rescbr

I wonder how a RWKV variant of this would fare...

Lately I've seen this statement that I agree with: for some people, a hobby is a means to an end (which means using AI to achieve the goal is a heresy), for others, it is the end in itself (which it isn't).

> Public got the last word, industry suffered a 40 year ice age.

This is the worst outcome.

I just heard about them today. The only other person mentioned in the story that I've heard about once was Aella.

I mean, why should I have heard about them if I'm not in SV?

Me too. I worked in two enterprise tech companies that definitely everybody here knows and I had the same experience with YC and HN. The two colleagues that knew about it, one went to found his own startup and went through YC, the other left the big co to work for a startup after about a year.

> Less useful for experienced users, but for beginners they’re a lifesaver.

The last emerge conflict I had was handled 100% by an agent.

I'm not wasting my weekend on stupid sysadmin tasks anymore.

Well, it's the enterprise software house pipeline... The software architect writes the spec, hands it down to the implementation team, senior leads, junior devs or offshore teams codes it.

I also disagree with the approach, this is cargo-culting the existing ways of working.

Exactly. Sometimes I just need a 3D printed part to finish a DIY project, I don't want to waste hours designing it. I had my fair share of time playing around with SketchUp when I was a teenager, but that's it.

Criminals don't need Astra or Fable. Get a bunch of GPUs, a good enough open weight model and a custom harness. What the model doesn't have in its weights it can research the Internet. How do you think black hat hacking is being done right now?

On bio: while it's easy to order all the raw material, there's the whole process of culturing bacteria/viruses/etc. that isn't trivial, and while a LLM might help in explaining stuff for rookies, there needs to be somebody physically working on it. It's not automatic. Once you get to this level, you'll find that any undergrad biologist or chemist can already make bio/chem weapons, and you don't see it happening. Terrorist groups already employ biologists or chemists that are supporters of their cause, no need for LLMs.

On cyber: aside of the question of why critical infrastructure is on the Internet on the first place (ah yes, lowest bidders, people not caring enough, business not giving IT/OT budget, S in IoT standing for security etc), if the available models can't handle cyber tasks, how can you defend yourself? See the HF "attack" that HF had to use GLM-5.2 to investigate what happened. This is the best argument for open models. Unless of course, you are the AI model creator or somebody they authorized to use their sanctioned model/harness and want to create a moat for their business.

This is FUD. Creation of business moats by fear.

I'm getting "spike" for a while now, and just found out the newest word which is "gauntlet".

Same feeling as well. Bluesky is also a place amenable to bots.

Huh, maybe that’s why I’m more empathetic towards them compared to my fellow millennials?

Uh, actively trying to hack an embedded device that runs Linux over the network, specifically an IP security camera, could be considered red teaming, no?

You never mentioned which exact activities you were getting flagged on and getting refused.

What is the harness that you're using?

I found that GLM-5.2 was pretty happy helping me reverse engineer/hack devices.

Maybe the system prompt you're injecting is making it refuse?

For my own stuff I simply don't care if Z.ai is learning straight from me or from my GitHub repo a few days later when I push the code.

When I use it for reverse engineering, same thing, the code is already out there, but in binary form, which is becoming trivial for these LLMs to work on.

I even bought some shares from them and got some money out of it!

Why do you assume $20k as the cost to run the fiber?

The cost is the contractor's hours to get up a ladder on each electricity pole from the closest PON splitter to the customer's home, plus the fiber itself which is peanuts at volume.

Everything else (permits, pole rent, etc) is accounting money that can be negotiated. If there's a strategic national program to fiber up the whole country, this could be exempted on a pen's stroke, or depreciated in many decades.

Direct democracy doesn't mean everyone votes on everything.

At the extreme you could do something like localized e-voting: citizens vote on issues that affect a given geographic radius of where they live. Because the boundaries are purely distance-based, gerrymandering is impossible.

Maybe we could have a decaying distance boundary by some measurement like population density or something like it?

The issue is that those marketplaces got too mainstream, and the average consumer is too stupid.

10-15 years ago if you were importing stuff from China you'd know that the trade off for cheaper prices is that you'd lose your consumer laws protection and you'd be on your own. If you wanted those, then you'd go to your local retailer and pay extra.

They go to the decent computer hosting the model, which can be yours if you pay the electricity costs

I know how mmap works. The JVM is/was terrible on freeing allocated memory though.

If the program is actively using that allocation, that's fine. My problem is with the runtime hoarding RAM when it should have been freed after GC back to the OS.

Then there's also the JVM not handling peaks well because it hit the max heap size, while you still could rely on the OS doing its job to shuffle stuff to swap temporarily. I still see JVM OOMs in my $dayjob's product while the OS has plenty of free physical memory. It is stupid.

I mean, we have malloc() and free(), they are in the stdlib for a reason :)

The JVM seems to follow a philosophy where it assumes it is the only process running besides PID 1, which is valid for some scenarios, but not for others.

The JVM has a fixed size heap which for me it is wasteful.

IMHO, Microsoft made the correct approach on .NET.

For LLMs, I prefer C# and C++ instead of TypeScript, JavaScript or Python as the static + compiled language factor keeps the coding agents on track. Plus, they have a true threading/async implementation.

I guess it’s ages of fake download button ads in download pages.

My instinct is “if it is an ad, then whatever it is offering is a scam”.

Eh, the ugly side of humanity was just a click away on the darknet, and I got my share of bait and switch beheading videos on IRC chats.

It was always there, society hasn’t changed. This shit wasn’t monetized back then though.

Agreed. I was a middle-schooler in the 2000's, had my own PC in my own bedroom, we had broadband access and I, of course, was exposed to porn while I was trying to download cracks and keygens on Astalavista and other such sites. Mislabelled files on eMule too. Links to porn also were readily shared on IRC channels. But yeah, I can't say it had a damaging effect on me.

It was a mix of being afraid of getting a virus from a porn site, knowing that porn wasn't for my age, and late-night FTA TV airing softcore anyways.

My parents told me to be responsible and kind of explained the dangers of the world the best way they could, and it was expected that I'd be grounded and lose computer access if they found me on a porn site. I preferred playing games. I definitely encountered some dark places on the clearnet (and on the darknet as well) and promptly GTFO.

At the school's computer lab they had plenty of content control which I had the most fun trying to bypass them. For what reason? Just so I could.

AI doesn't get tired. I can tell it to profile the application and find quick wins and it usually does a decent enough job.

Sure, but you can review the git repo's content/commits in plain text, while curl-pipe-bash would require you to reverse engineer the binary that's downloaded.

If somebody hacks the project's home page and switches the download location to a hacked binary, you'd be none the wiser. Of course, somebody could hack the repo and add a deliberate vulnerability as well, but at least you would have a trail of it.

I personally also don't like this, so use cargo to install it, it's in the readme. Takes a reasonable time to build on my X220, but on a modern computer it is plenty fast.

  cargo install --locked --git https://github.com/tontinton/maki.git maki

I'm using maki (https://maki.sh), it has 1, 2 and 3.

This is definitely not my experience with GLM-5.2. It is writing pretty good ARM SIMD code.

It also is an excellent radare2/rizin/ghidra driver as well. Maybe it's because its ~cyber~ capabilities (pretty much linked with assembly-level knowledge) aren't guardrailed off?

> questions about the modern Chinese history, to which a Chinese LLM will not answer

This has been debunked here on HN so many times. The Chinese open models do answer the hairy Chinese political questions, and the raw APIs pass-through the response. Now, the answer might be blocked by the agent who's calling the API, specially if you are using a Chinese endpoint instead of the RoW (i.e. Singapore) endpoint.

That's the reason why you should always prefer a open agent/harness as well instead of using the provider's.