encrypt it and hand somebody the entire key. There is nothing in between, so nobody puts anything that matters on chain.
Self custody has no recovery path. Lose the device and the identity, the records and the credentials are gone permanently. Ordinary users will not accept that risk, and no institution can underwrite it.
The standard workaround for privacy. It is all or nothing: the moment you share, the recipient can pass it on forever and you cannot revoke.
The standard workaround for recovery. It moves the failure from the software to a piece of paper in a drawer, and the failure is still total.
Recovery bolted on above the protocol, so it stays optional. Optional recovery is no recovery, and no institution will underwrite an optional guarantee.
What almost everyone actually does. It restores privacy controls and password resets by handing one company the ability to read, alter or lose everything.
What if a private file could be opened by exactly three named people and nobody else; what if losing your phone did not mean losing your name, your records and every credential you ever earned; and what if neither of those required the user to know a blockchain was involved at all.
Both move to the protocol level, so consensus prices a transaction by the future AI utility it creates rather than by gas alone. The chain pays for useful work instead of block space.
Borrowed from C++. An owner grants named parties scoped read access to a private encrypted object without ever surrendering the key.
SOLVES PRIVACYCredentials and records are issued on chain by the institution itself, turning issuance into a graph of verified relationships rather than a pile of documents.
Multi party computation over that graph reconstructs a lost private key from the institutions and people who already vouched for you.
SOLVES USABILITYBets a compute marketplace produces correctness. With no slashing, a wrong answer costs the provider nothing, so correctness stays a preference. Their wall is economic.
Bet that storage and access control are separable. Without protocol level permissioning, sharing still means handing over the whole key. Their wall is architectural.
Bet that verifiable credentials could be made usable above the chain, then recentralized the moment usability mattered. Their wall is custodial.
Treats privacy as an opt in module rather than a consensus rule, so it protects only the applications that bother to use it. Their wall is coverage.
MOVEit, National Public Data and PowerSchool all broke the same way, exposing 1.7 billion records in 2024 alone. The workaround everyone actually uses, handing one company everything, is the one that keeps failing.
Apple, Google and Microsoft put recoverable keys in front of a billion people. After that, an unrecoverable seed phrase reads as indefensible rather than merely inconvenient.
Apple Silicon and consumer GPUs reached the point where distributed inference costs less than renting cloud capacity, which makes a compute subprotocol economically real instead of theoretical.
a16z Crypto closed a $2.2B fund aimed at turning crypto infrastructure into everyday products, and infrastructure is now the dominant pre-seed deal type.

RUMUS — Rust-native deep learning framework, written solo. UCLA Math/CS. Alpha Speedrun: 100% coding assessment.

NASA lunar mission telemetry published in SPIE. DARPA-backed AML subgraph matching engine in C++. Architected SNIP from zero. Shipped SUMaillet to Chrome Web Store.

Univ. of Delaware Applied Math PhD candidate. Proved OmniNode model-sharding architecture. Shipped Graduate Journey — live product for underserved first-gen and international students.
Prof. Andrea Bertozzi (UCLA) · Prof. Paolo Climaco (UCLA) — Co-directed Leonard & Michael's graph-matching research (signed MATH 199). Foundational to SUM Chain consensus layer.
What we're raising and what we need beyond capital:
Validators and Archive Nodes to secure SUM Chain and serve SNIP at scale; core engineering to harden every layer.
School and institution deployments for SUMail — travel, installation, on-site onboarding.
Marketing and sales experts to form GTM, build pipeline, and reach institutions.
Patent filings across SUM Chain, SNIP, OmniNode, SUMail — plus IP and corporate counsel.
Multi-month operating stability through the next milestone.