The loop that cannot promote itself
An MIT set of revenue loops that climb only after a reviewed record, and cannot raise themselves.
- Category
- AI
- Who it is for
- A revenue team that already reviews what an agent decided overnight.
- Possible model
- Open source
- Difficulty
- Hard
- Inspired by
- Overnight GTM · Overnight GTM
- Original launch
- Watch the launch · Nov 12, 2026
Difficulty is an editorial estimate of technical and operational complexity, not a score.
Startup idea analysis by Startup Ideas. Last reviewed Nov 12, 2026. How we read a launch.
The startup idea
Cameron posted Overnight GTM on November 12. He said he built a working answer for AI-native rev ops and open sourced it. Twelve loops run every night, write down why they decided, and earn autonomy from their review record. The demo is overnight-gtm.vercel.app. The repository is cfagan17/overnight-gtm. The license is MIT. It sits on Atomic CRM and needs Postgres 16 and Python 3.11 or newer. The readme says the public demo is a fictional company, Northwind Data, with two months of generated history and one real Claude run. No account. It is wiped and rebuilt every night, so the history always ends last night. A loop moves up a level only when its record clears a bar: reviewed decisions, replay evals, holdout experiments, and zero policy violations. Agents have no direct write to the ledger, and they cannot raise their own autonomy. STATUS says one live run shows the loop works, not how often it is right. The promotion thresholds are starting guesses. Salesforce and HubSpot are not tested against a real system. The post says the loops earn autonomy. It does not say the company on the demo is fictional. This page did not run a loop.
The opportunity is the bar a person still has to clear, not a night shift that promotes itself. A smaller product is one loop, a written decision, and a rule that the agent cannot raise itself. A demo that treats Northwind as a customer is not this readme.
Do not call the demo a live book of business. Do not let the loop set its own level.
The problem
A revenue agent that writes its own promotion has no one left to refuse it.
Who would pay for this?
- Primary
- A revenue team that already reviews what an agent decided overnight.
- Secondary
- A person who will read that Northwind Data is generated.
- Early adopter
- One loop, the written reason, the cannot-promote line.
How could it make money?
Possible models. Nothing here is a claim about what Overnight GTM charges.
- No price. MIT.
- You still pay for the model the loop calls.
- The post names no fee.
What would the MVP look like?
- One overnight loop.
- A decision written down.
- A level the agent cannot raise.
- The fictional-company line on the demo.
How to validate the idea
- Ask five people whether Northwind was a real customer.
- If they say yes, the demo line failed.
- Do not raise a loop because one Claude run finished.
How to differentiate
- The record decides the level.
- The agent does not.
- MIT.
- Do not become the CRM.
Original product launch
This launch is inspiration and evidence that someone shipped, not a partnership. The idea above is our reading of the broader opportunity. Overnight GTM did not write it.
- Product
- Overnight GTM
- Source
- Overnight GTM
- Film
- Overnight GTM, the company is made up
- Posted
- Nov 12, 2026 · 1:21
Product that inspired this idea
Cameron's post says he built a working answer for AI-native rev ops and open sourced it. Twelve loops run every night, write down why they decided, and earn autonomy from their review record. The demo is overnight-gtm.vercel.app. The repository is cfagan17/overnight-gtm. The license is MIT. It needs Postgres 16 and Python 3.11 or newer. The readme says the public demo is a fictional company, Northwind Data, with two months of generated history and one real Claude run. It is wiped and rebuilt every night. A loop moves up only when its record clears a bar. The agent cannot raise its own autonomy. STATUS says one live run shows the loop works, not how often it is right, and the promotion thresholds are starting guesses. Salesforce and HubSpot are not tested against a real system. The post says the loops earn autonomy. It does not say the company on the demo is fictional. This page did not run a loop.
Posted by @camfagan. Watch the original launch. No separate product site is listed unless the post itself is the source.
- agents
- sales
Related startup ideas
- A coworker that stays on after the chat
An agent with its own computer that keeps a project page, and can be reached from the chat app a team already uses.
- A self-hosted assistant for personal apps
An assistant a person can host, pointed at the apps they already use, including computer use.
- A worker that lives in a folder
Give each worker a folder on the machine, a department, and a job, and say on the first screen whether it asks before it acts.
- Ask the computer, then approve the change
Put the agent on the desktop the person already uses, and stop it before it changes a file.
- Memory that survives the next coding session
Sell a memory layer that learns how one developer works, then carries that into the next agent run.
- The coach that gets skipped first
A MIT camera guide that checks the step, and drops the live look when the key hits 60 a minute.
Looking for more ideas?
The full list is startup ideas for 2026, each one tied to a launch you can watch.
Browse startup ideas for 2026