6,048 Contributions and Nothing to Show for It
I made 6,048 GitHub contributions in the last year. If I moved back to England, that would make me the sixth most active contributor in the country. In the US, where I actually live, I'm eighteenth.
I have 12 followers.
That ratio tells you something about my year, and it isn't flattering. I've been writing a staggering amount of code. Most of it sits in private repositories, powering products that almost nobody uses, built by a bloke who works full time at a bank and doesn't use automation tools for his coding. Just me, Claude Code, and whatever hours I can scrape together between incident response plans and sleep.
The numbers sound impressive until you ask the only question that matters: so what?
The builder's trap
I know how to build things. I can architect a multi-agent system from scratch. I can design a 132-factor assessment framework and implement it across edge functions. I can ship a benchmark platform, an llms.txt generator, a visibility scanner, and a trading bot, all while holding down a day job in business continuity.
What I cannot seem to do is sell any of it.
This is the bit where I'm supposed to have an insight that reframes the problem as a hidden advantage. I don't. The problem is exactly what it looks like. I'm a builder who doesn't distribute. Six thousand commits with twelve followers is the GitHub equivalent of writing novels and leaving them in a drawer.
The pile
While I've been committing code nobody sees, Lovable has been generating 100,000 new projects a day. Twenty-five million projects total at last count, valued at $6.6 billion. Every person with a product idea and twenty quid a month can now conjure a full-stack React app from a paragraph of English.
I'm fairly confident that 99% of that output will be rubbish. The data supports this. A CodeRabbit analysis of 470 GitHub pull requests found AI-generated code contains 1.7 times more major issues than human-written code. Security vulnerabilities run at 2.74 times the rate. Lovable itself had 170 of its 1,645 apps shipping with personal data exposed to anyone who knew where to look. Forrester reckons 75% of enterprises will face serious technical debt from AI-driven development this year.
The vibe coding hangover is real. These apps launch beautifully and rot quickly. They can't handle their first dependency update. They fall over at a hundred concurrent users. The developers who prompted them into existence can't debug them because they never understood the code in the first place.
I know all this. I find it reassuring. And that's the problem.
The moat I can't see past
Being right about the quality of the pile doesn't help me. The pile exists regardless of my opinion. A hundred thousand new projects a day creates a noise floor so thick that quality becomes invisible. When everyone has a landing page and a prototype, having a well-architected product with proper test coverage and a thoughtful API is like bringing a library card to a nightclub. Technically superior. Completely beside the point.
The market doesn't reward code quality at the discovery stage. It rewards distribution.
I've spent the last year optimising for the wrong metric. Contributions measure effort. Followers, stars, users, revenue — those measure impact. And my impact numbers are embarrassing.
What the contributions actually represent
Here's what 6,048 contributions look like when you're honest about it: a man with ADHD who finds building easier than selling, choosing the comfortable path and calling it productivity. Every commit feels like progress. The green squares fill up on the profile page. The architecture diagrams get more sophisticated. The specs get tighter.
Nobody cares.
I don't mean that as self-pity. I mean it as a statement of fact. The world does not owe me attention because I worked hard. Working hard is table stakes. The question is whether anyone knows the work exists, and whether it solves a problem they're willing to pay for.
My products might be good. AISearchArena might be the most rigorous independent GEO benchmark out there. LLMtxtMastery might be genuinely useful. I wouldn't know, because I haven't done the distribution work to find out.
The uncomfortable arithmetic
The vibe-coded pile will mostly die. That part I'm right about. Apps without maintainability don't survive contact with real users. Security holes get found. Technical debt compounds. The 100,000 daily projects will become 99,000 daily abandonments within months.
But some won't die. Some will find users, get feedback, iterate, and survive. And the founders behind those survivors will have learned more about their market in one weekend of shipping and talking to users than I've learned in a year of building in private.
That is arithmetic. Building in private at scale is still building in private.
What I should actually do
I'm writing this partly to think out loud and partly because someone else in my position might find it useful. Here's what I've landed on.
Stop treating GitHub as a scoreboard. The contributions graph is a vanity metric. It tells me I'm busy. I already knew I was busy. Busy isn't the problem. Invisible is the problem.
Pick one product and sell it. Not build it better. Sell it. Talk to potential users. Find out if the problem I'm solving is a problem they have. Do the uncomfortable work of putting myself in front of people and asking for money. Everything in my personality resists this. That's probably a sign it's the right move.
Accept that the vibe-coded pile is my competition, not my enemy. The pile lowers the barrier to entry, which means the barrier to differentiation shifts. It shifts from "can you build this?" to "can you maintain it, support it, and make it better over time?" Those are things I can do. But only if someone finds me first.
Use the 6,048 contributions as proof, not product. The contributions show I can build. The question is whether I can connect that building to someone who needs it. The code is evidence. The distribution is the trial.
The real question
I'm the eighteenth most active GitHub user in the US, with a full-time job, building real products with real architecture, by hand, in the hours between everything else. If I moved home to England, I'd be sixth.
And I have twelve followers.
The gap between those two facts is where the work needs to happen. Not more commits. Not better specs. Not another product. The work is: be findable. Be useful. Let people know the work exists.
That's harder than coding. It always was.
Jamie Watters builds AI-native products at jamiewatters.work. Find him on LinkedIn, X, and WIP.