The Autoresearch Loop
A continuous agent scans your analytics, your competitors, trending topics and AI-search citation gaps to surface what to create next. Keeps what performs, discards what doesn't, repeats.
ContentMachina is an autonomous engine for creators and companies. It researches what to say, writes it in your voice, ships it everywhere, and learns from what works — on a loop. You set the direction in plain English. It runs the experiments.
No dashboards to babysit · Self-hostable · MIT licensed
Stop being a meat computer for your own marketing.
Great content used to mean a human, manually, between meetings, writing and posting and measuring — forever. That era is ending. ContentMachina applies the autonomous-research loop to growth: you program the engine in plain English, the agents run the experiments.
Inspired by Andrej Karpathy's autoresearch — agents that experiment on a loop, overnight, while you sleep.
No 40-tab dashboard. You connect, you point it at a goal, and the loop takes over.
Plug in your repo, your channels and your past content. ContentMachina reads it all and learns your voice, your topics and what your audience actually responds to.
Tell it your goals in plain English — "grow our community," "launch v2," "own the GEO topic." No prompt engineering, no rule-builder mazes.
It researches, drafts, ships, measures and improves — continuously. You review a digest and steer when it matters. The engine gets sharper every cycle.
Each part feeds the next — perception, identity, leverage, distribution, and a feedback loop that compounds.
A continuous agent scans your analytics, your competitors, trending topics and AI-search citation gaps to surface what to create next. Keeps what performs, discards what doesn't, repeats.
Learns your tone from your existing posts, docs and repo. A brand guardian flags anything off-voice before it ships — so autonomous output still sounds unmistakably like you.
Drop in one long-form asset — a talk, a podcast, a blog, a release — and get shorts, X threads, LinkedIn posts and a newsletter draft. Each one tuned to its platform, not copy-pasted.
Connects to your stack — GitHub, Linear, your release notes. Turns every shipped update into launch posts, changelog entries and build-in-public content, automatically. Your shipping cadence becomes a growth flywheel.
Choose your level per channel: Draft → Copilot → Autopilot → Self-Driving. Slide it up as trust grows. Nothing ever ships outside the guardrails you set.
Schedules and publishes across X, LinkedIn, YouTube and your newsletter. Recycles evergreen winners and auto-plugs your product the moment a post starts taking off.
Every post feeds performance data back into the engine. It tells you exactly what to update, regenerates it, and gets measurably better at your audience with every cycle.
Tracks your answer share — how often ChatGPT, Perplexity and Claude cite you — and engineers content to win those citations. Generative-engine optimization, built in.
Mines Hacker News, Reddit and dev communities for conversations worth joining — and drafts authentic, on-voice replies for your review. Show up where your audience already is.
Borrowed from Karpathy's autonomy slider: start with the engine asking permission for everything, end with it self-driving. Per channel. Reversible anytime.
The engine researches and drafts every piece, then waits. Nothing leaves your review queue without an explicit green light from you.
ContentMachina is open source and self-hostable. Run it on your own infrastructure, with your own models — and turn everything your team ships into content.
$ npx contentmachina init
✓ connected github.com/acme/engine
✓ learned brand voice from 142 posts
✓ mapped channels: x · linkedin · blog · newsletter
$ contentmachina run --goal "grow our audience"
→ researching 6 trending threads in your niche
→ drafting changelog post from PR #318
→ repurposing 1 release note → 4 platform posts
✓ shipped 3 pieces · autopilot · digest sent
→ learning +2 winners promoted to evergreen ContentMachina is being built in public. Join the waitlist for early access, or follow along as the engine takes shape.