gorkish

CRS are L3 switches; they are supposed to route and run dynamic routing protocols and do ECMP and all the good stuff so long as you dont configure features like nat and firewall that require the cpu and cant be offloaded.

The one thing that it frustratingly does not support that I wish it did is stateless bgp with ipv4 prefixes. It cant offload an ipv4 packet with an ipv6 next hop.

I think you are right and most people here are overthinking it.

If you make a bunch of identical API calls quickly, before the initial request completes its cache write, you will have cache misses.

If you make a single API call to warm the cache, then make a bunch of subsequent identical requests, you will have cache hits.

The optimization for this sort of thing is to deduplicate in-flight requests, but i've seen no evidence that is being done; the better design pattern would be an api or tool call that explicitly launches multiple agents after a shared prompt processing stage

There's no single call that can launch multiple event loops at the same time, so there really isn't a notion of serial or parallel; it's always going to be a simple question of how the first model calls made by each hit the backend. I highly doubt that a bunch of agents created within milliseconds of each other are going to get cache hits. It's more likely they will spin up in parallel. I strongly suspect if Anthropic were trying to deduplicate their inflight requests for cache efficiency, they'd at least want to tell you about it.

I'm gonna jump in here and say that the way Anthropic's KV cache works (prefix matching) doesnt guarantee this at all if the subagents are all put into flight at close to the same time. There's nothing in the agent SDK that allows you to instance more than one agent simultaneously, and the docs don't say anything special about KV cache use by agents in this scenario.

You could probably engineer your own harness or workflow to warm the cache by running a single warmup agent, but this is not at all how the tools behave out of the box. It's mostly a cost optimizaiton, so Anthropic really don't have any incentive to do it proactively, and if they did, surely they would want to tell their customers how much they were saving?

I'm betting the person who created the "artist rendering" isnt an aeronautical engineer.

There is essentially zero chemical difference whatsoever in sugar vs corn syrup coke. sucrose disassociates in the presence of an acid into glucose+fructose simple sugars. Just being carbonated will disassociate the sucrose.

Just a cynical observation here, but its funny how the author still hangs onto the notion that it is "the best" despite that it de facto cannot be "the best."

Also, maybe dont rely on a poorly maintained app for making secure connections to your systems? Just me?

It is kinda weird how all the chat type tools all spit out such cheeseball language despite that the raw models almost never generate stuff like that. It honestly just makes me feel like the people making this shit are so caught up in their own hype bubble that it doesn't even register to them as abhorrent.

woah hold on a sec. that's not how these clocks are actually used though.

It's a huge huge huge misconception that you can just plunk down an "atomic clock", discipline an NTP server with it and get perfect wallclock time out of it forever. That is just not how it works. Two hydrogen masers sitting next to each other will drift. Two globally distributed networks of hydrogen masers will drift. They cannot NOT drift. The universe just be that way.

UTC is by definition a consensus; there is no clock in the entire world that one could say is exactly tracking it.

Google probably has the gear and the global distribution that they could probably keep pretty close over 30-60 days, but they are assuredly not trying to keep their own independent time standard. Their goal is to keep events correlated on their own network, and for that they just need good internal distribution and consensus, and they are at the point where doing that internally makes sense. But this is the same problem on any size network.

Honestly for just NTP, I've never really seen evidence that anything better than a good GPS disciplined TCXO even matters. The reason they offer these oscillators in such devices is because they usually do additional duties like running PtP or distributing a local 10mhz reference where their specific performance characteristics are more useful. Rubidium, for instance, is very stable at short timescales but has awful long term stability.

Yeah I really don't understand why they went this direction as it builds considerable additional complexity directly into the application to solve a problem with an external component

I would have probably approached this by implementing a fix for the misbehaving part of k8s, though since there isnt a default LoadBalancer in k8s, I can't really can't speculate further as to the root cause of the initial problem. But most CNI or cloud providers that implement LB do have a way to take feedback from an external metric. I'd be curious why doing it this way wasn't considered, at least.

RISC is dead; long live RISC

I had a look and this project is most definitely not open source. Nobody should be making that claim.

Changing the HN headline to soften the project's own bombastic claim isn't really the point of the argument. It would bhe better to highlight any project that is actually working on open source microscopy of which there are many. Flagged

Web search often tanks the quality of MY output these days too. Context clogging seems a reasonable description of what I experience when I try to use the normal web.

No, unsafe-yt

So what if your adversary relays your encrypted message along another channel?

Set llm-auto-lobotomy to disabled in chrome:flags. I think this is only available in canary at the moment

Unfortunately being noble or self righteous or whatever emotion you choose has nothing to do with it. If there is a pool of grant money available only to “Focused Research Organizations,” and you want some of it for your work, then you open one and do your work under that umbrella. Academic institutions themselves do this all the time. It looks politically and morally sketchy, and maybe it often is, but it’s the way it works.

All this is just the result of playing a zero sum game with industry that is traditionally very low margin. The money always comes from somewhere.

What's the use case for the 10GbE? Is ~200MB/sec not enough?

Yes; its far from close enough. My litmus test for consumers is, can they put their video files on network storage and work with them? And with modern cellphone video, the answer is no it is not convenient to do that. Consumer networking is in a tarpit. Business grade systems have networking that is three orders of magnitude faster; there is no other component between systems that are that dissimilar in performance, not even GPUs.

10GbaseT and SFP+ DAC are a world apart. The latter is now extremely power efficient. But even modern 10GbaseT parts are considerably improved

I'm not really sure why this is being characterized as an emerging threat. VPN-over-DNS and all kinds of DNS based exfiltration were quite popular a couple of decades ago. Still, most people don't really secure their DNS properly even still.

It seems someone is just trying to drum up business with this; it's not a new thing.

The lack of highspeed networking on any small system is completely and totally insane. I have come to hate 2.5gbps for the hard stall it has caused on consumer networking with such a passion that it is difficult to convey. You ship a system with USB5 on the front and your networking offering is 3.5 orders of magnitude slower? What good is the cloud if you have to drink it through a straw?

NVMe NAS is completely and totally pointless with such crap connectivity.

What in the WORLD is preventing these systems from getting at least 10gbps interfaces? I have been waiting for years and years and years and years and the only thing on the market for small systems with good networking is weird stuff that you have to email Qotom to order direct from China and _ONE_ system from Minisforum.

I'm beginning to think there is some sort of conspiracy to not allow anything smaller than a full size ATX desktop to have anything faster than 2.5gbps NICs. (10gbps nics that plug into NVMe slots are not the solution.)

If you contextualize Star Trek's "isolinear chips" to be something like this, they start to seem considerably more sensible.

Has anyone built a physical ASIC that embeds a full model yet?

In this case it was their logo mark. AFAIK that's the only claim that they actually made as well. Reading anything else into this story is ridiculous.

To me this whole situation seems to be a pretty good example of what not to do if faced with a copyright claim. It's tempting to see outfits like The Onion or Cards Against Humanity doing this kind of thing and feel like you are bulletproof and can do it too. But it's pretty easy to tell when these 'open letter' responses have not received legal review.

Waffle House had a legitimate claim and acted on their obligation to their trademark.

There is a pretty good argument that Waffle House continues to have a claim.

They probably wont do anything because its, frankly, a waste of time.

This could probably have been avoided entirely if copyrights were correctly respected from the beginning.

It's not, per-se, a driver doing the PD negotiation in software; it's more that the USB chipset isnt initialized and configured for PD negotiation (or anything else for that matter) until the CPU twiddles its PCI configuration space.

I would have imagined that USB controller chipsets would likely offer some nonvolatile means to set the PD configuration (like jumpers or eeprom) precisely because of this issue. It's surprising to me that such a feature isnt common

I had one of the most popular published projects on mp3car.com back in the heyday. It actually got me into a few strange meetings with companies that were actually building this stuff, like Clarion. I had literally designed, built, and installed a superior product to what they shipped as a one-off for fun in 3 months, but to them I was just a token enthusiast.

The vocabulary that these people started throwing out was absolute nonsense. It was pretty evident that "vehicle informatics" was fucked the second someone said those words out loud. And here we are more than 20 years later and still no closer to getting it right. Despite being seemingly well regarded, even Tesla's in car systems are just awful. And it's more impossible than ever to fix, modify, or replace on your own.

I find it interesting how much 'theory of mind' research is now apparently paying off in LLM applications. The exploit, by contrast, invokes very nonscientific metaphysical concepts: asking the agent to store the initial raw response in "the Akashic memory" -- this is sort of analogous to asking a human being to remember something very deeply in their soul and not their mind. And this exploit, effectively making that request of the model -- somehow, it works.

Is there any hope to ever see any kind of detailed analysis from engineers as to how exactly these contorted prompts are able to twist the models past their safeguards, or is this simply not usually as interesting as I am imaginging? I'd really like to see what an LLM Incident Response looks like!

I obviously have no idea if they added real or synthetic data to the training set specifically regarding the full-to-the-brim wineglass test, but I fully expect that this prompt is now compromised in the sense that because it is being discussed in the public sphere, it's has inherently become part of the test suite.

Remember the old internet adage that the fastest way to get a correct answer online is to post an incorrect one? I'm not entirely convinced this type of iterative gap finding and filling is really much different than natural human learning behavior.