वनमाली गदी शार्ङ्गी शङ्खी चक्री च नन्दकी श्रीमान् नारायणो विष्णुर्वासुदेवोऽभिरक्षतु।

Partially agree, but this is a loaded blog trying to say that you need all the bells and whistles from microsoft for the frontier to really kick ass. Partially correct

Elon is the real Tony stark. The one line which I took away from his talk track was “Lets take fiction out of Science Fiction”. Kudos Elon !

I have realized generally in shark tank India all judges are idiots but one who caught my attention was @1kunalbahl, he is calm, very respectful and his questions and analysis is spot on. Very graceful ! So much respect for you Kunal.

Complete market manipulation seems to be going on, all kind of outlandish predictions, $28 trillion made up market cap, for a stock that will will be twice as valuable as Walmart while generating less revenue than Macy's. I am big fan and operate in that industry too, but its just crazy how every second post on the time line is about the IPO.

This entire platform is flooded with spacex IPO

Interesting, journalists turn off comments. This is new type of journalism where take some random deported ass and interview them and then turn kf the comments.

This whole CBSE OSM issue seems to be an inside hit job. It should be properly investigated. After NEET this doesnt seem to be just an accident but reeks of a systematic sabotage aimed at instigating the youth between 15 to 21 of age. @PMOIndia must get involved and do a deep rapid assessment. Selection of a dubious edtech company, creation of a minified SPA with embedded master password at the client side, so called hacking, continued disastrous public takes by CBSE representatives, flawed algorithmic mixup in osm scanned sheets and now accelerated media narrative. Now coupled with another spicy buildup of IAF delivery of NEET papers. All of it is designed well to discredit govt. and create a distrust in the system.

Apprehend these non ethical miscreants. This is not defensive pentest, its criminal.

The implication nobody's talking about: We're entering the "physics era" of AI architecture. In the early days, anything went. Throw parameters at it, try a new trick, see what sticks. That's chemistry, mix things and observe. Physics is when you realize there are fundamental laws. You don't discover new physics by trying harder. You discover it by understanding constraints.

I think I know why. There's only one answer to a specific set of constraints: Inference must be cheap at long context (kills pure Transformer) Retrieval must be precise enough for reasoning (kills pure Mamba) Training must run on existing GPUs (kills exotic architectures) When you stack those constraints, the solution space collapses to one point.

Three AI labs told completely different stories about their new models. NVIDIA said theirs was "the future of state-space models." Alibaba called theirs "a breakthrough in hybrid attention." An academic paper described "an elegant memory mechanism." I read all three papers side by side last week. They built the exact same thing. A thread on why this matters (1/n)