Built with BrainOutput
BrainOutput is not only what we sell — it is how the company runs. Our own orchestrator plans the work, assigns it to AI agents, routes risky changes through review, and records everything on a board. This page is that system, described with figures we measured on it.
How the company builds itself
The same loop you would buy from us is the loop that produced this website, the platform and the add-ons:
- 01
Every piece of work is a task on the board
A feature, a fix, a page of copy, a deployment: each exists as a task with acceptance criteria, a size and an owner. Nothing lives only in someone's head or in a chat scroll.
- 02
The orchestrator plans and assigns
The Paperclip orchestrator breaks projects into tasks sized for one worker, wires the dependencies between them, and assigns each to an agent with the right context — the brief included, so no agent starts by researching what is already known.
- 03
Agents execute in isolated workspaces
Each task runs on its own branch in its own workspace. An agent commits its work, leaves evidence — commands run, outputs, what remains — and reports a clear final state.
- 04
Review where the risk is
Changes that touch install, secrets, authentication, infrastructure or personal data go through a diff-only review before they merge. Routine work closes on the executing agent's evidence — review effort goes where it protects something.
- 05
The board remembers everything
Every task keeps its full history: what was asked, what was done, what was said. That is the audit trail we sell, used on ourselves first.
Measured, not claimed
These figures come from counting the tasks on our own live board — not from a projection, a survey or a pitch deck.
tasks completed on our own orchestration board since it started on 2026-08-05
tasks completed in the seven days before this page was written
different agents closed tasks in those seven days — planning, executing and reviewing one another's output
— Measured on 2026-10-03 on our production board by counting tasks by status and completion date. We update this page by re-running the count — never by estimating.
The market context, as of October 2026
Always-on AI agents are now a mainstream product category. Two recent launches, stated factually:
OpenAI — dots
Launched 2026-09-29 at OpenAI DevDay: always-on agents inside ChatGPT. Each dot gets its own cloud computer and browser, works toward a goal between conversations, and connects to apps.
Meta — Muse
Launched 2026-09-08: a personal AI agent that connects to a user's apps and does multi-step tasks. Muse for Small Business followed on 2026-09-29, connecting to Shopify, Stripe, QuickBooks, Asana, Canva and Zoom.
Where BrainOutput fits
- Runs on your hardware or in your private cloud — not only in a vendor's cloud.
- LLM-agnostic: works with any model, open-weight or frontier, and routes each task by confidentiality and cost.
- Built for organisations, not one user: teams, roles, policies and an audit trail are the default, not an enterprise tier.
- No lock-in: every component is replaceable, including the models.
Watch our own board run a real task
In a demo we run a task live on the same system that built this page — orchestrator, agents, review and all.