The trust stack of every era of computing
Sixty years of computer networks; six trust layers. Each emerged when its predecessor's assumptions broke. Each became boring within a decade. The agentic era is the seventh moment, and the layer being built now will be boring before.
ARPANET: trust by enrollment
The earliest hosts trusted each other because the network operator personally enrolled them. Trust was a clipboard. It worked at four nodes and broke at forty.
DNS: trust by name
When hosts outgrew clipboards, DNS gave the world a hierarchical naming protocol. The nameserver said "this name belongs to this address." It was advisory, not authenticated; for fifteen years, the world cheerfully ran on advisory trust.
SSL/TLS: trust by certificate
When commerce showed up, advisory trust failed. SSL added cryptographic certificates that bound a domain to a public key. The certificate authorities became the new mediators of trust. Boring within ten years; foundational still.
DKIM, DMARC, SPF: trust by signed delivery
Email's spam crisis drove three layered protocols that authenticated the path of a message, not just the contents. DKIM signs; SPF binds path to sender; DMARC composes both. Within a few years these were as invisible as DNS.
OAuth: trust by scoped delegation
Web 2.0 needed a way for users to let apps act on their behalf without handing over passwords. OAuth invented scoped, time-bound delegation between services. Two decades later it is plumbing for every B2B integration.
WebAuthn / FIDO2: trust by hardware
When credential theft outscaled password rotation, FIDO bound authentication to hardware. The secure element became the root of trust. Today's passkey flows ride on that decision; the day-to-day login is calmer for it.
HATI: trust by attested human
The agentic era's specific failure mode: actions running across systems with no record of which human stands behind them. The seventh layer answers it. Hardware-bound human DIDs sign delegations to agents; relying parties verify the chain; audit trails name the human. This is the layer Manav is building. In a few years it will look as boring as DNS.
What the pattern teaches
Three lessons. The next layer always comes from the previous layer's failure mode, not from a marketing ambition. The protocol that wins is open and the implementation that wins is commercial. Whichever layer is being built now is unrecognizable as infrastructure for at least five years; that is the test, not whether it ships in three months.
What we cite our predecessors for
DNS for hierarchy. TLS for cryptographic root authorities. OAuth for scoped delegation. FIDO for hardware binding. Each of those primitives shows up explicitly in our protocol; nothing about Manav is invented in a vacuum. The seventh layer is not new — it is an arrangement of old primitives that nobody had reason to assemble until agents made the assembly necessary.
Common objections
Two objections worth answering. Stated values do not survive growth pressure — true historically, which is why we put structural mechanisms (open-source, governance, protocol-enforced custody) behind the words rather than just the words. This sounds like marketing — the test will be the audit hashes, the protocol design, and the operating agreements, not the prose.
Frequently asked questions
What does this commitment cost us if we honor it? Real money in the years where the temptation would have been highest. We are pricing it in upfront because the commitment is structural, not aspirational.
Where do we publish this commitment? Here, on the protocol governance page, and in the operating agreements with our investors. Anyone can audit whether the commitment is being kept by reading the audit hashes we publish quarterly.
What if leadership changes? The commitment is structural enough that a new leadership cannot quietly reverse it. The protocol mechanics make the breach detectable; the legal commitments add a second layer; the cultural commitments add a third.
Where to start
For the wider posture, read dns of human trust and seven layers of trust. The values, the protocol, and the operating model only fit together when read in that order.
Why the stack always rewrites itself
Every twenty years, the trust substrate gets rewritten. The notarized signature gave way to the typed name. The typed name gave way to the password. The password gave way to the OAuth token. The OAuth token is now giving way to the delegated agent action. Each rewrite happened because the prior layer broke under a pressure it was not built for: scale, then remoteness, then automation, now agency. The next pressure is already visible — the volume of agent actions is rising at a rate that exhausts every prior trust primitive within the decade. We are not building Manav because the OAuth-era infrastructure is bad; we are building it because the substrate has run out of pressure-headroom. The history of identity is the history of substrates that worked beautifully until they suddenly did not. The lesson is to build the next layer before the prior one breaks publicly.
You do not invent a trust layer. You arrange one. The arrangement that survives becomes the next decade's plumbing.