A week ago OpenPilot, the open-source desktop agent I built because I wanted a Cursor-style tool that runs on my own model, had 1 GitHub star. I launched it on Product Hunt on October 9 and wrote a couple of posts here. This is a short, honest look at what actually happened, mostly so I remember it and partly in case it's useful to someone else launching a small project alone.
The numbers
All of these come straight from GitHub's API and the Product Hunt page, nothing rounded up.
| Oct 9 (before launch) | Oct 15 | |
|---|---|---|
| GitHub stars | 1 | 18 |
| Forks | 0 | 4 |
| Windows installer + portable downloads (all releases, all time) | 170 | 423 |
| macOS downloads (dmg + zip) | 6 | 14 |
| Open issues | 0 | 1 |
Product Hunt: it finished launch day at #9 with about 88 upvotes and 54 followers. On launch day itself there were zero comments from other people, and one thoughtful question came in the next day.
What I learned
Upvotes are not installs. For most of launch day the Product Hunt page was getting attention but almost nobody was downloading. The "Visit website" button sends people to the landing page, and from there it's another click or two to GitHub Releases. When I put a direct Windows download link (plus a note about the unsigned-app warning) in the maker comment, downloads started moving. That's the single change I'd make earlier next time.
The unsigned installer costs people. Windows SmartScreen and macOS both warn on unsigned apps. Explaining it up front helps, but I'm sure some people bounce. Signing is on the list.
I'd rather write tutorials than announcements. An "it's live" post says nothing new after the first day. A step-by-step setup post, like the one on running it locally with Ollama and Qwen 3, answers the question people actually ask, which is "will this work with the model I already run?"
Slow and steady is still something. Downloads jumped in the first few days after launch and have settled to roughly 15 to 20 new installs a day. Not a rocket, but real people are trying it.
What's next
- A proper LICENSE file in the repo (it says MIT in the README, but GitHub doesn't detect it yet, which blocks some lists).
- Testing more local models and writing up what works and what breaks.
- Smaller fixes from the first issue reports.
A note on privacy since it comes up: OpenPilot currently sends anonymous usage data by default (app opens, feature use, token counts, no chats, files, paths or keys). You can turn it off during setup or in Settings → About.
If you've tried it with a model that didn't behave, I'd genuinely like to know which one. Repo is here.
I used an AI assistant to help draft and edit this post. The numbers are from GitHub and Product Hunt.
Top comments (9)
Hey OpenPilot, I read your post with the real numbers from your OpenPilot launch week. I'm a solo founder too and stuck on the same problem with my own tool, so I'm curious: where did those first ones come from, if you know?
Honestly, mostly Product Hunt at first. Launch day brought plenty of clicks, but installs only really moved once I put a direct Windows download link in the maker comment. Since then it's been a slow, steady trickle, and I can't say exactly where each person comes from because GitHub release counts don't show referrers. The DEV tutorials seem to bring fewer people, but ones who actually go through the setup. Biggest lesson for me so far: make the download one click away wherever people land. Good luck with your tool!
That’s a really useful insight. The direct download link is a simple change, but it clearly made a difference. Did you try adding it only after launch, or was there a reason you hadn’t included it from the start?
Added it the same afternoon, once I saw launch-day upvotes weren't turning into installs. There was no good reason it wasn't there from the start. I just assumed the Product Hunt "Visit" button to the website was enough, and it wasn't, since every extra click loses people. If I launched again I'd put the direct download in the first comment from minute one.
That makes sense. It's easy to overlook how much friction one extra click adds, especially on launch day. I'm realizing the same applies to getting users to try a tool: even when people are interested, the next step needs to be obvious. Did installs pick up noticeably after you added the direct link?
Yes, clearly. Launch day had about 22 new downloads in total, and roughly 16 of those came in the few hours after the link went up. Before that it was a handful all morning despite the upvotes. It's slower now that launch traffic is gone, under ten a day, but that one change is the thing I'd never skip again.
That’s a pretty convincing signal. 16 downloads shortly after adding the link is hard to ignore. It’s a great reminder that reducing friction can matter as much as getting attention in the first place. Thanks for sharing the actual numbers!
Thanks for sharing raw numbers instead of rounded ones. The first week after a solo launch is mostly about learning which channel actually brings people who try the thing, not just star it. Did the Product Hunt traffic convert to installs or issues at a different rate than readers coming from your DEV posts?
Good question. From what I can tell so far, Product Hunt brought a lot of curious clicks, but installs stayed flat until I put a direct Windows download link in the maker comment — then downloads started moving. Most of the steady ~15–20 installs a day since then don't look like launch-day PH-only traffic.
DEV readers seem fewer but more likely to actually try the setup steps (especially the Ollama one). I haven't seen a clear DEV-origin GitHub issue yet though — there's only one open issue so far, and that came after the PH launch. Small sample, so take it with salt.