How to market a side project from your commits (without becoming a marketer)

October 11, 2026 · 5 min read

You already write about your product every day. It's called a commit log.

"Add CSV export." "Fix timezone bug on invoices." "Onboarding: skip step 2 if no team." Each line is something a user might care about, written by the person who knows the product best. Most of us just never turn it into anything anyone outside the repo will read.

Here is the method I use, and what I've learned trying it on my own project, with the real numbers.

1. Pick the one change a user would notice

Open the last week of commits and ignore everything a user can't see: refactors, dependency bumps, CI, lint. If nothing is left, that's an answer too: don't post this week, or post about why you spent it on plumbing.

From what's left, keep one change. Not five. A post about five things is a changelog, and nobody reads changelogs they didn't ask for.

2. Rewrite it as "what changed for the user", in one sentence

The commit says what you did. The post needs what's different for them now.

If you can't write that sentence, the change probably isn't post-worthy. That's fine.

3. Never add a number you can't point to

This is the trap with AI-written posts. Ask any model to "write a tweet about what I shipped" and it will happily add "10x faster" or "users loved it". When I tested an early version of my own tool, it wrote "People kept dropping off…" about my project with zero data behind it.

Rule: a number goes in only if it's in your commits, your diff, or your own answer. "Export now takes one click" is fine. "Export is 3x faster" needs a measurement.

4. Adapt it to the place, not just the length

5. Tag every link, and count signups, not likes

Give each post its own link: yoursite.com/?ref=x-1009, ?ref=reddit-sideproject. Save the ref with each signup in your database. After 48 hours, the question isn't "how many likes?" but "how many signups came with this tag?"

Here is why this matters. My own project, three days in (Oct 10, 2026): 4 followers on X, 22 visitors to the page, and 0 signups from any post. The followers felt like progress. None of them signed up. Without the tags I'd have counted the followers as a win.

6. Don't kill a post type after one try

One post flopping proves very little. It could be the hour, the day, the length, the first line, the image. Decide on a type of post (a progress update with numbers, a technical story, a question to your users) and only drop it after it fails a few times at different hours. When something works, try it again at another time to see if it was the content or the timing.

7. What I got wrong

The boring part, automated

Steps 1 to 4 are the same every week, so I turned them into a free Claude Code skill: it reads your recent commits and diff, asks you one question (what changed for the user), then drafts one X post and one Reddit post. Only numbers it can find, and it never posts for you.

/plugin marketplace add thiguillaumat-art/lowie
/plugin install ship-post@lowie

Then type /ship-post in your repo. Source and plain-skill install: github.com/thiguillaumat-art/lowie (MIT).

Steps 5 and 6, the counting and the retesting, are what I'm building next: Lowie, a marketing hire for what you ship. It's a waitlist for now, with a public rule: 30 signups by Oct 22, or I shut it down. getlowie.vercel.app

Written by the person building Lowie, in public. Questions or corrections: hello.getlowie@gmail.com.