Hacker Newsnew | past | comments | ask | show | jobs | submit | novok's commentslogin

I've heard the harder part is to have data centers to put the nvidia hardware than getting the hardware currently.

And the secret ingridient to building those faster than others is... crime.

https://www.theguardian.com/technology/2026/jan/15/elon-musk...


Holy shit:

> This article was amended on 16 January 2026 because a megawatt is a unit of power, not energy as an earlier version suggested.


Micropayments would be another one, but then governments and banks have to give up ~~financial control & surveillance~~ AML essentially to make it financially viable. AML also has a horrible track record of how much money is spent compared to the amount recovered.

Anti-money laundering laws are a deterrent. If you know moving around $10,000 will be reported to FinCEN, as will any discovered pattern of structuring transactions below $10,000, then you are forced to pursue riskier ways of moving your money than Western Union.

Wouldn't micropayments be aggregated into a money laundering operation by some third party? (Wasn't IIRC even Spotify used for that?) Or would Cloudflare take all the money in this hypothetical scenario?

Security is often an excuse to block other teams to do legitimate work and so often it's fairly braindead. Security IMO needs to get it's act together, passkeys is a great example of security gone wrong from a UX design perspective because you can't hold them to the same standards as product or infra teams, they have the special privilege of breaking things and it increasing their metrics.

Tell them to make a better UX and they lose their minds in a huffy puff of fake crisis mode or get avoidant with stonewalling 'secret security stuff' that you can't hold them to account for. Or eat 50% of developer machine performance for "endpoint security" and the carnival of sadness goes on and on.

Signal is an example of security as a product that was actually designed for user UX in mind to give one example.


Yes we do, continue keeping it a faux pas to reduce the over verbose LLM speak put elsewhere and ask people to just share the original prompt and save us all time. Label your LLM usage to respect other people's time.


If you've used UIKit to any large degree, how bad SwiftUI is, almost a decade later, continuously punches you in the face. About 20% more time to write initially, for 90% less bugs and potholes, and I've tested this with iOS devs who never used UIKit before too. Now with AI, there is even less of an excuse IMO.

Leave SwiftUI to the settings pages. The gulf between AppKit and SwiftUI in macos desktop I'm not as sure about.


Also many have power even cheaper or even free unused surplus power with solar.

I don't do local inference other than hobby & learning reasons because electricity is so expensive where I am at.


I have an M1 Pro, and a M4 & M5 max to play with at work and the speed difference is very significant between all 3 machines, the M1 Pro is far slower, and the M5 is significantly faster than the M4. And a windows 3090 beats all of them but eats twice the amount of power per token. This is all running the same 24GB memory friendly model with LM studio.


I played with classifying and summarizing my entire email history (per email) with small models, but that only took about 12h of GPU time at most. Using a coding agent cli wrapper in that case is far slower because of all the spin up cost and the system prompt they inject even if you want to turn it all off.

If I used an actual direct API it probably would've been much faster, but I'm doing it for hobby / fun reasons. You also get to fiddle with a lot more params.


whats funny is if it looked up the source code on github it would've figured it out faster


He is implying he doubts the accuracy of that 'match'.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: