Likely because the last 75 years of government policies have focused on calorie maximization and not optimization. The structural setup needs to change.
I’m not joking—every time I’ve traveled to India with my wife, who’s from Taiwan, her first comment during a meal is always, “Why is the daily diet so low on protein, especially meat?”
@hainbehen Not hopeless, however, AI hasn't matured to that level yet (primarily limited by the inherent unreliability of language models).
If one can come up with reliable agents built without much knowledge of data science, then such a course can be useful.
Happy to discuss further.
@prathgodbole William Dalrymple doesn't portray Marathas as anything beyond raiders. It is stunning to me how difficult it is to find honest history texts.
@aadilbrar India will never be what China is. For the simple reason that the political systems are extremely different. India cannot and will not do the 10% growth numbers china pulled off. But 7% consistently? Sure.
For decades, a select group of organizations has published reports that shape public perception of India, Indian immigrants, and Hindus. Today, I'm launching a new project that scores these reports on their methodology. Read my twitter article and 🧵
https://t.co/jrZFDbCHZ2
Field Marshal Asim Munir is desperately trying to build his legacy as a successful leader- both domestically & internationally.
Deteriorating security. Diplomatic cold shoulders from UAE, Saudi, Iran, Afghanistan. When reality doesn't cooperate, narrative becomes survival.
🧵
@JaidevJamwal If you're building muscle, you can avoid cardio and you wont get exhausted. Plus, the hormone rush gets you more energized for the rest of the day.
In case you still feel exhausted: very likely you have electrolyte loss. Take an ORS and carry on with the day.
@rasmus1610 And consider the low enjoyment because you aren't in a flow state. It's still useful because it takes away writing the "obvious" code. Just not sure how much one can trust the reasoning abilities.
It feels wrong to ascribe to my name, so I'll call it Khattab's Law (@lateinteraction ): "Any sufficiently complicated AI system contains an ad hoc, informally-specified, bug-ridden implementation of half of DSPy."
Ran a small eval today on an LM using GPT-5.2 as a judge. Model scores 10%, but paper reports it scoring 34%.
I see that the paper uses GPT-5.1 as a judge; for the sake of consistency I change it.
Switch to GPT-5.1 as a judge. Model now scores 43.5%...
bro