explainx / blog
commit-history.com ranks GitHub users by lifetime public, private, and total commits — like star-history.com for stars. Who leads, how it works, and whether commit count still means anything in the AI coding era.
Jun 27, 2026
Git explained from scratch: what it is, how to install it on Mac and Windows, and exactly how to push code to GitHub with real commands you can copy right now.
Jun 25, 2026
World Monitor (github.com/koala73/worldmonitor) is an open-source real-time global intelligence dashboard aggregating 500+ curated news feeds across 15 categories, AI-synthesized into briefs. It ships a dual 3D/flat map engine with 56 layer types, a Country Instability Index covering 31 nations, a finance radar across 29 exchanges, local AI via Ollama, and a native desktop app. It has accumulated 59,200 GitHub stars and 9,300 forks. Here is what it is, how it works, and how to run it.
Jun 24, 2026
BirdClaw is an open-source CLI and web app by Peter Steinberger that imports your Twitter archive, syncs live updates, and stores everything in a local SQLite database. Daily AI digests, bookmark search, follow graph queries, DM triage, and blocklist management — all offline-first, no ads, no algorithm.
On June 28, 2026, a small side project called Commit History hit developer Twitter the way a leaderboard always does: everyone looked themselves up, the people at the top posted screenshots, and everyone else argued about what the numbers actually mean.
The site — commit-history.com, built by peetzweg — does for GitHub commits what star-history.com did for repository stars: plot cumulative totals over a user's lifetime and rank the all-time leaders. Public commits. Private commits (for users who expose them). Combined totals. One URL per username. Instant bragging rights — or instant existential crisis.
In the first 48 hours, indie hackers discovered they had more private commits than most engineers have public ones. Open-source maintainers found themselves outside the top 100 despite decades of work. And the AI-agent crowd asked the question that defines 2026: if Claude Code commits for you, does the number still count?
| Question | Answer |
|---|---|
| What is it? | Lifetime GitHub commit charts + global leaderboards at commit-history.com |
| Who built it? | peetzweg — explicitly "a homage to star-history" |
| Who leads public? | Dicklesworthstone (~145k), rgbkrk (~132k), leodemoura (~98k) as of June 29 |
| Who leads private? | brahimhamichan (~601k exposed private), steipete (~208k), levelsio (~162k) |
| Who leads combined? | steipete (~268k public + private) |
| Is it official GitHub? | No — third-party site reading public GitHub data + opt-in private stats |
| Does high count = skill? | No — volume ≠ quality; agents and commit style distort totals |
| Is it free / need login? | Free, no login — any username searchable |
| Why is my count "wrong"? | Squash merges, wrong git email, org attribution, or private visibility off |
| How to show private commits? | GitHub Settings → Profile → enable private contributions |
| Does GitLab work? | No — GitHub usernames only |
| Do AI agents inflate this? | Yes — granular agent commits add up fast over months |
Most traffic to Commit History does not stop at the homepage. People look up steipete, feel inadequate, then search specific follow-ups. These are the ones that actually came up in launch-week threads — answered directly.
Nine causes, in order of how often they explain a gap:
[email protected] do not count until that email is verified on your GitHub account under Settings → Emails.Quick diagnostic: run locally on a repo you own:
git shortlog -sn --all --author="$(git config user.email)"
Compare that author count to what GitHub attributes — if local is higher, your email linkage is the problem, not Commit History.
Because ~77% of his counted commits are private (~208k private vs ~61k public as of June 29, 2026).
Public rank rewards long-visible OSS history (Dicklesworthstone, rgbkrk, theorem-prover maintainers). Combined rank rewards total git volume across a career, including private agent experiments, client repos, and unreleased tooling — which is exactly how OpenClaw-era shipping works.
Same pattern for fabpot (~88k public, ~168k private, ~255k combined): Symfony's public face plus years of private Symfony Ltd work.
Takeaway: Public leaderboard ≈ "who maintained visible open source longest." Combined leaderboard ≈ "who touched git most overall, including hidden repos."
At ~601k exposed private commits, brahimhamichan is 3× steipete and 3.7× levelsio on the private list alone — the largest outlier on the entire site at launch.
Commit History does not publish repo names for private totals, so explanations are inference:
| Hypothesis | Plausibility |
|---|---|
| Daily commits across many private repos for 10+ years | High |
| Granular commits (no squashing, frequent WIP pushes) | High |
| CI/automation committing under personal account | Possible — misconfigured hooks happen |
| Monorepo with automated snapshot commits | Possible |
| Data artifact or double-counting | Low but unverified — treat as unconfirmed until peetzweg documents methodology |
Do not treat 600k as a target. Treat it as proof the leaderboard measures events, not normalized engineering output.
Three steps — no Commit History account needed:
What this does not do: expose repo names, commit messages, or code. It only allows aggregate private commit counts to be read by tools GitHub permits — same as your profile graph.
Privacy note: If you enable this, anyone can see that you commit privately on some days — not what you committed. Recruiters and curious strangers already used this setting pre-Commit-History; the site just sums it.
Commits that land on the default branch and GitHub attributes to you count — whether you pushed directly or merged via PR.
What does not count the same way:
| Action | Counted as commit on leaderboard? |
|---|---|
| Commits on a feature branch, then squash-merged | Usually 1 (squash commit) |
| Commits on a feature branch, then rebase-merged | All preserved commits |
| Force-pushed away before merge | Gone from history |
| Co-authored commit | Counts for each listed co-author on GitHub |
| Suggested change accepted on someone else's PR | Often counts for committer, not suggester |
| GitHub web editor single-file commit | Counts — yes |
This is why two engineers with identical merged LOC can differ by 5–10× on Commit History: merge strategy, not effort.
Yes. Rough order-of-magnitude for heavy agent users in 2026:
| Workflow | Commits per focused coding day | 250 workdays/year |
|---|---|---|
| Manual, squash at end of day | 1–5 | ~250–1,250/yr |
| Manual, commit per logical change | 5–15 | ~1,250–3,750/yr |
| Claude Code / Cursor Agent, commit per loop iteration | 20–80 | ~5,000–20,000/yr |
Agent + no squash + long-running /loop sessions | 100+ | Tens of thousands/yr |
Over five years of agent-native shipping, that is how private totals reach six figures without "fake" data — just high-frequency git events.
That does not mean the human did nothing. It means the harness chose to commit often. See loop engineering with Claude Code for why agents default to commit-per-action.
Hiring implication: Ask for PRs and design docs, not Commit History screenshots.
Because the leaderboard rewards volume over employer prestige:
Arnav Gupta noted on X his public commits sit just outside the top 100 — between Guillermo Rauch and Addy Osmani — which matches "senior visible engineer, not daily micro-commit indie hacker."
None of this ranks impact. Stripe engineers do not need 150k commits to matter.
| Concern | Reality as of launch |
|---|---|
| Official GitHub product? | No — side project by peetzweg |
| Need OAuth or token? | No for basic username lookup |
| Can it read private repo code? | No — only aggregate counts if you opted in on GitHub |
| Does searching expose your username? | Yes — "Recently looked up" ticker displays searched usernames publicly |
| GDPR / delete my lookup? | Unknown — no documented removal flow at launch; assume lookups are public |
If you do not want to appear in "Recently looked up," wait until that feature changes or ask peetzweg — do not search yourself speculatively if that bothers you.
| Tool | Measures | Unit | Leaderboard? |
|---|---|---|---|
| star-history.com | Repo popularity | Stars over time | No (repo-focused) |
| commit-history.com | Person output | Lifetime commits | Yes (global) |
| GitHub profile graph | Person activity | Contributions/day (rolling year view on profile) | No |
| GitHub Achievements | Gamified milestones | Badges | No |
Use star-history in READMEs to show project traction. Use Commit History for career commit narrative — knowing the caveats above.
No — unless you want a higher number for its own sake.
If you want commits to mean something when agents are involved:
# Before merging agent work to main — squash WIP on a feature branch
git checkout feature/agent-run
git rebase -i main # squash "fix", "wip", "try again" into one commit
git checkout main && git merge --squash feature/agent-run
git commit -m "feat: add payment webhook handler (agent-assisted)"
Better metric for teams: merged PRs with tests, incident resolution time, or revenue — levelsio's actual scoreboard.
The homepage is deliberately minimal. You type a GitHub username, get a cumulative commit chart, and can flip between public, private, and both views.
Three features drove the viral loop:
Private rankings only include users who have enabled private contribution visibility on GitHub. That is a critical caveat: the private leaderboard is not "all private commits on GitHub" — it is "private commits from people who opted to show them." The true private kingpin is almost certainly someone who keeps private stats hidden.
Numbers move as GitHub ingests new data. These reflect what Commit History displayed at launch plus the first day of lookups.
| Rank | Username | Public commits | Known for |
|---|---|---|---|
| 1 | Dicklesworthstone | 145,459 | High-volume open-source / automation output |
| 2 | rgbkrk | 131,866 | Jupyter ecosystem, developer tooling |
| 3 | leodemoura | 97,966 | Lean theorem prover, Microsoft Research |
| 4 | emmikat | 95,817 | Sustained OSS contribution volume |
| 5 | fabpot | 87,760 | Fabien Potencier — Symfony, PHP ecosystem |
| 6 | Niek | 81,045 | — |
| 7 | fabaff | 75,292 | — |
| 8 | gaearon | 68,597 | Dan Abramov — React, Redux |
| 9 | steipete | 60,634 | Peter Steinberger — OpenClaw, iOS tooling |
| 10 | peppy | 56,430 | osu! game developer |
| Rank | Username | Private commits | Known for |
|---|---|---|---|
| 1 | brahimhamichan | 600,642 | Extreme private volume — an outlier even on this list |
| 2 | steipete | 207,708 | Agent-native shipping; OpenClaw creator |
| 3 | fabpot | 167,678 | Symfony ecosystem |
| 4 | levelsio | 161,515 | Pieter Levels — indie hacker, Nomad List, Photo AI |
| 5 | cheshire137 | 143,030 | — |
| 6 | designorant | 134,811 | — |
| 7 | zkochan | 129,923 | pnpm creator |
| 8 | aarmful | 112,836 | — |
| 9 | ageesen | 111,584 | — |
| 10 | tommoor | 95,071 | — |
Add public + private for the names X cared about most:
| Username | Public | Private | Combined |
|---|---|---|---|
| steipete | 60,634 | 207,708 | ~268,342 |
| levelsio | (lower public) | 161,515 | ~161k+ private alone |
| fabpot | 87,760 | 167,678 | ~255k |
| Dicklesworthstone | 145,459 | 43,369 | ~189k |
Peter Steinberger's combined total made him the poster child for the launch — not because public commits rank first, but because he is visibly building OpenClaw and shipping with AI-native workflows while sitting at the top of the both leaderboard. Pieter Levels posted that he had the most private commits among exposed profiles alongside steipete — which, for an indie hacker famous for shipping dozens of micro-products, scans as on-brand.
Vincent Koc landed in the top 25 all-time on combined lookups. Jay Meistrich pointed people to the site for exploration. banteg noted finding himself "in good company." The pattern was consistent: the leaderboard rewards people who have been committing daily for a decade or more, often across many repos, often with agents now accelerating the pace.
Before treating Commit History as a skill score, understand what increment the site is summing.
GitHub counts commits authored on the default branch (and contributions via merged PRs depending on settings). What it does not normalize:
| Factor | Effect on leaderboard |
|---|---|
| Squash vs merge | Squash-merge = one commit per PR. Rebase-merge = many commits preserved. Same work, different totals. |
| Monorepo vs many repos | Splitting work across 40 repos does not change commit count; neither does consolidating — but CI bots and auto-formatters can add noise. |
| Commit granularity | "WIP" commits every 10 minutes vs one commit per feature — same code, wildly different counts. |
| Bots and automation | Dependabot, formatting bots, and release scripts commit as the bot user — usually not attributed to you, but misconfigured hooks can blur lines. |
| AI agent loops | Agents that commit after each file change inflate totals fast. See loop engineering with Claude Code. |
The Git beginner's guide on explainx.ai covers what a commit actually is — a snapshot pointer in git history, not a unit of value.
Bottom line: Commit History measures git event volume, not engineering impact. A single well-designed API migration might be three commits. A vibe-coded prototype might be three hundred.
This is the debate that made the launch more than a novelty.
In June 2026, a meaningful slice of the leaderboard's upper tier ships with coding agents in the loop. Peter Steinberger's public persona is explicitly agent-native — terminal agents, harness design, high-frequency iteration. Pieter Levels has talked for years about using AI to accelerate indie product development. Neither hides that machines write a lot of the diff.
Three honest positions emerged on X within 24 hours:
The optimistic read: cumulative commits correlate with years of showing up. The top public names include maintainers who held critical open-source infrastructure for a decade (fabpot/Symfony, gaearon/React, leodemoura/Lean). The number is a proxy for ** sustained output**, not a lottery ticket.
The skeptical read: when Claude Code or Cursor Agent can commit, run tests, and retry in a loop, commit count measures harness aggressiveness as much as human effort. Two developers with the same merged LOC can differ by 10x in commit count based on agent settings alone.
The pragmatic read: star-history was never a quality metric either. It was a storytelling tool — "look how this repo grew." Commit History is the same for people. Look up your username, share the chart, get back to work.
explainx.ai's read: all three are partially true. The leaderboard is real data with real career correlation at the extreme top — nobody accidentally commits 200,000 times. But in the AI-agent era, the margin between ranks 500 and 5,000 tells you less about skill than it did in 2020.
For the full stack of what actually makes agents ship reliably — prompt, context, loop, harness — see Context vs Prompt vs Loop vs Harness Engineering.
peetzweg (GitHub: @peetzweg) built Commit History as a direct homage to star-history.com, the charting tool developers use in READMEs to show repo growth curves. The insight was simple: stars measure project attention; commits measure person persistence — and nobody had made the second one easy to look up.
The site itself is lightweight: username in, chart out, leaderboard for comparison. peetzweg added a "Buy me a coffee" link — standard indie-tool monetization for something that will cost real money in GitHub API quota if it stays viral.
Recently looked-up users on launch day included a mix of indie hackers, Canva engineers, and agent builders — a cross-section of exactly who commits most in 2026.
| Limitation | Why it matters |
|---|---|
| No LOC / impact weighting | Deleting 10,000 lines and adding 10,000 lines counts like steady progress |
| No org attribution split | Work commits vs personal commits are blended |
| Private data is opt-in biased | Private leaderboard over-represents people who expose stats |
| No AI vs human split | Impossible without commit-message analysis — and even that fails |
| No time decay | A burst in 2024 counts equal to a burst in 2014 |
| Not affiliated with GitHub | Data freshness and accuracy depend on peetzweg's pipeline |
For engineering evaluation, use PR review quality, incident ownership, system design — not a public scoreboard. For curiosity and culture, Commit History is excellent.
If your number is lower than expected: check whether you squash-merge, whether most work lives in org repos under a different email, or whether you simply have not enabled private visibility.
If your number is higher than expected: check for old bot commits, fork sync patterns, or an agent harness that commits aggressively — all common in 2026.
If you are building with agents and want commits to mean something: adopt intentional git hygiene — meaningful commit messages, squash before merge on shared branches, and separate agent WIP branches from main. That is loop engineering applied to version control, not just model prompts.
Commit History arrived the same week developers argued about token economics pushing agent adoption and types of AI agents maturing from chatbots into autonomous loops.
The pattern across all of it: easy metrics get popular fast; hard metrics stay rare.
None replace the others. Commit History fills a gap star-history never could — personal output over a career — at the exact moment when "personal output" is the hardest thing to define.
Whether that makes you feel inspired or skeptical probably says more about how you ship than where you rank.
Commit History (commit-history.com) is peetzweg's star-history-style leaderboard for lifetime GitHub commits. It went viral around June 28–29, 2026, with Dicklesworthstone leading public commits, brahimhamichan leading exposed private commits, and steipete (~268k combined) dominating the conversation on X.
The site is fun, flawed, and culturally timed: it turns git history into a scoreboard at the moment AI agents make git history easier to inflate. Use it for curiosity and storytelling — not hiring decisions.
External: commit-history.com · star-history.com · GitHub profile contribution settings
Leaderboard figures and usernames reflect commit-history.com as observed on June 29, 2026. Rankings update as GitHub data refreshes and users change private visibility settings.