woooooo

12 ms·

151.
by woooooo 1y ago
Matrix vector multiplication for feed forward layers is most of the bandwidth as I understand things, there's not really a way to do it "better", its just a bunch of memory-bound dot products. (Posting this comment in hopes o
152.
by woooooo 1y ago
Yeah, but that is a continuum and not binary. Xi isnt personally in a room pulling levers that control everything. There are thousands of bureaucratic orgs and corporations making their own calls, deepseek and qwen were both trained on nvi
153.
by woooooo 1y ago
Im saying that people live in a culture, and that sets the parameters of what's conceivable. You're commenting from the post-reagan consensus, which inevitably leads to Intel's current situation. But it doesn't have to
154.
by woooooo 1y ago
It's worse than that, its culture war driven in a lot of cases. Doesn't matter if it's anywhere close to their back yard.
155.
by woooooo 1y ago
They've lifted a literal billion people out of third world, no-indoor-plumbing poverty in the last 30 years. And, to the point of the article, SMIC is already doing 5nm manufacturing. Political systems are more complex than dictator&#x
156.
by woooooo 1y ago
I think we could do with some different incentives. A lot of the readers here have worked at mega corps and seen what the incentives are in the real world, its not Ayn Rand superman doing the best thing for the company because their interes
157.
by woooooo 1y ago
Thats 3600 for a family of four. 5% of median household income in a country where most have zero savings.
158.
by woooooo 1y ago
Metaverse and this AI turnaround are characterized by the LACK of perseverance, though. They remind me of the time I bought a guitar and played it for three months.
159.
by woooooo 1y ago
A new hardware/software rollout is one of the more common breakage situations, though. It definitely could have been on purpose but my gamble is still on a fuckup with a new system rollout.
160.
by woooooo 1y ago
That's actually interesting! My experience is different, especially compared to the late 90s and early 00s, most people avoid Oracle if they can. But yes, its always worth having someone who's job is to think about the database
161.
by woooooo 1y ago
Im actually doing a ton of research in the area myself, the caution was against becoming an Nvidia expert narrowly rather than a general low level programmer with Nvidia skills included.
162.
by woooooo 1y ago
It's a useful bit of caution to remember transferrable fundamentals, I remember when Oracle wizards were in high demand.
163.
by woooooo 1y ago
Hanlon's and Occam's razors point to it being a mistake by the GFW operators, imo. If it's on purpose, I think you have the most likely motivation.
164.
by woooooo 1y ago
Fair enough, I actually only did that once or twice.
165.
by woooooo 1y ago
> trying to shame them into not doing it is pointless. Speak for yourself, its at least satisfying. Insult their choice of music and provide suggestions for what they should play instead.
166.
by woooooo 1y ago
Engineering leader | ex-something ... Is that person more likely to be a leader or a follower and ass-kisser in your experience?
167.
by woooooo 1y ago
We don't have a formal model of how/why any given LLM works, and incidentally we're also short on proofs for real-world software and organizations. Empirical facts are the strongest thing we have in this domain.
168.
by woooooo 1y ago
Great irony that his moment of maximum leverage with DOGE was spent slashing basic research.
169.
by woooooo 1y ago
They've got home offices full of MBA/finance people. It runs itself.
170.
by woooooo 1y ago
I think there's a distinction between inventing new terms for utilitarian purposes vs ideological and in-group signalling purposes. If you have groups talking about "expected value" or "dot products", that's di
171.
by woooooo 1y ago
"Controls dozens of companies" is a pretty strong allegation without naming a single one. "Has a stake in hundreds" is true of anyone who owns a global index fund.
172.
by woooooo 1y ago
Ok, but if we have to go looking for it, then its not exactly a juggernaut.
173.
by woooooo 1y ago
> "Build more" is direct and actionable. Regulatory reform is only one dimension. I could get behind "build more" much more easily than "abundance". You're onto something there IMO.
174.
by woooooo 1y ago
For what it's worth back, a lot of the opposition towards "establishment libs" is based more on optics than policy IMO. Voters feel ripped off by the establishment. Running on "we're the establishment and we'r
175.
by woooooo 1y ago
It's standard among self-described leftists, to distinguish themselves from establishment liberals. I'm not even very far left and I identify with the term because party leadership are all 80yo and out of touch.
176.
by woooooo 1y ago
Its about the message. Centering the message on something that has indirect, timelagged effects and isnt even actionable at the federal level is terrible messaging strategy for the national party. "The dairy industry ran ads saying mi
177.
by woooooo 1y ago
> Keep in mind: Klein and Thompson's political project is a plan to organize the Democrats. This is the crux of the opposition. It's not that leftists necessarily have a problem with zoning reform, I don't at least, its f
178.
by woooooo 1y ago
I think "innovativeness" is massively overrated compared to network effects and consolidation. Spotify is European. Any innovative SV companies going to unseat them with sheer pluckiness? Same goes for Meta or Amazon going the o
179.
by woooooo 1y ago
The 80s and 90s devs who built our current software infra were, on average, FAR less credentialed than today's juniors and mids who mostly don't understand what they're building on.
180.
by woooooo 1y ago
2 months but you're right, it wasn't "immediately". Bad memory on my part. https://en.m.wikipedia.org/wiki/2025_Gaza_war_ceasefire I still maintain that Netanyahu would gladly sacrifice hostages t

More ›