tucnak

21 ms·

391.
by tucnak 2y ago
I own Keychron devices with QMK firmware, and Brave (Chromium) is really explicitly communicating whenever the keyboard is being configured. I have also installed GrapheneOS using it, & honestly I cannot imagine getting tricked into lau
392.
by tucnak 2y ago
I was under the impression that the purpose of "system" prompt is to encode the instruction boundary explicitly to reduce the risk of injection. Do you enforce some kind of security invariant that we could rely on? For example, do
393.
by tucnak 2y ago
I agree with your comment in principle, but also disagree in this instance. I'd also recommend Futhark even though I'm not using it in production based on positive experience I've had with it previously. I happened to own an
394.
by tucnak 2y ago
You should learn more about FPGA's!
395.
by tucnak 2y ago
> My personal opinion so far is that this whole thing is teetering between media frenzy and actual issue. Welcome to hybrid warfare! This is the point.
396.
by tucnak 2y ago
Surely there's more to economics than controlling who you get to marry brother!
397.
by tucnak 2y ago
> It takes a decent amount of coordination, equipment, and crew to pull off these missions. They're dragging anchors on the marine floor; all coordination, equipment and crew are already there as a medium-sized fishing vessel is per
398.
by tucnak 2y ago
Check out mailcow
399.
by tucnak 2y ago
> I just find your position that btrees are optimised for every query structure But that is not my position! Postgres has many index access methods: hash, btree, brin, gin, gist, and there are extensions for rum, bloom, skipscans, geospa
400.
by tucnak 2y ago
Much appreciated, & good luck with your project!
401.
by tucnak 2y ago
It's simple, really. Make it so the experts have 0 leverage. Maybe have "the workers" make all the decisions! ??? Profit? :-) They tried this in Soviet Union...
402.
by tucnak 2y ago
Everything is a monoculture if you squint hard enough.
403.
by tucnak 2y ago
> Firefox has better memory usage At the cost of a subpar cache; it's not like Chromium is leaking memory, & its memory pressure effects are both well-studied and well-understood. Yet, Firefox stans keep touting lack of comprehe
404.
by tucnak 2y ago
So the underlying storage is conventional, it's still tuples of some kind, and it's only a matter of how indexes are laid out? Otherwise, I'm struggling to see how it could "optimise" for certain access patterns. Ho
405.
by tucnak 2y ago
Completion latency is one thing, bandwidth would be another. There's apparently a whole world of Alveo SmartNIC's and related FPGA platforms, and it can totally get in nanosecond range for whatever nails that may fit the compute-i
406.
by tucnak 2y ago
I wouldn't be surprised if it's some stock AWS instance type, probably not even NVMe-enabled, or some kind of metal instance bottlenecked by EBS. The downside of Cloud push in SaaS is that they were unable to tap into the ever-cha
407.
by tucnak 2y ago
How is it different? Isn't a graph basically two sets of tuples: edges and nodes? I played with Cayley (Google) for a little while, & that was my impression.
408.
by tucnak 2y ago
To accomplish what exactly? The mirror would be similarly limited.
409.
by tucnak 2y ago
Many ISP's provide /56 or at least /64 these days, but at any rate you can always get some from cloud providers and use Wireguard to tunnel the rest... There really isn't much excuse for not supporting IPv6 at homelab sc
410.
by tucnak 2y ago
If you're homelab is proper, you likely own a /56 range, also known as 256x /64 which is what they're limiting. I've always known my prefix would come in handy! Now, I only need to work out how to make it work witho
411.
by tucnak 2y ago
He works for 127.0.0.1 you should look them up!
412.
by tucnak 2y ago
The API is compatible, and even if it weren't, it wouldn't matter anyway; everybody has been writing OpenAI API-compatible proxies, including Google, for months now. The only thing that matters is availability, throughput, cost pe
413.
by tucnak 2y ago
> DeepSeek is open source and is on par with o1-pro (is it?) There is no being "on par" in this space. Model providers are still mostly optimising for a handful of benchmarks / goals, like we can already see that Grok 3 is
414.
by tucnak 2y ago
I don't know if they'd changed the submission title or what, but it says quite explicitly "Deepseek R1 Distill 8B Q40" which is a far-cry from "Deepseek R1" which would be misrepresenting the result, indeed. Ho
415.
by tucnak 2y ago
Of course you cannot use these machines "for confidential workflows for enterprise use", at least with AWS you know whose computer you're working with, but also keep in mind that it's really hard to steal your data as lo
416.
by tucnak 2y ago
Nobody ever made that claim
417.
by tucnak 2y ago
> can be tricky > extremely important > this article showcases and so on...
418.
by tucnak 2y ago
Bad, bad AI.
419.
by tucnak 2y ago
Mistral exists
420.
by tucnak 2y ago
Just search for "System 1 / System 2" and add word "researcher" there, filter by time <1y< and you'll find plenty. OP is right that a bunch of people made far-going predictions, & now doing mental gym

More ›