- 👾 Founding engineer @ formly.ai
- 🤓 Founder of weirdfriends, skintheory
- 🎸 Post punk musician spotify, bandcamp
- 🇩🇪 Berlin Based
v1.3.0 - fancy new corrections, points, and stats 📈
Copy of an email update from mailing list: weirdfriends Hello gang, I spent my 2 days off basically just coding. The wife isn’t the happiest about it, but: WE’VE GOT SOME COOL NEW THINGS! 🤩 Fancy New Corrections & Points 🔧 I changed up the agent architecture and now have a dedicated (and fancier 💅) model/prompt for better corrections of your messages. There’s now a grammarly-inspired UI where it highlights issues in your messages and you can click on them to learn more! Also, you will get points for perfect messages in a language! 💯 Example of a correction ...
congrats on failing to launch; weirdfriends log 1
7 months, 221 commits, and I still haven’t told anyone to try my app. A reflection on what went wrong.
My personal blog's traffic is 95% AI crawlers this week
Happened to check my analytics this week and saw this. Admittedly I don’t have much real traffic. I haven’t been writing much recently, but this irks me. And its super rude. Not only is all my traffic fake, but its actively stealing my personality and voice. Potentially even cataloging me across my virtual profile. Like this is nothing new on the internet, but this is kind of egregious. I’ll probably just fully block Singapore and Chinese IPs. It doesn’t sit great with me honestly, but I’d rather not be someone’s training data. ...
How not to make AI character
Intro: I’m almost finished with my language learning app weirdfriends and just thought I’d share some thoughts on what not to do for generating your ai characters to chat with. Don’t make an agent Don’t provide a random seed Don’t create the characters by hand Don’t use structured outputs Don’t put all your dynamic vars first
Reviewing Others' AI PRs
When you give me a PR to review, it should be complete and tested by YOU. You should be proud and ready for your work to be shipped right now. I’m just here to double check things. I should not be: the first set of HUMAN eyes to look at your AI slop making your code DRY/maintainable because the AI didn’t cleaning up the style/formatting to fit with the rest of our codebase because the AI didn’t The Bottom Line ...
My Next Project
I’m optimizing my next product for: An eye popping headline (clicks) As much conversations with users as humanly possible (pmf) Extra credit: it’s launched in a community interested in the topic, so the release post doesn’t feel like an advertisement
Unfiltered Thoughts On A Failed Product Release
I want to cover 2 things, before I get caught up in thought; My feelings releasing modergator today and getting no new users What to do next with modergator My feelings releasing a project and getting no users It sucks. And its basically the worst outcome. And its what I expected to happen. Our marketing plan was an after thought: Make a post on Reddit Make a post on indie hackers Make a post on twitter And the results, of course, are nothing. Maybe 10 people came to our site. ...
The Chrome Debugger
You should use the chrome debugger if you do any frontend js/ts work. /thread console.log is dead. The debugger saves my ass every time. You don’t need to set up anything or even import it if your using Chrome. You just: Add debugger; anywhere in your code: (stick it behind an if if you like!) If you’re debugging on the fly, find your file in the sources tab and click “Add Breakpoint”. Now you can really debug all those god damn rerenders. ...
Why I Make Waitlists For Side Projects
1. External validation for a new idea At the start of a project, I’m not sure if only I have the problem I want to solve. The more emails that join my wait list, the more excited people are about solving the problem. I try to be super straight up about who my product is for on the landing page. I also try to scope down the communicated project features as much as possible. ...
My 3hr Landing Page Build Strategy in 2025
Today I built a wait list landing page for a new project of mine. I did it in around 2 hr 55 min. This is a PB for me. I could definitely have done it faster, but with the trade offs I wanted, I am quite happy. The Strategy Note: I’m not affiliated with the brands mentioned. 1. Audit current state of drag-and-drop website builders Time: 45 min | Cost: $0 ...