apt-apt-apt-apt

Aren't there many places where a highway with no sidewalk is the only walkable option? So if someone saw your not-walkable, they might think not walkable at all.

Seems like you guys had two different perspectives, both arguably right. Maybe it needs to be split into technically-walkable and comfortably-walkable layers.

How can you say there are no challenges and things to master when working with AI? Unless you are just doing "Make this. LGTM. Now make this".

Fable 5 seems like it got nerfed when 5.1 came out.

I'm reading tons of Claude, all I do everyday is read all day

Wonder how much of this is caused by SwiftUI

Claude, create Claude. Wall clock time of 33 hrs so we can impress SJMG.

AIUI, it is a type of factory that produces software.

I think everyone wants the experience of living, but it's not a great one without money, and money commonly requires full-time work until financial freedom. But that point usually comes at 65.

So it's not that only children appreciate life, it's just that adults have to deal with that whole money problem, while children don't.

It showed VIZIO revenue numbers ($1B hardware sales at -0.8% profit margin, ~$500M ad revenue at 61% profit margin).

Based on that, it seems to me that TVs are a one-time-sale commodity with low margins, while ads are continual revenue generators with high margins. This makes the incentive extremely clear.

Unfortunately, considering that 80-90% of people simply don't think or care about this, I can see the incentive.

What's the incentive for these manufacturers to do this?

Are they getting paid for the data (e.g. we'll pay you $100M to install spyware), or are they trying to be an ad platform, or is it simply analytics gone out of control?

That made me double take too, but he must be talking about the software getting better, and thus higher value.

I'm so suspicious of this after Opus 5 benchmarks scored it higher than Fable 5, yet Opus 5 was untrustworthy (overconfident, error-prone).

Is this the kind of thing that works when you have a tight domain-specific language, but devolves into natural language (English) eventually?

E.g. A -> B, B -> C so A -> C (works, great). A -> B when A is sort of red and blue, also A is intermittent (what to do now?)

Sub-humble-wizard here, how do I sign up?

I'll be launching in a week.

"Claude, a business is making $30K/mo converting bank statements!

Use all of my max 20x plan to build a BETTER version. Must have animated per-page progress indicators"

Perhaps – but it's a useful grammatical construct in everyday use, handy for structuring sentences.

And with the prevalence of AI, it rubs off on the smelly humans, who also find it useful and use it more as a result. (I used an en dash here instead of an em one – also valid, easier to type, and shorter).

Huge bloated article that unclearly expresses a simple thought.

It's like the author prompted, "write me a big article on what is and isn't a commodity with AI, and make it confusing"

Wish I had known it was fiction – read half of it thinking it was real.

I was surprised when I bought my 'cheap, crappy' 2024 M3 Air ($1.3K) and it ran circles around my top-of-line last Intel MBP ($4.6K).

HN makes me laugh like twice a year – you're either the first or second!

This happens all the time and drives me bananas.

It seems to be regular file edits rather than parallel tool calls.

I'm sitting on 1.6m cache write even now with 468k in /context. It drives my session costs above $100 regularly.

Can someone from Anthropic look into this?

I don't use sub-agents, only the normal linear message-response flow.

I'm finding that unexpected cache rewrites cost me huge.

I have 1h cache TTL set, and do nothing to cause rewrite (response in time, no model/effort/tool changes).

At 400K tokens in, I'll write a message, and /usage shows only a small increase in cache write. On the next message, cache writes shows 800K, and by the end, I often hit 2M cache writes with no explanation.

This seems to happen when: using /btw, asking it to review code, other random times. Anyone know what's going on?

Honestly, how are you guys running multiple agents, presumably with them all writing code?

I find myself to be a huge bottleneck, having to do technical design and design reviews, to make sure they're actually working on the right things.

I feel like there's not enough coding work to justify multiple agents, that part doesn't take long, and then it's back to design again.

There is. They had a civilization long ago, but now they're down to a small village and only have a radio that they use once every 1000 years.

I managed to talk to them a few years back on my ham. They asked me to send them energy, I told them we'd need to figure out fusion energy first and that we'd get back to them next millenium.

I feel like this is going to add 100K tokens and 10K rules to everything that the agent is trying to do. Sort of like dumping the Clean Code book into context and saying, hey now you know how to code cleanly, write great code now

CS6 was peak, everything we needed, nothing we didn't. That's pretty much what photopea is, with improvements, which supports this view.

"We've now re-explored this question three times. While there were small differences along the way, one more would be pathological. Perhaps meditation or therapy would be more fitting."

This is basically how Fable told me to get therapy.

These are getting really good. Much more interesting than the average music video already.