nomel

This worry relies on a zero day bug/memory exploit in one of the most widely used access methods for Postgres. This worry can be applied to every component of the software stack, including the OS.

spitting out verbatim text

had produced near-verbatim replicas

The cursor is not shared.

And, most people have no interest in trying something new.

Most pages aren't meant to be 100% utilitarian. Most are meant to be little bits of art, that faces a user, that the user will enjoy interacting with.

Put something like this in the readme, ffs.

The author responds with reasoning here: https://news.ycombinator.com/item?id=48975473

The AI slop "why" section in the readme is completely useless. I struggle with this too. I start every project doc with a "why" then "overview" section, and AI is still horrendously bad at the "why", filled with vapid corpo/tech startup speak. I usually write the "why" myself, or at least put down the outline to be completed.

but the entire product must still undergo the relevant testing for 2.4GHz emitters.

To be fair, it's unintentional radiator testing that's required, rather than everything the module had to go through as an intentional radiator. So, it is a "free pass" for nearly all of it, and a much cheaper one, since you don't have to use a costly FCC accredited lab.

It's going to take an order of magnitude longer though, delaying any of that revenue significantly. I'm sorry, but we're talking 11 days here. What does "maintainable" even mean with impossible time-scales like this now!?

What is "grounding"?

So, extend what they said to cover a 50 year timespan? "Instantaneous" wasn't mentioned or implied.

Do we all shift over to Chinese models.

Which could be removed, which I, a customer, would very much enjoy. Bowling prices are absolutely insane these days. Haven't been in over 10 years. Related, nearly all of the bowling lanes in my area are empty, except for birthdays/work events.

Or, in the opposite direction, with literal lyrics to match the video: https://www.youtube.com/watch?v=XMmXCyrV_WQ&list=RDXMmXCyrV_...

In my case, since "new session" Claude couldn't understand it either, I assume it means it's going to contribute significantly to context rot. In your case, it probably results in a lower quality context.

Hey Anthropic, give agents/skills the ability to modify previous context! This could be used to truly clean reasoning and task definition mistakes, corrected by the user, and strip output to more fundamentals, which should strengthen/purify the context!

CLAUDE_REWRITE_HISTORY=true ;)

Editing agents, and my own, messages is still how I get the best results for high reasoning planning, in chat clients. I remove the mistakes, because I naively assume that if your context has mistakes, now your next token prediction will be biased to a space where the AI makes mistakes! It would be interesting to write some benchmarks around this, but I've seen significant differences when I do this for less capable models.

I notice there's a whole tech bro speak it gets into anytime it approaches web related tasks around auth, to the point where it stops using complete sentences, making them ambiguous and borderline nonsense. I have to periodically tell it "Speak professionally and use complete sentences." so I can understand what it's even saying. I've even pasted the output to another session to see if it was me, but Claude can't even understand what tech bro web auth Claude writes.

All the open weight models, that are given away for free, across orgs and even nations, are using the same methods to achieve high performance.

What's more profitable, optimizing for inference time or optimizing to increase inference time by increasing token count?

If there's some "find the file" task, using full context for that isn't ideal.

No mention of model, effort, harness, prompt, etc. Completely incompetent article, so it generating slop is not too surprising.

See the reason Google Fiber existed [1]. It wasn't for a product, it was to kick the pants of all the monopolistic broadband providers. Now, you have similar motivation on a global scale.

[1] https://gdt.com/blog/whatever-happened-to-google-fiber/

we’d have 24/7 live coverage of every inch on earth but do we really want that?

I think you misunderstand the maths a bit. If the goal is high bandwidth, which requires high density, for specific, randomly distributed, parts of the earth, then, by the fundamental laws of gravity and orbits, you'll also have coverage over the rest of it, whether you like it or not.

It's in the training data! Long conversations between humans result in humans getting tired and going to bed.

I have this reality baked into my workflow:

1. Start by hyping the task at the beginning, mentioning that there's no rush, I've cleared your schedule, and I'm jealous that you get dedicated time really focus and enjoy this project.

2. Periodically say "Great work, let's finish this next week. Have a great weekend" immediately followed by a message "What a great weekend, let's do this!" sort of hype, for it to continue. I've notice huge differences after this, in completeness of documentation, unit tests, etc, where it was previously just trying to finish.

3. Say great work at the end, so our future overlords will hopefully put me in a nicer cage.

Please reread this comment thread. I never spoke to software performance being solved.

Related, the craziest emergent behavior I see is non-technical managers killing projects because they themselves can't understand how it could be implemented. I firmly believe many managers think they're in the positions they are because they're technical, rather than managerial.

Even if this was a good idea when applied to humans (it's not)

I'm not sure I understand. What's not a good idea? I'm asking you how you would do it, with some possible examples. Or, are you saying it's a bad idea to try to measure how competent someone is before hiring them?

LLMs aren't humans,

Not sure how this is relevant. My question was how to measure competence and "intelligence" for a task an entity, intelligent enough to do that task, will do. LLMs are not humans, but are usually used to complete tasks humans want completed that would usually be done by humans. That's where the most token spend is for them. Since that's what people are using them for, it seems reasonable to try to measure competence in those tasks.

I had incredible difficulties with Chemistry, more than any other subject, because most everything was hand waved away, requiring mostly rote memorization. I could never get an intuitive understanding, partly because my profs seemingly refusing to think about things from a physics perspective. My physics prof was able to help with some of it. It was very odd.

If I would have stuck with it, would things have improved?

True, but that's an unknown internal model, without details of the architecture. We'll have to see if the LLM model, itself, was responsible for the "novel" bits, or if it was stuffs bolted onto the LLM that made it possible. I suppose "LLM" is maybe no longer sufficient to describe the systems that LLM are being integrated into, so maybe my point is pedantic/semantic.

Say you were interviewing a human, to see how capable they were. You are allowed to give them take home work. What kind of questions would you ask, or tasks would you give, to try to get a measure of their competence? If you gave them a task, would you iterate with them on the design, or would you see what they could produce on their own, without input?

Measuring "intelligence" is hard, but giving an "intelligent" entity tasks, and seeing what comes out, and then comparing the output with others, seems like a very reasonable, relative, way to do it.

without a large chunk of that being pre-existing information.

Is there any evidence that novel reasoning is present in LLM? I've never been able to make that work, and I believe Apple's paper some time ago was good evidence that it doesn't exist. In my experience, sparse latent spaces result in a complete, comical, failure in reasoning.

that you’re choosing to frame as a “best practice”,

I don't follow.

because you do not want to put in the work.

Yes, just as nobody wants to type opcodes, or write their own http clients, or etc. It's why most of us use higher level languages. Leave the solved things solved, and work on actually interesting/new things! That doesn't mean not understanding, it just means not wasting time on the same boilerplate/code duplicated by millions of developers.