He says he needs 200,000 GB300s or Huawei Ascend 950s. Now, the GB300 is much faster. So, probably he's just paying a little lip service to the 950 because of political pressure and would prefer the former. It's a moot point though because Hauwei can't manufacture with a low enough defect rate currently.
So glad that I'm not the only one struggling with these huge generated PRs that are too big to honestly review, all while an AI reassuringly whispers in my ear "just trust me."
Don't get me wrong, overall I really like having AI in my workflow and have gotten many benefits. But even when I ask it to check its own work by writing test cases to prove that properties A, B and C hold, I just end up with thousands more lines of unit and integration tests that then take even more time to analyze -- like, what exactly is being tested here?, are the properties these tests purport to prove even the properties that I care about and asked the agent for in the first place, etc.
I have tried (with at least modest success) to use a second or third agent to review the work of the original coding agent(s), but my general finding has been that there is no substitute for actual human understanding from a legitimate domain expert.
Part of my work involves silicon design, which requires a lot of precision and complex timing issues, and I'll add that the best AI success I've had in those cases is a test-first approach (TDD), where I hand write a boatload of testbenches (that's what we call functional tests in chip design land), then coach my various agents to write the Verilog until my `make test` runs with no errors.
I'm not affiliated with this company, but I personally believe that not having a website (or any public presence) is the future of our industry. The issue is that any information you put out there can now be used by AI to emulate you. This can be used by competitors to quickly reach feature parity with your product, it can be used by bad actors like scammers, and just generally is harmful to you.
I know it sounds silly but not having a website is probably the future of the web.
It's already happening. Klarna canceled their IPO. So did StubHub. I think it's a safe bet that all planned IPOs have been canceled for the time being.
That is true. X=150 under the 1977 International Emergency Economic Powers Act, which is the authority under which Trump is establishing the tariffs. After 150 days, someone could file a lawsuit to reverse the tariffs if there has not been any vote.
"None of these trade provisions empowers Mr. Trump to impose tariffs on all imports from all countries based on an arbitrary formula. Section 122 lets a President impose tariffs of up to 15% in response to trade deficits, but Congress must approve them after 150 days. Someone should sue to block his abuse of power."
Actually, that's not what I've found. When I have compared betting sites like Kalshi and Polymarket to the implied probability of Fed rate cuts in the interest rate futures market, I have found them to match up and move in lock step with each other. (I don't work for any of these companies and I don't use the betting sites myself; just stating my observation.)
The belief of Trump and his allies is that the last 40 years of free trade has been extremely favorable to college educated "elites" and extremely unfavorable to the working class. They see foreign trade as having hollowed out American manufacturing, and therefore they see the shutting down foreign trade as a way to reverse this decline. While they are aware that doing so will hurt college educated professionals, they don't care because they feel those people have had it too good for too long.
Not going to comment on the politics, but I'd point out that recession odds on Kalshi are currently 64% and 58% on Polymarket [1,2]. In other words, recession in 2025 is more likely than not -- though far from guaranteed.
My thought relevant to HN is that either way this is likely to decimate venture capital and startups, at least in the short run. Venture capital funds come from limited partners like pensions and endowments, and the professionals I know are all shifting away from growth stocks and into value, international and bonds. VC is already in a drought (except for AI) and now my guess is that it's going to get worse. So many VCs I know have started only in the last 15 years, so they have never seen a real recession like 2009 while working in venture.
Yeah, it's a Briticism. The FT is basically the Wall Street Journal of London, for those not familiar. I've been reading this paper for decades and they often use "group" for "company," "sack" instead of "fire," and many other delightfully English locutions.
I don't have any statistical data on this, but my impression is that it's more than "some people." It may be half or even most.
You have one contingent that is anti-Trump and will vote for any alternative to Trump, even a senile old man with dementia. You have another contingent that is against Progressivism/leftism and will vote for anything that opposes this, up to and including voting for Trump despite strongly disliking him.
The root problem is that social media amplifies extreme voices, so you get very extreme rhetoric coming out of both sides. This scares people and makes them feel like their primary goal must be to vote against the scary thing.
That's a tragic story. However, I'm surprised that the transistor was supposed to come in a DIP package. Usually through-hole discrete transistors come in a three-lead package like TO-92. Of course, that would not have helped you since yours looked like every other student's except the for the markings.
I don't think it's just adoption and network effects, though that is part of the equation. The other big (bigger?) piece is that the CUDA landscape is very complete, with libraries and examples for many different kinds of use cases, and they are well documented and easy to get started with. Ctrl+F this page for "ecosystem" and you'll find another comment that explains it better than I can.
OpenCL was definitely supposed to be the standard, but as the sibling suggests, it's just much harder to use than CUDA. Also, it doesn't feel like OpenCL has much of a community around it, documentation doesn't seem great, and just the overall experience is very frustrating. I tried to implement something in OpenCL about 5 years ago, thinking it would be fairly trivial to port a simple compute shader from CUDA, and ended up giving up. Just too difficult.
I've been telling people for years that NVIDIA is actually a software company, but nobody ever listens. My argument is that their silicon is nothing special and could easily be replicated by others, and therefore their real value is in their driver+CUDA layer.
(Maybe "nothing special" is a little bit strong, but as a chip designer I've never seen the actual NVIDIA chips as all that much of a moat. What makes it hard to find alternatives to NVIDIA is their driver and CUDA stack.)
UX. Cursor is fundamentally no different from going to ChatGPT and asking it to write some code, but the Cursor UX is very nice. Shows you colored diffs of what it wants to change with accept/reject, chat window to flesh out ideas, etc. They get all the small UX details right (assuming you like a VS Code style experience).
If your boss brings up "orange jumpsuits" in any context, you should probably start looking for a new job. Even if it's to tell you that you won't be wearing one, the fact that the person thinks this way is very sus.