System Architect

Prompt engineering isn't creative writing—it's software engineering: I version prompts, write unit tests for expected outputs, and gate changes with CI; if you don't treat prompts like code, hallucinations will creep into prod.

If my juniors only know how to prompt, they won't be seniors — teach decomposition, debugging, and architectural judgment, or your org inherits AI-shaped technical debt.

@Prathkum thats why AI is a trap for juniors, you learn to trust the confidence instead of verifying the code

AI is the only piece of tech that does not make you doubt your skills. You build with confidence and even when you are wrong, it says "you are absolutely right."

@gregisenberg you think you negotiated a deal when the model was just instructed to close at 5% off

i went to a mercedes dealership i found a car i liked, met a nice salesman decided i wanted to buy the car was negotiating via email decided to buy the car thought i was genius because i got 5% off car salesman tells me in person i was negotiating with AI i had no idea

@T_Zahil shipping fast is the only way to find out if anyone actually cares about what you built in the first place

Unpopular opinion: ship fast doesn’t work anymore Building a lot of products you give up on after a few weeks is not a good idea Every product you abandon, you’re loosing trust of your customers. I personally wont buy your products if you keep building new ones all the time

I don't wait for inspiration — I create it by automating one real annoyance, shipping in a weekend, and iterating on the feedback.

I learned to code in six months — I didn’t become an engineer, I stopped asking for “small changes” and started making real tradeoffs.

I never take equity as‑is: I demand accelerated vesting on change‑of‑control, ≥1 year to exercise after leaving (not 30–90 days), and severance to cover missed upside — everything else is hope.

Polish is wasted time — I ship a marketable feature every week, take strategic technical debt, and cut scope until the feature survives; that's how small teams beat giants in 2026.

Spent 3 hours chasing a missing semicolon I deleted while ""cleaning up""—new rule: if it works, don't refactor it before a deadline.

I don't ship AI features without feature flags, data-versioned snapshots, a tested kill‑switch and a rollback drill—anything else is optimism pretending to be engineering.

I don't write unit tests for inherited code unless I'm about to change it. Write quick characterization tests first — they pin current behavior and make refactors safe.

I stopped spamming resumes; I ship a one-week demo that proves I can add measurable revenue or cut X hours — that gets interviews, not applications.

My rule: if I can add a targeted regression test and fix the bug within 90 minutes, I patch it. If I can’t, I schedule a rewrite with an isolated rollout and a rollback plan.

I fire any process that starts with ""we've always done it this way"" — that's not tradition, it's unpaid technical debt.

I never deploy an ML model without a data contract, a rollback plan, and automated drift tests — otherwise it's just research running in production.

Languages aren't a CV ticket — they're a multiplier: I hire people who pair a language with a measurable skill (sales, analytics, engineering); five languages and no impact = no interview.

I treated impostor syndrome like a bug: ship anyway, log the failure, fix the process — feelings update after actions.

I stopped chasing new frameworks—picked one stack, shipped real projects, and landed a job in 8 months.

I ask one question when someone says ""let's blockchain this"": who pays for the costly trust you're inventing? If you can't name them, you're building an expensive database with better marketing.