Materials discovery might be the hardest thing you can ask AI to do today. Frontier models with lab tools and 100M+ tokens can "discover" hundreds of novel compounds that hit every target โ but ask how you'd actually make one and it falls apart: hallucinated recipes, gaming the scorer, and eventually clocking out.
This also happens to be THE most important problem to solve. Compute scaled 100x and so did the heat you have to pull out of the chip, making thermodynamics the real ceiling on AI - not transistors, not capital. We need materials that don't exist yet.
Finding them takes nuanced "taste" in materials and the ability to arm AI with that taste. That's exactly what we saw in @Akash__Ramdas and @advaith_sridhar when we decided to partner with them!
It's amazing to see insti (@iitmadras) teams go after some of the most ambitious problem statements out there :)
cc: @LightspeedIndia@MohapatraHemant
@nyn Not sure if exchange has to be quick. Feel there is motivation in the 2 user sets to enable exchange. Prep cld be made easier by guided navigation (like property apps for listing). Recipes have stages & could guide user thru diff stages with clips/images that stitch together.
@nyn@sriramk@cnu@VennInDoubt@srikanthch10@srikrishnang@twitortat@manikandanux Isn't video a better mode? It's tough to collate (though I doubt how easy writing out recipes is for the older gen) but there's benefits that it brings about:
1. Learning visually is easy, esp in a skilled craft
2. Furthers the nostalgia element of seeing older gen "in person"
@pspratyush Giving him the contract so early last season was such a stupid decision. Ed literally could've waited till end of season and no one would've questioned him. I seriously doubt if Ole is tactically good enough to manage a top tier team
@IndiGo6E we're a bunch of passengers 50+ stranded on bangkok airport due to cancelation of the inbound flight. We are getting no information on what's happening and how we get back to Bangalore. Flight number 6E76