Also, most of NATO's existing resources are currently not doing much but standing around. You can question how useful NATO's 'pre-drone' material is, and how much hasn't already been sent to Ukraine, but it's not nothing.

This is about remote control. People who think Apple/Google/Microsoft will just pass on the reigns of that remote control are dangerously naïve.

Apple has locked down whole classes of third-party utilities for its mobile system and Google is copying them as we speak. We are slowing progress at great cost by letting them lock-in and rent seek.

But more to your point, I think what makes this httpd scheme especially attractive (for C users especially) is that you can't try to "optimize" anything. There is no temptation to lock/mutex/cache yourself into more problems. Its 1 process that then dies. Even 'free' becomes optional in some cases.

To torture a different analogy, we're all prospecting and yes i can see that the labs have the option to do strip mining which I dont. But I have tried various forms of bulk processing.

The samples i've collected point to the fact that strip mining just adds in more dirt, and i'm asking if there are more positive samples or if you've inferred their existence.

Now even without that, another anti-signal to me is that we have 1 unquestionable result afaict which is the openai math proofs; and they didnt even take the time to wash out the dirt (https://github.com/offline-ant/wiggums-proof-loop) and the trick of scaling really isn't all that mind-bending beyond increasing an integer somewhere to gamble with more money.

I'm sure there are 5000 people experiencing something unique and categorically different with dealing with problems, but unique does not mean they're the place for a glimpse into the future or holders of exclusive insights. So I'm questioning the use of the separation.

Well, the use beyond establishing a flattering hierarchy of course.

The internal models are more eager, and there is the scale difference of being able to spin up 10.000 vs 10, but to get something valuable out requires skill.

Agents that produce good results 90% of the time vs 93% of the time are not enough of a difference. Multiplied too many times, that error rate is too high for scale to become "magic".

Finding proofs is one of the exceptions.

Just look at the codex/claude harness. They regularly lag behind 'external' users.

Use Rust and the chances of an application overflow drop off a cliff. The debug-ability also increases the surface for bugs themselves. The majority of cases we're talking about chips that can comfortably be hardware-simulated. And/or all you need is some way for a dev machine to edit memory and an AI can work through the night to explore a configuration space much larger than you would ever have.

I might be interested in this because that is cheap, but what exactly are the networking rules/costs? ip4/6, bandwidth, egress, throughput, connection limits?

You are self-assured enough to think you know better, articulate enough to write down what it ought to do.

I'm stuck because i do not know how to talk to you. I can empathize with people not mentally flexible enough to use these tools that feel stuck. Thinking about workflows is not trivial. I can argue with people who do build their own workflows; about our experiences what works and what doesn't and why.

Your frustratingly falling in between. You obviously have ideas, but you didn't bother to have them meet reality and get refined by it, yet you seem to want others to consider their merit - as though that's a good use of their time.

In the analogy the printer does work.

I'm needling you - since you've claimed my attention to show me your draft of the manual - to turn the damn thing on before we continue discussing the merit of your ideas.

I dont know how long that's going to take you the first time, but once you've found ideas that do survive contact with reality - instead of presuming they will - you can delete the code and absolutely rebuild them from scratch with pi in 5 hours.

Or - if you're happy to - i'll politely bikeshed the color of the title or something and move on.

The thing is, a year ago the models started to be good enough that i cut my dev time down by 50%, and i started to use 30% of my time on improving my tools; and still feel like a win overall. Exhausting it is though.

The big chasm i saw grow in the early months of this year are primarily between the people who had and took the time/opportunity to toy with workflows, and the people who didn't.

So to your problems i do not recognize them:

Give the harness their own user account and launch it with no exec limits. Domain specific knowledge needs their own markdown files. AGENTS.md has important rules, and references to a well-kept set of source-of-truth Markdown documents. You need to remember what is in the context. Avoid compression at almost all costs; You want to be able to restart a session without it being too much of a setback (i.e. keep Markdown docs).

But it all boils down to the ability to: spot a workflow issue, think it can be improved, and then just improving it.

Every once in a while I go to my abandonware and strip out the things I thought would work but didn't.

Why aren't you using the printer?

If you can't use the printer, why do you believe that other people should believe your ideas about the printer are worth listening to?

Feed your blog post to pi - have it implement it - and you will realize pretty quickly that a lot of them are flawed. This would have costs you about the same time as writing this blog post.

I'd rather be reading the blog post that comes after you reflect on what you got wrong and right. Better yet, the blog post after you've iterated enough that you believe your ideas are right.

I don't care for your hand drawn instruction manual on how to use the printer.

I started to gloss over hard after a few paragraphs because I don't really have any the problems you describe anymor; at least not to the point i'd be blogging about them. Instead, I've just been iterating with an agent on various pi extensions that solve the issues.

As with operating systems - you can hold out in the hope somebody solves the right mix of issues in general, and they match your situation well-enough.

That's your choice.

I would note though that being the passive consumer gets you either Mac or Windows UX & prices.

The goal is to add a data point to the discussion.

Comments URL: https://news.ycombinator.com/item?id=50026202

Points: 1

# Comments: 0

'Emulation' is a general English term applicable outside digital systems.

A group of people had problems and solutions that required an expanded vocabulary in that space, and they choose to re-uses the generic term to be a specific technique, and to use other terms for other techniques.

You're complaining that this doesn't align with your memory of a different group of people who picked different.

Complain all you want, but this is just how language works. The kids playing Roblox right now will grow up, and replay this whole spiel in their own way.

Have we tried a good ol Ralph Wiggum:

> simplify/shrink this proof

and see what we get?

The obvious next step is to get enough memory throughput to run that SOTA model itself so that it develop its own hardware.

But perhaps the more interesting question is this: Can an AI be given a big FPGA and design a model architecture that takes advantage of the fabric being reconfigurable.

Or just use the other model.

NVIDIA has bought HuggingFace.

Jensen, in my irrational hope he is susceptible to random comments on HN, should grow some balls and do the world a huge favor, by suing OpenAI to set the legal precedence.

> Claude/codex/deepseek, please replace 25% of words with synonyms or slight rephrasing because i dont like the current version.

Not sure if that counts as: `a solution that's robust against "common alterations and adversarial attacks"`. Is there a sort of adversarial attack that is more common?

Instead, I have a (pi.dev) tool called `do` which forks the agent and spawns a task on top of it - then returns the last message to the parent agent. (or `delegate` to start with a fresh context).

That's it.

claude/codex use it for a lot of things (impl/review/search/misc), and thus including situations RAG would otherwise be used. By skipping RAG, these search sub-agents instead use tools they've been trained on to use (find, grep, etc), and can decide to look deeper if required.

I'm not saying RAG couldn't be made to work, but with my set-up i've automatically gotten improvement every time a new range of models was released.

When you have knowledge distributed in markdown files; finding them puts the path/filename into context as well as some indication of document size. (If its on line 1200 or line 20). This is extremely valuable for picking what ought to be focused on next.

RAG on the other hand creates the hardest challenge for these models. It instead puts 5 ideas with the highest similarity into the context in full.

Its the difference between having to remember a set of numbers when in a crowd that's talking about stuff, and having to remember them when the crowd is shouting out random numbers. The similarity in the task makes things harder. SoTA models work despite this, but its extra-gambling while you're already gambling.