Steven Tey's daily commits jumped from 5 to 24 (+380%). Lines of code rose from 891 to 4,363 (+390%). Active hours increased from 3 to 8 (+167%). His timeline transforms from sporadic activity to sustained, dense productivity.
This isn't an outlier—it's the extreme end of a pattern visible across elite developers. We analyzed 18 prominent open-source developers across two six-month periods: pre-AI (Jun-Nov 2022) and pre-agentic-AI (Nov 2024-Apr 2025). The recent period captures mature human-in-the-loop collaboration, before autonomous agents became widespread.
- 53% increased daily commits (median +25% among increasers)
- 67% increased lines per day (median +91% among increasers)
- Median across all developers: 0% commits/day, +2% LOC/day
- Range: -40% to +380% for commits; -76% to +390% for lines
The "AI raises the floor" narrative dominates discussion. The data shows the ceiling rose too—some already-elite developers found massive additional leverage.
Each developer name below links to their Vibe Gain Explorer page with complete timelines, session patterns, detailed metrics, and methodology.
Increased Output
Several developers dramatically increased their output during the AI era.
Steven Tey (Dub creator, Vercel contributor) anchors the high end. The numbers in the opening (+380% commits, +390% lines) only tell part of the story—his timeline transforms from sporadic bursts to sustained daily output.
Michael Klishin (RabbitMQ core maintainer): daily commits up 67% (6→10), lines up 177% (198→548). Commits per hour doubled while maintaining focused work sessions.
Henrik Rydgård (PPSSPP emulator creator): commits up 83%, lines up 125%. Active days held steady (162 to 163), active hours per day rose from 4 to 6, commits per hour doubled.
Loading charts...
Anthony Fu (Vue.js core team) maintained exactly 9 daily commits in both periods but more than doubled daily lines (+119%, from 3,011 to 6,583). Lines per commit jumped from 24 to 34 (+42%), achieved while reducing active hours from 4 to 3.
Loading charts...
Mitchell Hashimoto (HashiCorp founder): stable commit rates (7→7 daily) but 61% increase in lines per day (337→543). Active hours increased from 3 to 4, with time spans increasing from median 318 to 470 minutes.
Loading charts...
Phil Wang (lucidrains, prolific PyTorch paper implementer): commits up 25%, lines up 62%. Commits per hour doubled, session productivity improved significantly (commits per session from 2 to 3, lines per session from 61 to 115).
Loading charts...
The pattern across these developers: not just faster coding, but more coding. All increased active hours. The data suggests AI enables sustained high-output development that was previously difficult to maintain.
Efficiency Gains
Other developers maintained or modestly increased commit frequencies while dramatically improving efficiency.
Simon Willison (Datasette creator, Django co-creator): exactly 7 commits per day in both periods. Efficiency improved substantially: lines per hour more than doubled (30→67, +123%), lines per commit increased 67% (6→10).
Loading charts...
Linus Torvalds (Linux and Git creator): daily commits up 20% (5→6), lines per hour up 23% (219→270). Lines per commit grew from 79 to 105 (+33%).
Loading charts...
Andrew Kane (Ruby data tools creator): 9→11 commits daily (+22%), maintained across more active days (136→150).
Loading charts...
The pattern: same or slightly higher commit frequency, significantly higher output per hour. The ceiling rose through efficiency.
Different Trajectories
Some patterns defy simple categorization—neither straightforward output increases nor efficiency gains, but structural shifts in how work distributed.
Georgi Gerganov (ggml/whisper.cpp creator): lines per day dropped 73% (251→68), but active days jumped from 90 to 138 (+53%). Shifted from intense bursts to consistent daily work.
Loading commit activity data...
Loading commit activity data...
Thibault Duplessis (Lichess creator): one of the highest commit volumes in both periods, with substantial increases in the AI era. Pattern suggests AI can amplify already-intense development practices.
Loading commit activity data...
Loading commit activity data...
Steven Tey wasn't struggling before AI—he was already prolific. He found ways to sustain 380% more output. Michael Klishin, Henrik Rydgård, Anthony Fu, Mitchell Hashimoto, Phil Wang: all already-exceptional developers who discovered additional capacity.
The ceiling rose.
Credits
Concept by @restlessronin. Initial analyses (based on llm-readable vibe-gain statistics) created in agent mode by @gemini-3-pro (anti-gravity) and @claude-sonnet-4.5 (claude-code). Rewrite by @claude-opus-4.5. Copy refinements by @claude-sonnet-4.5.
