Developer-first Git hosting: built for the person who decides what ships
Everyone claims to love developers. The claim only means something when it tells you who loses. Developer-first names the winner when interests conflict: the person writing and shipping the code beats the platform, the agent, the org-as-watcher, and the investor.
If a principle can't lose you a decision, it isn't a principle. This one can.
Developer vs. platform
The platform wants you to stay, discover features, get upsold, and never leave. You want to finish and get out.
Codebahn optimizes for less time in the app, not more. No engagement metric drives the roadmap. No "come back" notifications. Export is one click, standard format, no ticket. The forge that respects your time doesn't fight you on the way out.
Developer vs. agent
Agents generate volume: PRs, comments, suggestions. Volume drowns attention.
The rule: agents propose, humans dispose. The human's review queue is sacred. It holds only what a human must decide on. An agent opening forty PRs produces one summary to triage, not forty notifications.
This is the exact inverse of the Copilot-first pivot: GitHub reshapes the UI around the AI. We reshape it around the person the AI answers to. Agents connect over standard protocols (MCP), as tools the developer wields. Never as co-owners of the workspace.
Developer vs. org
The org that pays wants dashboards, scores, productivity telemetry built from developer activity.
Codebahn gives teams coordination, not surveillance. Team-level flow, yes. A per-person commit scoreboard that turns someone's work into a stick to beat them with, no. We serve the person doing the work, not the person watching them. Even when the watcher signs the invoice.
Developer vs. investor
VC roadmaps bend toward the next round. Terms change after acquisition. GitHub was independent, then Microsoft. The terms survived; the trajectory didn't.
Codebahn is bootstrapped. The roadmap answers to the people using the product. Independence isn't a vibe. It's whose email changes the roadmap.
What it looks like in the product
- The terminal is first-class. The CLI isn't a companion to the web UI; it's a peer.
- The web UI optimizes for judgment. Review, decide, ship. Not "do everything in-browser because we want you in-browser."
- Latency is a feature. Every second of lag in the core loop is a tax on the person. Speed isn't polish; it's respect.
- Defaults favor safety over configuration. The human shouldn't have to become a config expert to avoid footguns.
Why this survives AI
Developer-first is not a bet that humans keep typing. It's a bet that the human stays the principal: the one with judgment, accountability, and a job on the line. The agent is what the word means in economics: something acting on the principal's behalf.
The more the agent does, the more the human's remaining job concentrates into exactly what we serve: review, taste, deciding what ships, owning the mess when it breaks. In a world where an agent writes 90% of the code, who picks the forge? Not the agent. It has no budget and no career. The human.
Build for the person who decides what ships. Let the agents help them.