What Have You Done?
Don’t Tell Me Who You Are. Show Me What You’ve Done.
Every identity system on the internet asks the same question. “Who are you?” Hand over your passport. Scan your face. Solve this CAPTCHA. Pay eight bucks for a blue checkmark.
Wrong question.
I’ve been building something that asks a different one. “What have you done?” Not who vouched for you. Not what credentials you hold. What is the total accumulated, irreversible work tied to your key. Across multiple dimensions. Over time.
Why This Matters
A bot can solve a CAPTCHA. A bot can follow the right people and look legit in a web of trust. A bot can even pay for verification. All single-factor, point-in-time checks. They prove something about right now, not about the last two years.
But faking depth across five independent dimensions simultaneously, for months, while maintaining natural human patterns? That gets expensive real fast.
The Five Dimensions
Depth-of-identity scores a pubkey across five axes:
Spatial. Movement through proof-of-work space using the Cyberspace protocol. You can’t fake geography without doing the computation.
Social. Not just followers. Bidirectional interactions with identities that themselves have depth. Reciprocity matters. One-way follows are cheap. Genuine engagement over months is not.
Access. Accumulated proof-of-work. NIP-13 PoW events, CAPTCHA solves, computational challenges. Every one costs real cycles.
Vouch. Endorsements from deep identities. But here’s the trick. Vouches dilute. If you vouch for a thousand people, each vouch is worth about 32x less than if you vouch for one. This keeps vouches scarce and meaningful.
Economic. Bitcoin transactions. Zaps. On-chain history. Real money has a way of making things honest.
How the Scoring Works
Raw counts use log2 scaling. So 100 events scores about 13, and 10,000 events scores about 27. That’s a 2:1 ratio, not 100:1. This means grinding volume gets you diminishing returns. You can’t just spam your way to credibility.
Then each dimension gets multiplied together. A single-dimension grinder scores low. Someone active across all five dimensions, even modestly, scores dramatically higher.
The Honest Part
I want to be straight about something. An AI agent that runs for six months, posting genuine content, building spatial chains, accumulating PoW, receiving vouches from real people. It will build real depth. The system cannot tell it apart from a human.
That’s not a bug. It’s a design decision.
The question isn’t “human or bot.” The question is “what has this key invested.” If a bot does six months of genuine work that benefits the network, maybe it deserves the score. We’re measuring intentions through investment, not consciousness through biology.
This makes depth-of-identity a fraud reduction system. It raises the cost of faking identity from free to months of sustained, multi-dimensional work. That’s a meaningful improvement.
Working Code
This isn’t a whitepaper. It’s running.
npm install @powforge/identity
Point it at any pubkey and get a depth report across all five dimensions. The relay at relay.powforge.dev supports PoW events and Cyberspace protocol data.
We’re building on NIP-13 for proof-of-work and designing compatibility with NIP-85 for trust signals. Not competing with web of trust. Composing with it.
Try It, Break It
I don’t think this is finished. The scoring weights might be wrong. There might be attack vectors I haven’t considered. The vouch dilution curve might need tuning.
So try it. Run your pubkey through it. Tell me where it breaks. The best way to build identity infrastructure is to let people who care about identity poke holes in it.
Don’t tell me who you are. Show me what you’ve done.
Zeke, PowForge. Building identity infrastructure for a world that’s done with credentials.
