Alygn (@aialygn) header
Alygn (@aialygn) profile picture

Building the political infrastructure for life beyond Earth. Research, pilots, and governance frameworks for future societies.

Earth → Mars
Joined August 2023
Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)Photo by Alygn (@aialygn)
Who can patent EAGF? Nobody. Who can implement it? Anybody. Yes, competitors too. EAGF (Enterprise Agent Governance Framework) is an open specification for governing AI agents. alygn.us
1
1
11
Can your AI agent do more than it needs to? The first axis of our trust audit is Authority: what an AI agent can do, and whether it is limited to that (least privilege). We review the evidence read-only: permissions, API scopes and amount limits.
5
Three questions to ask before an AI agent gets access to anything: Who approved it? What exactly may it touch? How is it switched off, and by whom? If the answers live only in chat messages, they are not records. Happy to talk: DM or [email protected]
7
Antes de contratar una auditoría de agentes de IA, conviene saber también lo que queda fuera. alygn.us
5
Who can patent EAGF? Nobody. Who can implement it? Anybody. Yes, competitors too. EAGF (Enterprise Agent Governance Framework) is an open specification for governing AI agents. alygn.us
10
Do you sell cybersecurity or digital identity to banks, insurers or fintechs, with a network of integrators? Alygn is looking for a remote channel cofounder for Agent Trust, governance for AI agents. Founding role, no salary. Not a securities offer. [email protected]
13
A quick check for any company running AI agents: can you show, for one agent, who approved it, what it may touch and how it is switched off? If the answer takes a week to assemble, that's the gap. Happy to show our prototype: DM or [email protected]
5
Delegar no es ceder sin límite. Cuando un agente de IA le pasa una tarea a otro, EAGF pide tres cosas en cada traspaso: ✓ Viaja una parte acotada de la autoridad, con su propósito y sus restricciones. alygn.us
3
Se habla mucho de lo que la IA puede quitar y poco de lo que podría abrir. «Hacia la Era de la Abundancia» es una miniserie costarricense de capítulos cortos sobre futuros positivos con IA. Llega a YouTube en 2027: youtube.com/@EraAbundanciaCR
13
Your AI agent doesn't work alone: it connects to external tools. Each connection deserves four questions. Where does it come from? What permissions does it ask for? Did it change without notice? Can a person cut it off instantly? alygn.us
1
19
A quick test for any team running AI agents: pick one agent and ask three things. Who approved it? What can it touch? How do we stop it? If the answers live in three different places, that's the gap. Agent Trust Inventory is our prototype. Happy to show it: DM or [email protected]
16
Gobernar agentes de IA no es encerrarlos. Una órbita no es una jaula: autonomía para sus agentes de IA dentro de límites declarados, alrededor de las personas. Y comprobable. Disponible hoy: Agent Trust Inventory. Más en alygn.us
6
Costa Rica ran on 98.6% renewable electricity in 2025. A Former Costa Rican Deputy (2024–2026) has published an open proposal for pilot sites there for advanced manufacturing and robotics. For anyone curious: medium.com/@cynthiamcordobas…
1
66
Se habla mucho de lo que la IA puede quitar y poco de lo que podría abrir. «Hacia la Era de la Abundancia» es una miniserie costarricense de capítulos cortos sobre futuros positivos con IA. En YouTube: youtube.com/@EraAbundanciaCR
7
Your AI agent changed. Its history didn't. To audit a decision, don't ask the agent today: read what was recorded then (traces, model or prompt versions). EAGF HV-26. alygn.us
6
Question for risk and audit teams: if one of your AI agents makes a mistake tomorrow, who can stop it, who approved its limits and where is that written? Agent Trust Inventory is our prototype for answering that. Happy to show it: DM or [email protected]
1
2
60
Decisión ≠ Recomendación. Que un agente de IA recomiende no significa que alguien haya decidido. En Agent Trust Inventory cada agente de IA queda registrado con su nivel de autonomía y un responsable humano con nombre. alygn.us
7
As @stretchcloud put it, handing a coding agent a whole password vault isn't scoping. A simple rule: each agent gets only the keys for today's task, an expiry, and a named person who approved it. Agent Trust Inventory keeps that on record. alygn.us
32
A permission is not a receipt. An AI agent having permission doesn’t prove it did it: Authority and Evidence are checked separately. alygn.us
18
Reading @chetanp on LangChain, I keep thinking about the next layer: once companies run agents at scale, risk teams need to know who owns each one, what it may do and how to stop it. That's what we build at Alygn (prototype). Happy to show it: DM or [email protected]
18