Think like an economist Talk like an accountant Act like a technologist @dineshsingh2004 That a decade ago I think burned into my brain somehow.

The vibes couldn’t be more immaculate in America right now

WE’RE GOING TO AN ELLA LANGLEY CONCERT ON THE 18TH AND WE’RE EVEN GOING TO MEET HER!!!!

Welcome to y’all Street.

We have a new Toqueville.

Painful comments section. Let me spell it out. $LITE and $AVGO have superior laser design and fabrication process. They can deliver same noise performance and power with much shorter cavity length. Everyone else has to make lasers 60% longer, obliterating InP area and yield.

Babe, wake up. Freddy is looking at rifles at Bass Pro Shop

According to international law, the land on which any Buc-ee's stands is legally part of the state of Texas. Kinda like an embassy or whatnot. Therefore there is no such thing as a Buc-ee's located outside the state of Texas

DUDE LMAO THIS IS A GAS STATION😭😭😭

Chili’s & NBA Finals😋

This dude can sense I expressed interest in an IPO can’t he.

Interesting approach from Apple They are storing the shared attention block in the DRAM While the FFN weights stay in NAND and are loaded in the DRAM, depending on the request Apple is facing 3 constraints - 1) Limited DRAM size 2) Large model size (20B params) 3) Slow NAND read speed A super small model (sub 8B) won't be that useful, but they can't store a 20B model in DRAM (due to memory shortage). They also have to manage the KV cache overhead. If they streamed the weights completely through iPhone SSD, then it would take 2.5 seconds to generate just 1 token (0.4 tokens/s) So the big thing here is that a normal MoE activates different experts based on every token, but in Apple's case, a sparse mask predictor decides which parameters to activate based on the request/prompt, locks it in, and loads it into the DRAM (1B-4B depending on the request). They basically convert a 20B MoE (with 1B-4B active) into a dense 1B-4B param model for a request. The tradeoff: They are basically adding 0.3-1.5 seconds (1B to 4B params loaded) of latency to TTFT time by loading FFN weights from NAND to SSD per request (read speed is around 1.5-1.7 GB/s for iPhones) and taking a hit to performance They will get around 15-50 tokens/s of decode speed (depending on params loaded) Ideally, smartphones would come with 24-32 GB of RAM so that 20B param models could be loaded, but memory shortage won't allow it to happen But, their competitor here is ChatGPT Instant, which is a much smarter model that runs at 200 tokens/s and has a TTFT of 0.8 seconds (Apple's TTFT will be around 0.5-2 seconds, and decode speed is around 15-50 tokens/s), and is also free Apple's AFM on device models will be great for privacy-focused tasks. They get beaten by cloud models on other benchmarks (perf, speed, quality)

now that i can't eat beef i've decided to pivot into preachy environmentalism for example, did you know that the water used to make 1 cheeseburger could power ONE HUNDRED THOUSAND Opus 4.8-Max prompts? wasting that water on meat is completely unjustifiable