@alienorg

new mini app arena lets you access frontier AI models for free in exchange for ranking the responses you like the most available in the alien app https://t.co/X84uBDLA5Z

Every platform you join asks you to prove you're a real person from scratch. CAPTCHA, phone number, government ID, selfie. You've proven your humanity hundreds of times. None of those proofs carried over. This isn't a design oversight. It's the default architecture of the internet: every platform owns its own verification silo. The proof you gave Twitter when you signed up means nothing to your bank, your marketplace, or the forum you joined last week. Each one starts from zero. The cost is friction that accumulates invisibly. A few minutes here, a phone number there, an ID scan somewhere else. Multiply it across every service you use and the total is significant. But the deeper cost is that verification systems built this way have no memory - which means attackers can exploit each platform's ignorance of every other platform's records. Here's what Alien ID changes in practice: You only verify once. Biometric verification runs inside a Trusted Execution Environment on your device. Your face is processed and discarded. A cryptographic proof is generated: verified unique human. That proof is what travels - not your data. The proof works anywhere the platform accepts it. Any app or service that integrates Alien ID can check your verification status without seeing anything about you. The check answers one question: is this a verified unique human? The answer comes back yes or no. Nothing else transfers. Different contexts, different identities, same root proof. For a forum, you create a session that proves you're human without linking to any other account you have. For a financial service, a different session. Each one cryptographically derived from your root Alien ID, none of them linkable to each other. Five places this removes friction today: Community access — skip the phone number verification that bots route around anyway Platform onboarding — one proof accepted instead of repeated ID uploads Age-gated content — prove the relevant fact without submitting a document Marketplace trust — verified human status visible to counterparties before a transaction AI service access — human verification as the entry credential, not behavioral monitoring One verification. Any context. The proof travels with you.

What's the smartest thing your AI agent did this week? And the dumbest? Open thread👇

25 minutes until AMA with Kirill in Discord. We’ll cover what’s being built and where things stand. Join us! https://t.co/aBgGHyyYZN

The call came from his CFO's face and voice. The number matched. The background was the right office. He transferred $25 million before anyone realized the CFO had never made the call. This isn't a hypothetical. In early 2024, an employee at a multinational firm in Hong Kong joined a video call with who appeared to be the company's CFO and several other senior colleagues. The voices were right. The faces were right. The setting was convincing enough that nobody in the call questioned it. The CFO asked for a series of transfers. The employee authorized them. $25 million moved before anyone on the real team realized what had happened. Every trust signal that humans have learned to rely on was present and fabricated: visual identity, voice, context, authority, urgency. The deepfake didn't need to be perfect. It needed to be convincing enough that skepticism felt inappropriate. It was. The defensive instinct after cases like this is to add verification steps: call back on a different number, use a code word, require a second approver. All of these work until the attacker has enough context to route around them - and with AI, gathering that context is the cheap part. The structural problem is that identity verification in this scenario happened at the moment of the call, using signals that can be fabricated. By the time the employee was deciding whether to trust the CFO, the attacker had already won the information asymmetry. The fix has to happen before the interaction starts, not during it. A cryptographic proof that the CFO is the CFO - one that exists independently of what their face looks like or how their voice sounds, and that no deepfake can replicate - changes the question the employee is answering. Instead of "does this look like my CFO," they're answering "has this session been cryptographically authorized by my CFO's verified identity." That's what Alien ID creates. Not a better way to scrutinize faces. A proof layer that exists underneath them.