What if an AI could spend days learning your entire codebase before writing a single line of code?
Been thinking about why current AI coding tools still break down inside large enterprises.
2 massive problems:
- they never have enough deep context of huge old codebases
- companies don’t want proprietary code/context sitting on external AI clouds
So I started building this:
Idea is simple:
instead of temporary context windows, the system continuously learns the company over days by understanding:
- repos
- jira
- docs
- infra
- PR history
- architecture patterns
- ownership
Eventually it behaves less like a copilot and more like an engineer who has worked at the company for years.
Then teams can directly assign tickets to it and it investigates systems, writes code, tests and opens PRs inside the company’s own AWS/VPC.
Feels like AI coding tools will eventually evolve from “assistants” to “persistent engineering systems”.
Would genuinely love thoughts from people working in large engineering orgs. And I am raising pre seed, would love to get in touch with investors.