I've spent the last few months in the trenches building an email sending platform and along the way I picked up a few things that surprised me. Not about the code but about how poorly understood email deliverability is even among technical Founders who send email regularly.
We re a small group of 15-year-olds from Mumbai running Indian Instinct Studios. Every evening we used to open the fridge and ask kya banaun aaj? (What to make today?)
So we built TikkaMate an AI cooking assistant made for real Indian kitchens:
Fridge photo Jain thali in 10 seconds (no roots, no onion, no garlic)
I built something because I had 2 problems and couldn t find a good enough solution. Problem 1:
I code every day for work, and AI assistants make me much more productive.
But there s a tradeoff you need to actively practice and improve your coding skills or you forget.
Because let s be honest:
AI can write code faster than ever, but understanding it is still on you. Problem 2:
More and more often, I d have a quiet moment in the evening, pick up my phone and get stuck in a Shorts/Reels/whatever-your-platform loop.
Even if it s not hours, I really dislike that habit. So I started looking for quick, practical coding exercises I could do in those in-between moments. There are options, but none really fit: LeetCode great on a laptop, but not something I d replace my phone time with Mimo if you already know how to code, it feels painfully slow Other apps free trial then a subscription and still slow So I built it. Coding Drills Available on iOS!
https://apps.apple.com/lu/app/co... Soon on Android (waiting play store confirmation) In a nutshell: Short coding drills you can finish in minutes Bite-sized lessons that actually stick Multi-language: Python, SQL, JavaScript, Rust Choose your level: new, beginner, intermediate, advanced No subscription (3 challenges + 3 lessons free per day, restock with a simple ad) Give it a try and let me know what you think! To send your feedback or if you want to contribute with drills or lessons:
https://github.com/FedericoRaimo...
We're about to launch WriteMyPrompt a marketplace where you hire expert prompt engineers. Before we go live, I want a reality check. If you regularly use ChatGPT, Claude, or Midjourney for work and you're not getting the results you want, would you be open to paying $20 $100 to have an expert craft the perfect prompt for your use case? What would make you trust it or not?
I built a free tool that simplifies generating Access Tokens for Shopify custom apps.
If you've worked with Shopify's API since their early 2026 update, you know the process got significantly worse multiple screens, manual URL copying, and way too many steps for something that should be simple.
One pattern I keep seeing is that logging starts out useful, then gradually becomes inconsistent, noisy, expensive, and harder to trust. Field names drift, context goes missing, dashboards get polluted, and sometimes sensitive data ends up in places it never should have reached.
The deeper problem seems to be that most teams try to fix logging after the data has already entered downstream tools. By then, the cost, risk, and cleanup burden are already there.
I m curious how other teams handle this. What breaks first in practice: naming consistency, missing required context, sensitive fields in logs, alert noise, or ingestion cost? I m building in this area and want to learn where current approaches still fall short.
We're launching TeamAI today AI employees that get a real Ubuntu desktop, open Chrome, and actually complete tasks for you.
Not just answers. Real work.
The most surprising use case from our early users has been competitive research. They tell the agent "go through these 20 competitor websites and summarize their pricing" and come back in 20 minutes. No copy-paste, no prompt engineering, just done.
