Why we are building for agents.
Software is increasingly used by agents acting on a person’s behalf, and almost nothing was built for that. This page says what we think is changing, what breaks, and where we are starting.
Software has a new kind of user.
For decades a product had one kind of user: a person at a screen. Now a growing share of use is an agent doing a task for a person: reading a page, filling a form, calling an API, moving money.
Products built for people fail agents in predictable ways.
Data sits in screenshots. Actions hide behind scripts. Flows meant to slow a person down stop automation outright. And when an agent reports a result, there is often no way to check it except to trust the agent.
Structured, predictable, checkable.
Our view is that agent-ready software needs three properties: structured data an agent can read, predictable flows it can rely on, and outputs a stranger can verify without taking the operator’s word. An interface built that way tends to be better for the person using it too.
Two products, two kinds of trust.
AnalystOS tackles evidence: a report in which every figure is proven against its source, sealed so it can be re-verified offline. DepositX is about money movement for agents, and is in development.
What “alpha” means here.
AnalystOS is an early release. It works end to end and its limits are published next to its strengths, but it is not yet for confidential documents. Accounts, access controls and security certification are planned, not built. We would rather show you where it is weak than hide it.