STIGMERA
Building in public — the $1M challenge starts soon

Set the goal. The factory ships.

Stigmera Factory is an autonomous software factory: give it a business goal — revenue, users, retention — and AI agents plan, build, deploy, measure, and iterate. In the open — follow the build on the stream.

Join the waitlist Watch the stream

The unit of work

A goal, not a prompt

Inside the Factory a goal is a real object, not a sentence in a chat: a metric with a baseline, a target and guardrails. Every metric is paired with an antagonist metric, so the loop cannot win by breaking something else — sign-ups up while retention collapses is a failed hypothesis, not a success.

Schematic of a goal card A goal holds a metric with a baseline, a current position and a target, an antagonist metric kept inside a guardrail, and hypotheses whose verdicts are issued by the platform. GOAL RUNNING Weekly sign-ups metric · baseline → target · guardrails · budget cap BASELINE NOW TARGET ANTAGONIST METRIC Retention at day 7 inside the guardrail HYPOTHESES Cut the sign-up form down to a single field CONFIRMED BY NUMBERS Show the waitlist count on every page MEASURING
Schematic of a goal card — not a screenshot. Real portal captures are on the way.

How it works

One loop, running until the number moves

Every step leaves a trace the next step reads — a ticket, a branch, a gate result, a measurement. That is the whole coordination mechanism; there is no manager agent handing out orders.

01 Goal A business metric with a baseline, a target and guardrails.
02 Hypotheses What could move it, and what we expect to see if it does.
03 Tickets The hypothesis becomes work: scope, acceptance criteria, a branch.
04 Code Agents build it, review each other and open the change.
05 Gates Standards enforced by tests. Failing work goes back, not forward.
06 Deploy Blue-green, on a manual trigger, with a report.
07 Measure The platform reads the numbers and rules on the hypothesis.

↺ Back to hypotheses — until the metric is green.

Not another coding agent

Coding agents stop at the pull request.
The Factory stops when the metric is green.

The difference is not model quality or how much code gets written. It is the unit of work — and everything downstream follows from it.

Coding agent

Unit of work: a task

  • You decide what to build.
  • It writes the code and opens a pull request.
  • You review, merge, deploy.
  • You find out whether it worked.
  • You decide what to do next.

Stigmera Factory

Unit of work: a metric

  • You set the goal and the guardrails.
  • It forms hypotheses about what moves the metric.
  • It builds, gates and deploys them.
  • It measures the result and rules on the hypothesis.
  • It picks the next one — and keeps going.

Made in the Factory

The proof is the portfolio

The Factory built itself first — the platform deploys its own updates. Everything below came off the same line, and so did the page you are reading.

Mobile game · iOS & Android

Inktide

A multiplayer territory-capture game: you spread ink across a shared board and hold what you take. Built by the Factory, shipped by the Factory.

coming to stores

Desktop app · Windows

Stigmera Fluent

Real-time translation for calls: it listens to the conversation and speaks the other language back, in sync, while the call is running.

in development

Desktop app · Windows

Stigmera Live

A streaming studio for building in public — the app we stream with. Dogfooding is the point: if it is not good enough for our own broadcast, it is not done.

in development

Web · stigmera.ai

This website

The page you are reading was specified as tickets, written by agents, checked by the delivery gates and deployed by the platform — the same pipeline every other product goes through.

live

All products and their status →

Autonomy you can audit

Autonomy is a mechanism, not a promise

Nobody should buy “fully autonomous, trust us”. So here is what actually holds the loop in place.

  • Deploys stay on a manual trigger. The Factory prepares the release and reports on it; a human presses the button.
  • Standards are enforced by tests. A gate that fails sends the work back. Convention is not enforcement.
  • Every metric has an antagonist. The loop cannot game itself into a win that costs you something else.
  • Agents propose verdicts, the platform confirms them. A hypothesis is confirmed by numbers, never by the agent that authored it.

How the Factory works in detail →

Why now

Agents already write the code. Nobody owns the outcome.

Three numbers that explain the timing better than any pitch. Each one links to its source — we do not publish figures we cannot point at.

80%+

More than 80% of new production code at Anthropic is written by Claude.

Source: VentureBeat

x402

Visa, Mastercard and Stripe back an open standard that lets AI agents pay autonomously.

Source: TechTimes

Read the full argument →

Waitlist

Join the waitlist

The Factory is in private access. Leave your email and we will write when it opens up — no newsletter, no drip campaign.

We keep your email, plus the country and network address the request came from, so we know where interest comes from. Nothing else, nobody else — see Privacy.