Sri Rang
AI transforms open‑source.
And how open-source transforms the enterprise
@srirangan · srirangan.net · hello@srirangan.net
1 — The Old World - 1990-2025
Built on people, not just code.
- Pull requests as conversation, not just patches
- Code review as transfer of institutional knowledge
- Maintainers: the human who says yes, no, and not yet
- Trust built slowly, over years
1 — The Old World - 1990-2025
Slow by design.
A PR sitting for two weeks wasn't dysfunction. It was deliberation.
Linux. PostgreSQL. Python. Weren't fast. Were disciplined.
1 — The Old World - 1990-2025
Enterprise borrowed everything.
- Pull requests — came from open source
- Git — built for the Linux kernel
- CI pipelines — refined in public repos first
- Review culture — no one merges their own work
The best engineering orgs run like well-governed open source projects.
1 — The Old World - 1990-2025
That world still exists.
But something new entered the room.
2 — What's Actually Changed - 2025+
"AI helps you write code faster."
True. But the least interesting part.
What's really changing is the nature of participation.
2 — What's Actually Changed - 2025+
Agents don't just assist. They contribute.
They open issues, write code, submit PRs, respond to review comments, and iterate.
At a pace and volume no human can match.
2 — What's Actually Changed - 2025+
Already happening.
- Pull-Prompt request — human writes spec, agent writes code. The diff is real. The author is not.
- Continuous agent loops — Ralph loops runs AI agents against PRDs until every item is done. That's not a pipeline. That's an autonomous contributor.
- Spec-driven development — the skill shifts to writing precise, unambiguous instructions
2 — What's Actually Changed - 2025+
This is not the future.
Teams are doing this today.
And review quality across AI tools is now being benchmarked. The gap between best and worst is significant.
3 — What This Breaks
Three things under pressure.
- Provenance — who authored this?
- Maintainer burden — review volume scales with agent output
- Trust collapse risk — did anyone actually read this?
3 — What This Breaks
A single developer can generate 10 PRs a day.
Reading code nobody wrote — in the human sense — is different again.
The math doesn't work without help.
3 — What This Breaks
We measured it.
100 PRs · 8 production codebases Redis, Tauri, Firefox iOS, cal.com
3 — What This Breaks
"Who is watching the agents?"
Better tooling. Clearer norms. Humans who've decided what they must own.
4 — A Builder's View
I build Tusk.
Native PostgreSQL clients for macOS and GNOME. Zero telemetry. No Electron.
When I started using AI in that workflow, two things happened.
4 — A Builder's View
The barrier to entry collapsed.
Features that took a week now take a day.
Move fast through the unfamiliar. Slow down where judgment matters.
4 — A Builder's View
Speed is a double-edged sword.
AI gives you an enhanced ability to go in the wrong direction very quickly.
The F1 driver analogy: a faster car doesn't make you safer. It makes the cost of a mistake higher.
Engineering fundamentals became more important, not less.
4 — A Builder's View
Staying power is the differentiator.
A lot of projects spring up. Most of them die. Not because the code was bad. Because nobody committed to the long game.
AI lowered the floor. It didn't raise the ceiling.
5 — The Bigger Picture
Open source is the leading edge.
CI came from open source. Code review culture came from open source. All of it percolated into enterprise before becoming standard.
AI-native development will follow the same path.
5 — The Bigger Picture
The developer identity question.
Coder → Supervisor
Define intent clearly. Review outputs critically. Own the decision about what ships.
Taste, judgment, understanding of the problem. Things that are not in the model.
5 — The Bigger Picture
This is not a threat.
It's a reframe.
That hasn't changed.
The best software comes from people who care about it over the long run.
If anything, it matters more now.
Sri Rang · srirangan.net · @srirangan