Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
tux3
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
9 ms
·
181.
▲
by
tux3
11mo ago
So with the water used in datacenters. It's just a cooling loop, the output is hot water.
182.
▲
by
tux3
11mo ago
The original Lenna is controversial, but I'm delighted to share the "ethically sourced Lenna": https://mortenhannemose.github.io/lena/
183.
▲
by
tux3
11mo ago
>I think cargo is better than cmake That is an understatement. I can't think of a build system that has spawned more attempts to find something better than CMake has. There have been so many people trying to make their own C/C+
184.
▲
by
tux3
11mo ago
That has definitely not been my experience. I like Gitlab, but they've had regular incidents all along. If a git push failed I wouldn't question it, it's almost never my network. I'd just open Gitlab's Gitlab and fi
185.
▲
by
tux3
11mo ago
In this case AlphaEvolve doesn't write proofs, it uses the LLM to write Python code (or any language, really) that produces some numerical inputs to a problem. They just try out the inputs on the problem they care about. If the code gi
186.
▲
by
tux3
1y ago
Point taken, I might have formed the wrong impression at the time. With my C++ background, I'm very much at home with that philosophy, but I think there is room for nuance in how strictly orthodox we are. C++ does have optional languag
187.
▲
by
tux3
1y ago
I feel like select!() is a good case study because the common future timeout use-case maps pretty closely to a select!(), so there is only so much room to weaken it. The ways I can think of for making select!() safer all involve runtime che
188.
▲
by
tux3
1y ago
I think the kernel and microcontroller use-case has been overstated. A few bare metal projects use stackless coroutines (technically resumable functions) for concurrency, but it has turned out to be a much smaller use-case than anticipated.
189.
▲
by
tux3
1y ago
There was a proposal for a unit type for C++, it would be a zero-sized type just like void except that you could actually use it to declare a variable and pass around as a value, like other regular types. The less serious proposal: have it
190.
▲
by
tux3
1y ago
Quantum advantage papers have a history of overpromising, this one looks interesting, but it would still seem wise to wait for a second opinion.
191.
▲
by
tux3
1y ago
Most modern glues are synthetic. Super glue is usually cyanoacrylate, regular glue sticks might be Polyvinyl Acetate (PVA). Outside of specialty glues that are intentionally going for old school methods, it's not really a thing anymore
192.
▲
by
tux3
1y ago
Very nice to see an update! I have to admit the pushback on ET_STAT is not completely unexpected. I can see why the gABI thread would argue that it can be solved without a new e_type. It would be a major change that requires updating all to
193.
▲
by
tux3
1y ago
Same experience here. Both consumed a lot of CPU, but they looked very similar on the perf graph, and neither missed any frames.
194.
▲
by
tux3
1y ago
The effect of claiming that people act like NSA propagandists is indistinguishable from claiming they are an NSA propagandist, except that the wording allows you to weasel out of it. This turns a thread about cryptography into a thread abou
195.
▲
by
tux3
1y ago
Matthew Garrett is not a secret NSA propagandist. People can reasonably disagree with the djb position. His blog posts are notoriously divisive, and that doesn't make everyone on the other side a secret NSA influencer. Please assume go
196.
▲
by
tux3
1y ago
>RISC-V-Zxxx implementation can be as incompatible with any other RISC-V implementations (and even specification) as one wants. It can't. Extensions have a lot of freedom, but still have to follow the core spec. You can't just
197.
▲
by
tux3
1y ago
>So why is it fundamentally and obviously true for PQC? This isn't actually an engineering hill I'd die on, if more people I trust made clear arguments for why this is dangerous I'd take it very seriously, but right now we
198.
▲
by
tux3
1y ago
I know some users held back on the upgrade while Windows 10 still had support, but there's many improvements to privacy & consent in Windows 7 that are worth considering, now that both are about to be equally out of their support w
199.
▲
by
tux3
1y ago
I've seen Twitter use "AuDHD" for the intersection. It's big enough to have its own label and subgroup who identify with it.
200.
▲
by
tux3
1y ago
The side-effects are vaguely evocative of antipsychotics or some sort of antidepressant. There's no specific autism medication that I'm aware of, but psychiatric diseases often have plenty of comorbidity. There's some ADHD po
201.
▲
by
tux3
1y ago
Very funny to call Chrome an electron app. For the sake of words meaning things, it isn't one. Althought Electron uses the same core components as Chrome, the Chrome UI is not built with Electron.
202.
▲
by
tux3
1y ago
Having a niche and expanding into new industries are all fine, there is no problem with having a DB filling a particular sub-segment of the market. But writing that traditional SQL databases cannot go above these "100-1000 TPS" nu
203.
▲
by
tux3
1y ago
>Traditional SQL databases hold locks across the network; under Amdahl's Law, even modest contention caps write throughput at ≈100–1,000 TPS In fact large real world systems are not limited to 100-1000 TPS, or even 10 kTPS as the ca
204.
▲
by
tux3
1y ago
The RISC-V privileged spec describes their paging implementation (under the Supervisor-level ISA), while the unprivileged spec has a chapter and two appendices describing their memory model (RVWMO), formal axiomatic and operational models i
205.
▲
by
tux3
1y ago
Pssh, real developers read volumes 1, 3A, 3B, 3C, and 3D of the Intel® 64 and IA-32 Architectures Software Developer’s Manual Combined Volumes, pages 3125 through 4200 And if you have a steady hand, a magnetized needle and a hard drive can
206.
▲
by
tux3
1y ago
That's great given that the goal of OAI is to train artificial superintelligence first, hoping that the previous version of the AI will help us control the bigger AI. If GPT-5 is learning to fight and undo other models, we're in f
207.
▲
by
tux3
1y ago
They seem to be making their own eFPGA architecture, so there's no vendor tool to compare against for their primary purpose. The closest you'll get is xc7 (Xilinx) vs vendor, one of which is surely Xilinx 7. But the Yosys xc7 supp
208.
▲
by
tux3
1y ago
If you need to retrieve the values, the fastest would probably be a stored generated column, not a virtual computed on the fly. In case you only want to filter without returning values, you could also index directly on the expression withou
209.
▲
by
tux3
1y ago
I lesson I have yet to learn (along with everyone on the Fox go server, presumably)
210.
▲
by
tux3
1y ago
It hasn't, Gateway is still in development with most features being experimental.
More ›