AiOverload, governed execution for AI agent fleets

Console: attention-first dashboard

Thu, 08 Oct 2026 00:00:00 +0000

The console’s first IA/UX slice landed: a dashboard that leads with what needs a human, run detail organized around steps (where the decisions actually are), and cross-links that make the log navigable from any screen. The approvals queue your words land on is the same one the homepage mock shows — now visibly closer to it.

Stamp: 3fc7f47 — engine main.

Empty acceptance refuses to run

Thu, 08 Oct 2026 00:00:00 +0000

A forge-ingested row with an empty acceptance: could previously fold to done on any terminal run — a quiet Goodhart hole. Now the server refuses start/resume on such rows with a named acceptance_required error, and the operator gets an explicit accept path. The hole the status page lists as partial got smaller today.

Stamp: 022bde9 — engine main.

Forge-aware task actions

Thu, 08 Oct 2026 00:00:00 +0000

The console’s task rows advertised start/pause/resume/cancel unconditionally — including on forge-bound rows the server would refuse. Actions are now forge-bind aware: what you see is what the server accepts. Small change, real honesty gain: a button that always refuses is a lie with a border radius.

Stamp: b35a48f — engine main (this is also the commit the status page is currently stamped with).

About

Mon, 01 Jan 0001 00:00:00 +0000

One engineer, on purpose.

I’m Johannes Girard — a senior platform engineer, French, working fully remote from Normandy. (LinkedIn) AiOverload is my project: built in my free time, developed in the open from the public release on, with everything this site claims pinned by tests rather than by confidence. This page is the honest version of who I am and how the thing was actually made.

The path

Hotels, AWS certs, and a long way around.

Five-star hotels → engineering school at 25

Before tech, I worked in five-star hotels. At 25 I went back to engineering school — and kept working hotels on weekends and holidays to pay for the studies. Hospitality teaches you service under pressure; engineering gave me the system. Both stay in everything I build.

How it works

Mon, 01 Jan 0001 00:00:00 +0000

The engine, under the hood.

Four binaries on one bus, contracts before code, and a single rule that never bends: the event log is the runtime, and nothing executes outside it.

The pipeline

Four binaries, one bus, two governed doors.

The problem: agents today run as loose processes — their access, their steps, and their failures live in different places, and reconstructing “what happened?” afterwards is archaeology. The answer: one pipeline, one bus, one door — every transition an event on the log.

Legal & privacy

Mon, 01 Jan 0001 00:00:00 +0000

Legal, hosting, privacy — the unglamorous page.

Everything a visitor (or a regulator) needs to know about who publishes this site, where it lives, and what it does with data. Kept short and honest like the rest.

Publisher

Who publishes this site.

Éditeur / Directeur de la publication: Johannes Girard — senior platform engineer, France. Contact: johannes.girard@gmail.com · LinkedIn.

This is a personal, non-commercial pre-launch site for an open-source project. AiOverload’s source code will be published under AGPL-3.0; the prose and visual identity on this site are © 2026 Johannes Girard. When the announced commercial offer becomes real, this block gets the professional identification the law then requires — and this page will say so first.

License

Mon, 01 Jan 0001 00:00:00 +0000

AGPL-3.0, in plain words.

The short version: the engine is open source, not “source available”. The distinction matters, and this page exists so nobody has to infer it.

The license

What AGPL-3.0 gives you.

The core — engine, reflex gate, manifests, workflows, A2A, the sidecar contract, the CLI, the console, observability — is licensed AGPL-3.0, an OSI-approved open-source license. In practice:

Self-host, forever, for anything — internal tools, client work, production fleets. No seat counts, no usage telemetry, no feature gates.
Modify it — the source is yours to read, fork, and adapt.
Share it — redistribution comes with source obligations, the standard copyleft deal.
The network clause — if you run a modified version as a service others use, AGPL-3.0 requires you to offer those users the modified source. That clause is the point: the engine cannot be quietly closed into a competing hosted product.

The commercial boundary

Philosophy

Mon, 01 Jan 0001 00:00:00 +0000

How we think.

The vision is the why; this is the how — the doctrines the engine refuses to break, stated as opinions with reasons attached. Each one exists because the opposite cost something real.

Doctrine 01

Nothing is claimed before it exists.

Marketing writes checks that engineering cashes; when they drift, trust is the casualty. So the site carries a status page stamped with the engine commit it was generated from, the threat model grades its own containment layers and names its gaps, and a claim without a pinned test is a bug — whoever wrote it. Honesty here is not a value statement; it is a build constraint, the same way “unknown keys are loud errors” is.

Platform

Mon, 01 Jan 0001 00:00:00 +0000

The platform around the engine.

Where agents, tasks, knowledge, and your repositories become one governed system — all of it declared as contracts, all of it auditable.

Agents are configuration

The manifest: persona, boundaries, budget — one reviewed file.

An agent is a typed, schema-pinned entity in registry/agents/. The five-piece contract set — Go types + loader, YAML, JSON schema, contract tests, docs — means a manifest that doesn’t validate doesn’t exist:

Roadmap

Mon, 01 Jan 0001 00:00:00 +0000

Where it’s going — and what stays free.

A public pre-launch roadmap, kept honest the same way as the code: nothing is claimed before it exists, and nothing gets a date it can’t keep.

Roadmap

Available, in progress, planned — in that order.

The same vocabulary as the status page: Available exists at the stamped commit, In progress is being built on main today, Planned is designed or contracted — named, never dated.

Security

Mon, 01 Jan 0001 00:00:00 +0000

The threat model, without the sales filter.

This page summarizes the engine’s own docs/threat-model.md — the full document publishes with the source at release. The rule is the site’s rule: a security claim you have to dig for is marketing.

Containment, graded

What actually contains an agent today.

The boundary: per-run Docker containers

The engine’s real containment layer: on the docker runtime, a run gets an ephemeral container — its filesystem staged copy-in/copy-out, its network a private lane whose only peer is the gateway. When the threat model says “boundary”, this is the layer it means. Stated plainly: the shipped fleet opts into the cli tier today, so this boundary is wired on main but not yet the shipped configuration; container hardening (resource limits, cap-drop) is in progress.

Status

Mon, 01 Jan 0001 00:00:00 +0000

The honesty page, machine-stamped.

Every number on this site drifts the moment it is typed by hand. This page is different: its facts are generated from the engine repository at deploy time and stamped with the exact commit they came from. When this page and another page disagree, this page wins — and the other page is a bug.

Stamp

What this page was generated from.

Engine repository	the AiOverload engine repository — private until the public release; the commit below is verifiable against it then
Commit	`b35a48f`
Commit date	2026-10-08
Measured	95 Go packages with source (test files excluded from the count)

The generator runs at deploy time against the engine repository; until the engine is public, the data is regenerated from a maintainer-side checkout and committed — never typed.

Vision

Mon, 01 Jan 0001 00:00:00 +0000

Agents are about to hold the keys. Who governs them?

Every framework answers “how do I build an agent?” Almost nobody answers the question that matters once the demo ends: how do I govern one?

Why now

The missing half isn't smarter actors.

Agents are leaving the chat window. They take keys, push branches, message customers, spend money. An autonomous actor without an audit trail isn’t automation — it’s a liability with an API token. The industry keeps shipping smarter actors; the layer that makes an autonomous actor accountable is the empty one.