Consuming tokens and calling it productivity is like judging a developer by lines of code written per day.
High activity. Low signal.
In fact, reducing the number of lines is often a better metric than writing more of them.
The win is not generating more words, more plans, more summaries, more "thinking"
I used to have nightmares about being stuck in an infinite loop while coding.
Now I have nightmares about prompts not working while the LLM confidently keeps doing the wrong thing.
Progress, I guess.
@rvivek@albinder@letsblinkit Any engineer hyping this as “great tech” should zoom out: for this job, it’s mostly table-stakes. The stack/tools have evolved so much that this is primarily a solid DevOps/ops execution problem, not some deep tech/algo breakthrough. Honestly, there’s more engineering in IRCTC.
TIL
Parrotfish poop = White Sand 🏝️
Civet poop = $100 Coffee ☕️
Whale gut waste = Chanel No. 5 🐳
Yeast poop = Alcohol 🍺
Tree poop = Oxygen 🌳
Conclusion: The Earth is basically a giant toilet and we are just paying a premium to live in it.
Just discovered https://t.co/cpKvCMmsU2, a brilliant initiative for IITB students. After 3-4 years of intense, isolating JEE prep, a program focused on psychological resilience, growth mindset & life skills is exactly the missing piece many of us wish we'd had
#IITB#collegelife
Optimus is going to be one of the biggest equalizers humanity has ever seen. Bc when every family can access a robot that delivers world class medical care, it means the quality of your health no longer depends on your income, your zip code, or your background. I get it now.
AIOps investments are exploding.
But what’s the ROI?
I discussed this with Forbes contributor Kolawole Samuel Adebayo.
The issue isn't adoption—it's attribution.
Companies deploy AIOps so performance improves. But they can't isolate what drove it.
The AI?
Better data?
New workflows?
87% say AIOps met expectation but only 12% achieved full deployment
Proving value is hard.
Value shows up two ways:
1. Cutting noise (lower alert fatigue)
2. Speeding fixes (better retention, lower churn)
But you need benchmarks: MTTD, MTTR, auto-triage rates, downtime avoided.
After last month's AWS outage, executives stopped asking if AI can run operations.
They started asking for proof.
That's the shift.
Promise to proof.
What metrics prove your tools work?
PS. You can find the link to the article in the comments
Expecting Indians to skip crackers for AQI is cute.
We eat sugar with diabetes, and drink on failing livers.
We're not self-preserving.. we’re self-destructing, but festive!
@OneTicketIndia wtf.
I tried booking the ticket and the payment went through, but instead of QR, it says "something went wrong", which happens to everyone.
But now I can't raise an issue because the transaction is undefined.
Unless you read this tweet, you will never even know!
@SiddharthS85 Religion gives you rituals.
Spirituality shows you the vanity of rituals and lets you find your own path to God, based on your own definition, and peace that’s yours alone.
Rust is toxic.
Not the community (mostly)... the compiler. 🦀
🔸 Borrow checker: “You can’t do that.” - but won’t tell you how to fix it.
🔸 You fight lifetime annotations like your life depends on it.
It ruins your today to save your future🦀
Go just doesn’t let Go.
🔸 nil != nil — because interface nil isn’t really nil (Surprise nulls!)
🔸 map[key] on missing keys → no error, just zero value. (Good luck debugging.)
🔸 Errors as values — forever writing if err != nil {} like a mantra.
Go away. 🚀
Java can keep you awake☕
🔸 Integer.valueOf(128) == Integer.valueOf(128) → False (Because… caching?)
🔸 Saying hello world needs a ceremony: public static void main(String[] args)
Java: Safe, boring… until it isn’t. ☕
You can't trust Python.
It's a snake 🐍 — it bites you in ways no other language does.
🔸 '5' * 4 → '5555' (Strings multiply?!)
🔸 True == 1 → True but True is 1 → False (Wait, what?!)
Python lets you do 'ha'*3 but laughs when [] is [] is False.
Trust the snake.. carefully. 🐍