11 Years in Tech | Backend | System Design | AI Building https://xitter.cf/t.co/YyyZZNpQyL DM For Collab & work📩
Joined February 2026
- Tweets16.2K
- Following813
- Followers3K
- Likes16.8K
Vikash Singh Tomar retweeted
Product exists -> 200 OK.
Product doesn’t exist -> 200 OK, body: null.
Frontend shows a blank page.
Monitoring says everything is healthy.
Nobody gets an alert.
Same endpoint. Same code.
Zero errors anywhere.
What’s wrong? 🤔
Product exists -> 200 OK.
Product doesn’t exist -> 200 OK, body: null.
Frontend shows a blank page.
Monitoring says everything is healthy.
Nobody gets an alert.
Same endpoint. Same code.
Zero errors anywhere.
What’s wrong? 🤔
Firefox for privacy and performance. Plus, it’s always good to have an alternative to Chromium-based browsers. 🚀
C first, then D. Talking to users helps validate the problem before spending time building the wrong solution. A quick prototype comes next.
Input validation isn't enough. Use parameterized queries or prepared statements so user input is never executed as SQL. Validation is an extra layer, not the primary defense.
I’d inspect deadlock logs to identify conflicting queries, ensure consistent lock ordering, optimize indexes, and keep transactions short. Retry deadlock failures with exponential backoff while preserving transaction atomicity.
Benchmark scores are just a starting point. Test both models on real production data and compare accuracy, latency, cost, and failure rates. The best model is the one that delivers better outcomes for your users, not just better scores.
A leaked secret should be treated as compromised, even after deletion. Revoke it, rotate it, and audit its usage. Assume exposure means compromise.
Vikash Singh Tomar retweeted
Everyone’s building AI features.
Almost nobody is testing them.
With normal code, a test passes or fails.
With AI:
-> Same input, different output.
-> “Works” depends on who’s reading it.
-> It’s fine in the demo, wrong in production.
If you can’t measure whether your AI feature got better or worse after a change, you’re not engineering it.
You’re hoping.
Everyone’s building AI features.
Almost nobody is testing them.
With normal code, a test passes or fails.
With AI:
-> Same input, different output.
-> “Works” depends on who’s reading it.
-> It’s fine in the demo, wrong in production.
If you can’t measure whether your AI feature got better or worse after a change, you’re not engineering it.
You’re hoping.
Every platform is slowly becoming the same product with a different logo. The real competition is for attention, not originality.