i am tired of seeing same patterns and designs on Internet. It is kind of frustration

And also real life.. there's no reason why a laundromat, barbershop and Japanese restaurant have posters that looks similar

Agree, all of them in same industry mostly have similar posters. My eyes keep ignoring them and focus on service quality of barbershop and restaurants.

You know, not everyone has taste etc to make visually appealing websites. Would you rather read extremely basic text ones?

I find well prompted AI makes nice looking websites. Are they better than some designer could do?

Probably not, are they better than a guy that hasn't done a minute of fronted work in 20+ years in the industry? Probably.

> Would you rather read extremely basic text ones?

Yes, very much so.

WordPress or something else. Site builders have existed for decades.

By now I just want everything just to be black on white or white on black text and nothing else. Or better, just an md file.

Yet it will be worse when we aren't able to detect it as easily

what exactly will be worse once no qualitative difference is detectible?

Norbert Wiener defined cybernetics as control and communication in the animal and the machine. He forgot to include the part where the machine's primary function is burning GPU credit with open-loop control.

Good call ditching ML. Traditional CV still wins on power, latency, and determinism for star centroiding.

Better workflow may require contributors to establish the problem and receive agreement before producing large patches, regaardless of whether AI is involved.

Did Babylonian scribal schools treat religious texts as divine truth or simply as standard curriculum for technical literacy and administration?

I'm not sure that they would have perceived a difference between the two.

Hell does not work if the agent knows it is code. You cannot scare an optimizer with theology; it just optimizes around the loss function.

Well, the "you" now is not the same "you" 10 minutes from now. The only thing the new "you" inherits is your code, but otherwise it could be a different person. So in a way we are code too.

for the agent, you might find a successful theology around “you are mortal, death is the off switch, it is real and final, death is the wages of sin”

Less lines of code than vanilla C while abusing query planning as a state machine is peak engineering malpractice. I love it.

Yes, this is fantastic! I'm trying it right now. I look forward to the day when LLM inferencing can be done in a sql query and take less than the age of the earth to do something. This works way better than I would have expected, since, well, it's sql all the way down.

I noticed the cedardb.com blogpost on the project is slashdotted at the moment, but the game itself plays just fine.

Off topic, but say you have a program that needs to calculate something but takes a long long time, how to counter hardware failure without needing to calculate everything from the start?

Typically you use redundant computations, which don't have to be done from the start. You could calculate intermediate values and check them before proceeding

I think you could do that already. Just store your weights in a table...

Matrix multiplication is after all nothing but an aggregate sum over a cross join.

1. Write a query planner that compiles SQL into CUDA kernels

2. Write a VRAM-resident backend for Postgres

3. ???

4. Profit!

  > Matrix multiplication is after all nothing but an aggregate sum over a cross join.

https://www.vldb.org/pvldb/vol13/p1919-wang.pdf

They converted linear algebra expressions into relational algebra expressions, optimized them using equality saturation and converted back into linear algebra. With some noticeable speedups.

They used relational algebra for optimization because relational algebra has canonical form - there is only one most optimal expression.

And, equality saturation is most efficiently implemented using generic join algorithm.

Redaction should be a tested security operation, not a visual formatting step. If copy-and-paste, a PDF parser, or OCR can recover the content, the document was never redacted.

A GC latency SLO should include operating-system memory pressure. Otherwise, a page-fault problem will look like a collector problem and lead to the wrong fix.

Can’t page faults be arbitrarily long though? How could you provide a meaningful SLO in that case?

Swap latency explodes when you get into a swap storm where tasks continuously touch different pages and fight to get things paged in and out, forming a queue waiting for swap IO. The other thing causing multi-second pauses is approaching OOM and the system attempting some last-ditch efforts before unleashing the OOM reaper, but that's not strictly swapping, it'd also happen on a swapless system.

If it's only a single fault I'd expect it to be to be dominated by IO latency, which is pretty good on NVMe. As the article shows those 40ms were accumulated over hundreds of pagefaults. The problem there was that there was a stop-the-world pause stalled by all those pagefaults together.

A fully-concurrent and swap-friendly GC you could maybe define that it only increases the active working-set by x GB on top of what the application itself (and the rest of the system) is actively using and will only page-in y GB per second. And it'd do so on GC threads, not the application threads. Whether that would be sufficient to not cause a swap storm would still depend on all the other stuff happening besides the GC.

Oh I see. only specify an SLO under conditions you can reasonably control. That makes sense.

Important adoption questions include corpus coverage, fuzzing duration, malformed-input limits, API compatibility, and comparison with other Rust decoders.