a-dub

14 ms·

271.
by a-dub 1y ago
also most python usecases that are in the realm of things like high performance concurrent request servicing push it down into libraries that i think are often tied to a native network request processing core. (gunicorn, grpc, etc) python i
272.
by a-dub 1y ago
i don't see it like a typical technological race for dominance. user count seems silly beyond bragging rights. i think it's more like a multiparty multipath expedition where the results of each team reaching the top is yet another
273.
by a-dub 1y ago
maybe it was wince. it has been a very long time! the ps1 had caetla (programmable with a parallel port and cheat cartridge) and yaroze (official sony devkits for hobbyists), but i think that was on some custom sony os services library in r
274.
by a-dub 1y ago
iirc it had a browser and used linux under the hood for the os. i think they even had a downloadable unlocked linux distro for it that couldn't play games.
275.
by a-dub 1y ago
i suppose one nice property is that it is trivially scalable. if the problem gets really bad and the scrapers have llms embedded in them to solve captchas, the difficulty could be cranked up and the lifetime could be cranked down. it would
276.
by a-dub 1y ago
blame canada
277.
by a-dub 1y ago
i wish they had a more exhaustive list of all the features they offer beyond vscode or cursor, otherwise it's hard to evaluate what exactly it brings to the table.
278.
by a-dub 1y ago
the one thing i wish tmux supported was remote connections to several backend instances.
279.
by a-dub 1y ago
i thought all the cool data driven robotics stuff was like reinforcement learning from sensors that track moving effectors in the real world with online retraining that mimics the sensorimotor experimentation that is observed during the dev
280.
by a-dub 1y ago
would be interesting to see if there's a difference in how the 0-page is handled in systems under this condition...
281.
by a-dub 1y ago
idk. i don't really follow the argument. large projects in open source coordinate internally and engage externally when they need to- i suspect that isn't all that different from what you'd see in a large megacompany like app
282.
by a-dub 1y ago
A* = priority queue with heuristic function _with specific properties_ ... in particular it must be an "admissible" hueristic that never overestimates the true value.
283.
by a-dub 1y ago
pretty cool! in isolation looks awesome! i'm still a little curious about the impacts increased executable image size, especially in a complete system. if all the binaries are big, does it start to crowd out cache space? does static li
284.
by a-dub 1y ago
a lot of early ai was simply applied classical data structures and algorithms. although perceptrons go back decades and decades.
285.
by a-dub 1y ago
traditionally helps with thundering herd, but counterintuitively can help with just a single client. if the action potentially has an unknown duration long tail in terms of state or load on the service, backoff can help find that duration a
286.
by a-dub 1y ago
hmm.. i wonder how it would compare then with clang+linux, clang+stl or hotspot+j2ee. reminds me a bit of the days when perl programs would often outrun native c/c++ for common tasks because ultimately they had the most efficient strin
287.
by a-dub 1y ago
i'd be curious if they're using the same llvm codegen (with the same optimization) backend for the c and rust versions. if so, where the speedups are coming from? (ie, is it some kind of rust auto-simd thing, did they use the oppo
288.
by a-dub 1y ago
harry potter is likely to be excerpted a million times all over the web (in legitimate fair use context). wouldn't it make more sense to try out other titles that are still under copyright, appear in the research datasets, but have lit
289.
by a-dub 1y ago
seems to me that one approach is fueled by data and the other is fueled by understanding. in the former, the observations form a view of behavior which is then modeled with high fidelity. in the latter, active inquiry, adversarial data coll
290.
by a-dub 1y ago
"When working on your particular problem, start with simple comparisons and then fit more and more complicated models until you have what you want." sounds algorithmic...
291.
by a-dub 2y ago
it would be interesting to perturb the CoT context window in ways that change the sequences but preserve the meaning mid-inference. so if you deterministically replay an inference session n times on a single question, and each time in the m
292.
by a-dub 2y ago
yeah, i understand the premise. my question revolves around how well it actually works today for bi style applications. specifically, how close is it to being something that you can just drop in as a smart query and plotting interface rath
293.
by a-dub 2y ago
is that true? i'd like that, but i get the sense that this mcp stuff is more oriented around programming assistant and agent applications. i suppose the desktop app can use it, but how good is it for this general purpose "chat wi
294.
by a-dub 2y ago
it's pretty standard these days i think. i remember aws having some hyperlocal low latency POPs with graphics capabilities for this specific purpose. i think there have also been a handful of purpose built remote desktop packages that
295.
by a-dub 2y ago
i wonder if an llm could have spotted the malicious patches to autotools in the dist tarball... a public, deterministic and assured build facility for oss would be cool. also maybe a deprecation of autotools.
296.
by a-dub 2y ago
llm commit scanning might be an interesting approach to the oss supply chain security problem.
297.
by a-dub 2y ago
it was fun when having a network solutions/internic contact handle was a badge of honor. the early internet was fun. whois was always a fun dimension. is there a canonical rdap client that will end up everywhere? one of the nice things
298.
by a-dub 2y ago
precision != accuracy, correlation != accuracy let f be a continuous real valued function, let c be a fixed bias constant, let pcc be the pearson's correlation coefficient function. pcc(f) = pcc(f + c) f + c is precise, but not accurat
299.
by a-dub 2y ago
> LLMs are not good in some domains and bad in others. Rather, they are incredibly good at some specific tasks and bad at other tasks. Even if both tasks are in the same domain, even if tasks A and B are very similar, even if any human t
300.
by a-dub 2y ago
didn't early versions of netscape have a wysiwyg html editor? i think the bigger challenge was highly available hosting. there were authoring tools aplenty.

More ›