human.tech's banner
human.tech's profile picture

human.tech

@humntech14,041 subscribers

keeping humans in the loop of autonomous tech via proof of humanity @humnpassport & secure secrets management @waapxyz built by Holonym

Videos

humntech's profile picture

hello world, meet

human.tech

36,553 Aufrufe • vor 1 Jahr

humntech's profile picture

At HSC Asset Management in Hong Kong, hosted by Mpost Media Group, our co-founder S El Damaty gave a keynote titled “One Account, Every Chain.” On the surface, it’s about interoperability. Underneath, it’s about something even more fundamental: Would you trust an AI agent with your private key? --- Not theoretically. Practically. Would you give full signing authority to software running locally or in the cloud, knowing it can move assets, execute trades, deploy capital, or spin up new transactions on your behalf? IT'S ALREADY HAPPENING As you're reading this post, people are already doing it. Agents are being embedded into wallets, trading systems, and productivity tools. It's the new narrative: software is automating everything. Finance is next. But crypto, since Bitcoin, was built around self-custody. “Not your keys, not your coins” wasn’t marketing. It was architecture. If you hand an agent your full private key, you have reintroduced a custodian. It may look like automation. But structurally, it’s concentration of authority. Right now, most implementations fall into two models. Either the agent gets full control, or it runs inside trusted infrastructure like AWS Nitro. Both approaches shift trust somewhere else. Either into the agent itself, or into centralized compute providers sitting underneath the system. If the future economy runs on agents, that becomes a very large assumption. THE ALTERNATIVE The alternative presented in the talk reframes delegation entirely. Instead of giving authority, you give capability. With Two-Party Computation (2PC), a private key can be split into shares. The agent never holds the full key. It receives scoped permissions: time-bound, policy-defined, function-specific. It can execute within constraints. It cannot take over. That difference is subtle, but it's foundational. Delegation does not have to mean surrendering custody. The next step is anchoring those permissions onchain. Key derivation and policies can be written into protocol-level logic through threshold networks like Human Network, previously Mishti Network, rather than relying purely on offchain safeguards. Using Ika「🦑」 within the Sui ecosystem, keys can be generated across different elliptic curves, enabling a single account to operate across Ethereum, Solana, and other chains without fragmenting custody. We're doing it with WaaP, Wallet as a Protocol, the first decentralized seedless wallet on Sui, also operating on EVM chains. “One Account, Every Chain” is not only about asset interoperability. It’s about unified control across chains, without collapsing back into centralized trust. This is not an anti-AI argument. Agents will automate, trade, coordinate. The question is whether humans remain sovereign in the process. If we're serious about an agent economy, the architecture has to reflect that. Agents should execute. Humans should retain authority. Interoperability is powerful. Automation is powerful. But if we don’t redesign how keys and permissions work, we rebuild the same fragilities crypto was meant to escape. The question lingers long after the talk ends: Would you trust an AI agent with your private key?

human.tech

13,552 Aufrufe • vor 5 Monaten

Keine weiteren Inhalte verfügbar