Chinmay Nakwa

GenAI Engineer Intern @Tickernaut

[Chinmay Nakwa @primus2901]

Vision getting exorcised was NOT on my 2026 bingo card. 😭

[Chinmay Nakwa @primus2901]

https://t.co/yRN4HSLzPF

[Chinmay Nakwa @primus2901]

Ever read a book, looked up a word, and couldn't recall it a few days later?

I'm building Lingo to fix that. You save the words you meet, and the app helps you remember them.

[Chinmay Nakwa @primus2901]

1st page😶‍🌫️

[Chinmay Nakwa @primus2901]

Got a book just to get blinkit free delivery 😶‍🌫️

[Chinmay Nakwa @primus2901]

😇

[Claude @claudeai]

Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.

It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.

[Chinmay Nakwa @primus2901]

@LoopandPixels 👋

[Chinmay Nakwa @primus2901]

Couldn’t have found a better time to relate

[Chinmay Nakwa @primus2901]

Can’t wait for the Android version!!

[alphaXiv @askalphaxiv]

If doomscrolling X is part of your research workflow, we built something for you.

Introducing Paperscrolling 🚀

Get the most trending research with key ideas, figures, and audio explanations from alphaXiv Briefs

[Chinmay Nakwa @primus2901]

@MBashirzadeh Fair. Relevance-before-retrieval is basically a prediction problem: you're guessing what the task needs before you've seen the full task.That's why timing matters as much as filtering. Pull in less upfront, let the agent request more as steps unfold. Cuts stale-assumption risk.

[Chinmay Nakwa @primus2901]

Prompt engineering asks: "how do I phrase this?"
Context engineering asks: "what does the model even need to see?"
One's about wording. The other's about the entire information environment: system prompt, retrieved docs, tool outputs, conversation history.

[Chinmay Nakwa @primus2901]

If you want the full breakdown (prompt vs context engineering, context rot, the 4 pillars, live example): https://t.co/g5Yh3FpplZ

[Chinmay Nakwa @primus2901]

Bad: "Schedule an appointment for this patient."
Good: same prompt + clinic policies + doctor availability + patient prefs + live scheduling API output.Same model. Wildly different answer quality. That's context engineering.

[Chinmay Nakwa @primus2901]

Bad: "Schedule an appointment for this patient."
Good: same prompt + clinic policies + doctor availability + patient prefs + live scheduling API output.Same model. Wildly different answer quality. That's context engineering.

[Chinmay Nakwa @primus2901]

Counterintuitive finding: bigger context windows ≠ better performance.Dump in 4,000 pages of irrelevant Slack messages and the model struggles just like you would. Researchers call this "context rot."More isn't the goal. Relevant is.

[Andrew Ng @AndrewYNg]

New: A map of the most important skills in AI Engineering. https://t.co/VVkn1Dqp1N

[Chinmay Nakwa @primus2901]

Life is like riding a bicycle. To keep your balance you must keep moving.

~ Albert Einstein

[Chinmay Nakwa @primus2901]

Your REST API returns a user object with 40 fields. App needs 3. Multiply that waste across every screen and you're shipping megabytes nobody asked for.

GraphQL flips it: client sends a query describing the exact shape it wants, server returns exactly that. No more, no less.