90% of backend engineers are just people afraid of CSS.

LLMs speed up coding to the point where it doesn't feel like you were the one who created it. Because of that, the satisfaction at the end of the day is not the same. It begins to feel like the code isn't really yours and that feeling stays with you.

your biggest cheerleaders are often strangers, rarely the people close to you

Use AI as your co-pilot, never let it fly the plane alone.

Disagreements without data are just opinions.

Is CQRS good? Absolutely. Should you implement it? Depends😂

Ek din woh kisi aur ko commit kar degi, aur tum bas Git commit karte reh jaoge😂

The best engineers in the world are self-taught. Not self-taught in the sense of having no education, but in the sense that they built things nobody asked them to build.

50 engineers who can't communicate will build more complexity than a product actually needs. Because everyone needs something to do. 20 engineers who trust each other will build something that actually works.

No offense, but if a manager says 'fix this bug' and you do it without asking why, impact, or alternatives, you didn't engineer anything — you just executed a task. Real engineering starts with a business problem, not a Jira ticket.

Most people will understand things only after it breaks. By then, options disappear. Thinking early is the real edge.

Engineering starts on pen and paper, but it is validated in the IDE and in production.

It's always important to set context.

Ironically, this whole idea of role models, especially on social media, has turned into a flood of so-called professional role models and influencers, which is mostly hollow. People should look beyond idolising individuals and instead pick traits they admire from different people. Excessive, almost cult-like admiration is harmful, and social media has amplified the wrong kinds of role models for millions.

Optimize cloud costs where they are materially significant. Otherwise, the most expensive line item becomes engineer time, not infrastructure.

It's complex, but that doesn't mean it's complicated.

Poor engineering education has a hidden cost — when a university neglects core engineering, students don't graduate as engineers, they graduate knowing they must relearn engineering on their own.

Skills ≠ Opportunities

Downdetector was down too

So can we engineer these little butterflies today that flap their wings a decade later and help you succeed as a business, as a creator, as an engineer or as a hobbyist? Absolutely not. You can't engineer success like that. But you can change. You can create tiny positive shifts, knowing that each small change carries a probability of producing some impact, maybe small, maybe orders of magnitude bigger, at some point in the future. That's really the gist of it. You can't predict the future.

Building a small RAG pipeline is easy. Building a production-grade RAG with low latency, high availability, and consistent retrieval quality at scale is a nightmare.

In the age of LLMs, the smart are evolving faster and the dumb are proudly doubling down.

Most scaling issues come from the database, not from tiny code tweaks like data structures or function calls. When the database is the bottleneck, other optimisations don't really matter.

absolutely nothing is absolute, and everything is a trade-off

Scaling works best when it starts with good code. Only after improving the app should you scale hardware.

When it comes to this world, always quantify your time in terms of money.