The organisation's only active project
Agents today are applications. This is the argument that they should be an operating system — and that the difference is not branding, but three abstractions nobody has built yet.
An OS earns the name when it owns how work survives interruption, how state is addressed, and how a person perceives and steers the whole machine. QM already solves the hard part underneath: a real multi-tenant harness with scoped identity, permissions, sandboxes and audit. ai-os is the layer above it.
Four questions today's agents cannot answer
Four pillars
What if the unit of work outlived the conversation?
A flow is a declared, persisted, resumable object with a goal, a shape, a state and a lineage. It records where it forked from at its first commit, so two attempts can be compared. The turn becomes an implementation detail.
What if the interface were the state, not the transcript?
A spatial, live canvas composed by the system from the flow's state — and re-composed as it changes. The user overrides the arrangement; the override wins from then on.
What if memory had an address space?
Four levels — system, user, project, flow — with different lifetimes. Flow memory is expected to die. Promotion between levels is explicit, recorded and reversible.
What if we did not rebuild the part that works?
QM, vendored as a subtree and pulled weekly. Identity, scopes, sandboxes, policy, audit and six model harnesses. We did not write it and we are not rewriting it.
Where this actually is
Design stage. ai-base is vendored and runs. The other
three pillars are specified in doc/ and not yet
implemented. Nothing on this page describes running software unless it
says so.
Publishing a design before the code is deliberate. The predecessor of this project shipped eighteen thousand lines describing five subsystems and three test functions — an architecture written down and never pinned to anything that could contradict it. So each pillar here ships with the measurement that would show it is not worth building, written before the code: that a flow completes work a plain session loses, that four memory levels retrieve better than one flat file, that a canvas finds state faster than a transcript.
The memory claim has already come back flat once, on a closely related experiment — 80% either way. That result is why the burden of proof sits on the new axis and not on the baseline.
Twenty-six frozen experiments, 2025 – 2026.
Agent memory, self-modification, interpretability and constrained decoding — each labelled by how much evidence stood behind it, including the ones where the evidence went against us. Kept because they are still true, not because they are maintained.