shreya

Software Engineer | Backend & AI | Reading books, shipping code, learning every day.

[shreya @eshryaa]

The fastest database query is the one you never make.

[shreya @eshryaa]

Indexes is helpful to find data without scanning every row in a database but too many can drastically slow writes.

[shreya @eshryaa]

Connecting to a database is way too expensive...
thats why connection pool are important to reuse existing connections which improves performance and stability of the system

[shreya @eshryaa]

When one request touches 10+ microservices, logs alone won't tell the full story.

Distributed tracing follows the request with a single Trace ID across every service, revealing where latency or failures occur which make debugging distributed systems far less painful.

[shreya @eshryaa]

Agentic loop is just what we call trial and error

[shreya @eshryaa]

@rxhit05 Yeah right, plus too many tokens get exhausted fast in claude for same task compared to codex

[shreya @eshryaa]

@realmcore_ Its working!

[shreya @eshryaa]

Coding has becomes effectively free and now the scarce resources are curiosity, judgment, and trust.

That's a much harder problem to solve.

[shreya @eshryaa]

@harsh_dwivedi7 The startup may be winding down, but helping thousands of people find opportunities is a legacy most products never get to claim. That's worth being proud of.

[shreya @eshryaa]

@the_codewala Yes now writing code has become easy but logic behind code is still important

[shreya @eshryaa]

@shinemeriz This is so true

[shreya @eshryaa]

Vector databases are just proof that computers also forget where they put things.

[shreya @eshryaa]

@thenowhereway I'll probably explore ideas that I'd normally abandon too early...

[shreya @eshryaa]

@maharshii def, gotta put your all into it

[shreya @eshryaa]

The fastest algorithm is asking, "Do we really need this feature?"

[shreya @eshryaa]

Learned about request coalescing.

If 10,000 users request the same uncached item at once, don't let all 10,000 hit the database.

Let 1 request fetch the data while the other 9,999 wait.

Same result. 9,999 fewer expensive queries.

[shreya @eshryaa]

@neerajjj6785 Good taste and trust + money to buy tokens :)

[shreya @eshryaa]

@prasenx true, maybe they will extend it for some more time...

[shreya @eshryaa]

@rxhit05 Yeah basically same

[shreya @eshryaa]

We save hours of coding using AI, then spend those hours debugging it