The independent control and evidence layer for AI-generated software change.
AI now generates more change than teams can safely review. NEO is the independent layer that bounds what an agent may touch, proves what actually happened, and keeps a human in authority over what ships — making AI-generated change defensible and reversible.
NEO gives AI-generated change a controlled surface area: bounded authority, independent evidence, and a human decision at the edge.
Every change is bounded before it starts: approved intent, allowed paths, and protected files and secrets an agent may never touch.
Connects to: Protected execution, Rollback & human authority
Every change is bounded before it starts: approved intent, allowed paths, and protected files and secrets an agent may never touch.
Connects to: Protected execution, Rollback & human authority
Open ->Work runs inside enforced boundaries — write mediation and snapshots before any destructive step, nothing beyond the approved surface.
Connects to: Scope & path control, Evidence custody
Work runs inside enforced boundaries — write mediation and snapshots before any destructive step, nothing beyond the approved surface.
Connects to: Scope & path control, Evidence custody
Open ->Acceptance tests are authored and frozen before implementation, so the bar can't move to fit the result; judges are checked for vacuity.
Connects to: Independent audit, Evidence custody
Acceptance tests are authored and frozen before implementation, so the bar can't move to fit the result; judges are checked for vacuity.
Connects to: Independent audit, Evidence custody
Open ->A cold, context-isolated channel re-runs the evidence — not the builder's narrative — and recommends keep, remediate, or reject.
Connects to: Frozen acceptance criteria, Evidence custody, Rollback & human authority
A cold, context-isolated channel re-runs the evidence — not the builder's narrative — and recommends keep, remediate, or reject.
Connects to: Frozen acceptance criteria, Evidence custody, Rollback & human authority
Open ->Hashes, diffs, test output, residue checks, and decisions form a re-auditable record held independently of the agent that made the change.
Connects to: Independent audit, Protected execution, Rollback & human authority
Hashes, diffs, test output, residue checks, and decisions form a re-auditable record held independently of the agent that made the change.
Connects to: Independent audit, Protected execution, Rollback & human authority
Open ->Every change is reversible, and a human holds the final keep/iterate/reject decision. Authority is never delegated to agent confidence.
Connects to: Scope & path control, Independent audit, Evidence custody
Every change is reversible, and a human holds the final keep/iterate/reject decision. Authority is never delegated to agent confidence.
Connects to: Scope & path control, Independent audit, Evidence custody
Open ->Intent, authority, paths, protected files, budget, and risk are approved before execution begins.
The change runs inside the approved surface, with custody boundaries and snapshots preserved as it moves.
A separate review path produces findings, rerun output, diffs, and a re-auditable acceptance packet.
A human makes the final call. The decision can keep, remediate, reject, or roll back the change.
A convincing summary is not a result. What ships is judged by reproducible artifacts — tests, diffs, residue — not by how sure an agent sounds.
The party that writes a change never grades it. Building, judging, and auditing are different seats, so no one can move the bar to fit their own work.
Every change is scoped before it starts and reversible after it lands. Authority to accept stays with a human; nothing consequential ships on autopilot.
We label what is proven and what is not. NEO is proven in founder-operated development today; commercial proof is earned through paid evidence, not claimed in advance.
The party that writes a change never grades it. Builder, judge, and auditor are different seats.
The acceptance record is held apart from the agent that made the change, so it can be re-audited later.
A person makes the final keep/iterate/reject call; agent confidence is never evidence.
Scope, controls, and evidence travel across repositories, agents, and model providers, rather than living inside one platform.
These are the properties NEO is built to provide; whether teams will pay for independence over bundled controls is what our first paid evaluations are testing.
The first step is a bounded evaluation: an independent audit of one high-consequence change in a real repository, producing a re-auditable acceptance packet and a keep, remediate, reject, or rollback decision you can defend.
Early engagements run as hands-on, paid design partnerships - one supported stack, direct access to the people building NEO, and honest limitations stated up front.