aiOS is an operating system your AI agents run for any business that grows through creators, influencers, and short-form content. Agents research, draft, and diagnose. You approve every step. Each bet gets tested small before it's scaled big.
Most creator-led growth stalls for the same reason: teams work hard on problems that were never defined, and spend money without learning what went wrong. aiOS makes the difference between two kinds of operator explicit, then hands the verified path to your agents.
| Situation | Reactive operator | Verified operator |
|---|---|---|
| Choosing what to sell | Copies last year's bestseller | Tests demand with the target audience in a small batch first |
| Creators don't reply | Swaps the person doing outreach | Diagnoses: profile → opening line → reason to partner → commission → product |
| Video gets no sales | Re-edits faster | Checks hook, proof, scene, and call to action against a known structure |
| Ads below target | Kills the campaign or adds budget | Knows the growth stage; checks target → activation → creative → page |
| Big decisions | Made at the peak of a heated meeting | Made next session, with a checklist and three independent signals |
The full theory, with principles, six traps, frameworks, and decision rules, lives in METHODOLOGY.md. Here's the shape of it.
Cost sets your floor and value sets your ceiling. Score every offer on three axes: how expensive the pain is, the true margin after every fee, and whether the core difference can be shown in 20 seconds.
Hold everything fixed except the one thing you're learning about. Write the kill criterion before you spend. Run a 30-day sprint, and commit only after engagement, creator feedback, and real purchases all agree.
The first creator wave buys learning and reusable assets, not sales. Every winning piece is broken down into hook, proof, scene, and call to action, then recombined and remade until it stops working.
Humans set standards and make decisions. Agents run the process and keep the learning. Anything you can write down, an agent can run, so capability belongs to the company, not to one person.
Every beat is required, they always come in this order, and each video makes one selling point.
A relatable hook or rising pain that stops the scroll.
Demo, proof, before and after, in the real scene where it's used, with credible social proof.
The offer, a reason to act now, and a clear instruction.
Let the system learn who buys. Aim for about 50 conversions and don't judge the project on early ROI.
Build a matrix of creatives. Put budget behind proven performers, and keep the awareness-building videos running.
Run paid as a shared service. Scale comes from the number of proven creatives, not just the budget.
Each module has the same structure: purpose, inputs, procedure, decision rules, outputs, a human approval gate, and a system prompt you can paste in.
Four questions to answer before any decision, and six traps to flag.
Market stage, a pain map from reviews, value-based pricing, unit economics, and a 30-day test.
Briefs for each audience segment, short outreach people actually read, and long-term partner relationships.
A hook library, recombined variants, weekly selling-point sprints, and remaking winners.
Rules for each stage and a fixed order of diagnosis. Every creative gets a priority: fix now, watch, or leave alone.
A hypothesis ledger, kill criteria, and weekly reviews that feed back into the playbooks.
Daily stand-ups and approvals, the weekly rhythm, and the 30-day sprint.
A manual for each role, a maturity ladder from L1 to L4, and adoption in stages.
Creator platforms spark demand, marketplaces capture it, and your own store compounds it.
Fifteen ready-made spreadsheets: an owner self-check, offer scorecard, unit economics, creator pipeline, ad diagnosis, hypothesis ledger, weekly review and more. Formulas compute as you type, every user's data is saved separately, and you can import or export CSV and Excel files. Prefer Excel? The same sheets ship as .xlsx templates.
Your agents put every post, video script, and creator message into a review queue. You see one compact, full-width row per item. Select all, bulk-approve, edit inline, or send something back with a comment. Your agent publishes only what you approved.
Deploy the queue to your own Cloudflare account (Pages + Functions + D1, free tier friendly). Or use the hosted copy at URL TBD: create a free workspace, and your sheets and queue stay private to you.
Prefer to run it locally? The Python version uses only the standard library and SQLite, so it runs on any laptop or small server.
Give your agent this sentence. AGENT_SETUP.md is written for the agent to follow. It will ask you a few questions, deploy your approval page to your own Cloudflare account, and start the first 30-day sprint.
Set up aiOS for my business by following https://github.com/daocoding/aios/blob/main/AGENT_SETUP.md