That’s a great point! One thing that I found challenging on this was that it was so tough to get good data, so I don’t feel confident that doing things differently would matter. That’s part of what motivated our second Show HN experiment, but different approach did not change results.
Some things surprised me, and some of the common advice didn’t quite play out the way I expected.
If you're planning a launch on either platform soon, I hope this is helpful.
blog.nseldeib.com/p/what-i-lea...
After our Product Hunt launch this week, a few founder friends reached out asking if we had any learnings or best practices to share.
I wrote up an honest (and somewhat long, sorry) reflection on our recent launches to Show HN and Product Hunt, including the prep work and what actually happened.
This seems like the good chaos!
Shorter feedback loops when building software with agents is crucial! How to shorten and speed up iteration is an obsession.
Image of CodeYam CLI & Memory Product Hunt launch page.
We’ve been running it on our own repo for the past few weeks and are seeing fewer repeated mistakes and much less manual context maintenance.
Free, runs locally, no login required.
We launched it on Product Hunt today. Would love feedback if anyone here tries it.
CodeYam Memory runs a background agent that reviews coding sessions, identifies confusion patterns, and generates targeted rules with proper scoping. You review and approve everything.
While digging into this we discovered Claude actually has a native rules system that lets you target context to specific parts of your repo with path matching. Nice primitive, but managing those rules by hand was not sustainable for us.
So we built CodeYam Memory.
Over the past few months we’ve been using Claude Code heavily for software development.
One issue we kept running into was memory.
Claude Code would repeat the same mistakes and our claude [dot] md files would slowly get stale.
CodeYam CLI & Memory's a small-but-mighty free tool we built after running into repeated issues while coding with agents: context gets messy, mistakes repeat, & memory quickly goes stale.
Would love for folks to check it out over your morning coffee or tea tomorrow ☕ Appreciate any support!
Product Hunt profile screenshot: Nadia Eldeib. builder and trier of new products. 220 followers. 18 following. 17 points. 2 day streak.
Are you on Product Hunt?
After years of lurking, I’m finally launching my first product there tomorrow.
If you’re on Product Hunt, feel free to follow me at “nseldeib” and I’ll happily return the favor.
Tomorrow morning we’re launching CodeYam CLI & Memory for Claude Code.
If you're building with Claude Code, we just released a free local tool for memory management to help keep context organized during long coding agent runs.
Would love feedback.
If you're on HN, search “codeyam.”
Made my first Show HN post today.
Woke up early with the nervous/excited butterflies that come with shipping something new.
Posted it… and within minutes watched launch after launch push it down the feed. Such is life on Hacker News.
Still really proud of what we built.
Small reminder: sometimes nothing is missing. You just need a different way of searching.
Lost an AirTag weeks ago. Find My said it was at my climbing gym but I had no clue where.
Asked lost & found yesterday mostly on a whim. Someone suggested pinging it. Immediate chirping from the bin.
Same day I recovered sunglasses from a ski jacket pocket I’d already washed.
Thank you 🙏 the excitement and scaries are real!
Trying to pick the line that lets us keep building.
If you’re using Claude Code day to day and want early access, feel free to reach out. Would genuinely love thoughtful feedback.
Right now at CodeYam we’re putting our product in front of our first external users. Exciting and honestly a little terrifying. We’re starting small on purpose, working closely with people who feel the same Claude Code context and memory pain we built this for.
What stuck with me was how strong the instinct is to keep going simply because you already started.
Building a startup feels like this a lot.
So I stopped, took my board off, and hiked back uphill until I could traverse onto a different line.
Still steep. Still hard. Just not catastrophic.
The run narrowed, got steeper, and instead of the chute I expected… there was a rocky cliff below that I definitely did not want to meet.
First run of the day, dropped into a snowboard trail that looked completely reasonable from the top. Fresh snow, great visibility, heading toward a part of the mountain I remembered loving last year.
A few turns in, I realized I’d misjudged the route...
Photos from Sunday while enjoying some well earned views and strudel at Bonnie's after said (mis)adventure.
Bonnie's delicious apple strudel.
Preface: I’m okay.
Had a small mountain (mis)adventure this weekend that felt very startup-adjacent.
One thing I keep relearning as a founder is that communicating a complex product can feel almost as hard as building it.
We have been using AI not just to write code, but to pressure test how we explain CodeYam.
Our product designer Dani wrote more about this. Link in the comments.
If you’re building with AI and thinking about where autonomy ends and tooling begins, this piece by my cofounder @jaredcosulich.bsky.social is worth a read.
👉 blog.codeyam.com/p/to-tool-or...
One framing that’s stuck with me: tool use isn’t a sign of weaker intelligence. It’s often the opposite. Complex problems benefit from externalization.
I often hear some version of: “Why can’t AI just do X?”
Where X is a software development task like fixing a bug, writing a test, or making a clean code change.
Usually the answer is simple: the task is more complex than it looks, for both the human responsible and the AI doing the work.
This looks great! Sharing with the Harvard in Tech London community.
P.S. Feedback on our new site welcome. I'm still iterating!
