HN Explorer

Ask HN: We got 100 stars. A weekend project got 12k. What are we missing?

2 points by lilwing ↗ HN
This week Graphify launched off the back of a Karpathy tweet. 48 hours to build and 1k stars. in 2 days. It uses an llm to INFER system structure literally labeling its own edges as "inferred"

we launched Ix a week ago. its an open source CLI that maps any codebase into a persistent architectural graph so ai agents stop relearning your system every session. 100 stars in 10 days. Graphify got 12k in 48 hours.

when a tool is built in 48 hours anf outpaces months of engineering in star velocity is that a signal about the product itself or the timing?? does the better technical approach win eventually or is the early momentum most important.

I think we have the better solution but also aware i might just be wrong about what the market wants right now. can anyone look through our repo and tell us?

would be super helpful.ill put the link in comments

7 comments

[ 7.1 ms ] story [ 35.8 ms ] thread
BeetleB
1. Karpathy is famous. You're not.

2. How many of those stars are bots?

sgbeal
> What are we missing?

To re-frame the question: why do you care how many stars (presumably you're talking about github stars?) you have? Does the number of stars change what you develop or how you develop it?

Losing sleep over github star count is akin to losing sleep over up/downvotes on HackerNews or thumbs on Your Favorite Social Media Site. Tying one's self-worth to them, or one's self-image of one's own works to them, is... well, kinda sad.

> i might just be wrong about what the market wants right now.

Github stars are no indication of "what the market wants" - they're an indication of how many people (or scripted bots) have seen the project, thought "huh, interesting," and clicked the star so that they have a bookmark of it for later reference in their github settings.

sph
People still care about Github stars? News to me. I used to star repos as a reminder to check them out at a later date, and always forgot to. Only inexperienced devs on GH care about these useless and gameable vanity metrics.
Madsn
I was deciding between Ix and Graphify today, and had Sonnet (free tier) compare them for my use case. Won't paste the full comparison, but here is the reasoning Sonnet gave for recommending Graphify over Ix:

  Ix is more polished and purpose-built for agent integration, with a Claude Code plugin that hooks into every file read and grep automatically. But it's a paid product in early access, less proven, and the Claude Code plugin is Claude-specific — the AGENTS.md integration works with OpenCode but is less deep than the hook-based integration.
For my use case I think I would be perfectly fine on the free tier so the comment about it being paid might not be valid. It also seems false that Ix is early access / less proven compared to Graphify. I do think it's interesting that Sonnet made these apparently incorrect conclusions based on your website and whatever else it found online.

In my case I went with graphify because I am using OpenCode, and it seemed plausible that Sonnet would at least have gotten that part right about Graphify having better support for OpenCode. Now that I take a second look though, I realize there is a plugin for adding Ix into OpenCode (https://github.com/ix-infrastructure/ix-opencode-plugin) so now I'm contemplating migrating to Ix - the hooks available in that plugin appear to be better than what I have now with Graphify. I really dislike that graphify hooks rebuild the graph post-commit, since I want to include the graph files in my git repo.