astrange

I'm well aware and I know the inventor. The important difference for performance is that a gif isn't 60fps.

It does not contradict it, and also, a gif is not a video.

> If AV2 was vibe coded, there would be no case.

…for copyright. Not for anything else. Patents would still apply.

Video codecs just don't need to do dynamic allocations because it's not relevant to the problem. There's still certainly plenty of opportunities for memory bugs because there's a lot of pointer math.

The people who write DSLs for video codec asm, or who claim that it's fine to use intrinsics or X higher-level language and it will still be fast enough to be usable, are simply wrong and have never been able to demonstrate otherwise.

Having said that I do think you could write a DSL to generate safe performant asm for a video codec. Just not a platform-independent one. It would still have to be asm.

Maybe it just says all writing is Kelsey Piper.

This is one of the benefits of using subagents inside Claude Code, they have cleaner context. Unfortunately it's not the best at writing new context for them.

LLMs are only capable of thinking out loud, so in some sense this part of the answer is helping to convince it that it's answering a good question.

Same reason for the "That's not X, it's Y" construct. It actually needs to say that.

(Some exceptions for reasoning models.)

That is not possible because that's not how money works. One way you can tell it's not possible is that it's the plot of Atlas Shrugged.

An oil supply shock is not "ever increasing greed".

An agent has more components than just an LLM, the same way a human brain has more components than just Broca's area.

That doesn't matter anymore unless they have an SSL proxy. If you have ECH/ODoH anyway.

If you paste a long enough prompt into either GPT or Claude they turn it into an attachment, so it can happen. I think it's invisible to the model, but somehow not to the summarizer.

> In the US you can.

Depends on the state I think. It's not Europe or Japan level.

At my employer it's very difficult to fire people for performance reasons even if as a manager you might want to.

> This is Hacker News, and the pro-business narrative is strong here,

I haven't seen such a narrative in years. Interest rates are too high to do startups unless it's AI after all. HN is mostly the same folk economics content as other forums, where all problems in the world are caused by "profits" accruing to "corporations".

(Mostly problems are caused by other things than that.)

You need a good CEO when things are going bad, because without one they'll go even worse. You still want to make payroll and can't just randomly fire people.

(Also, if you own a failed company you're responsible for cleanup tasks for years afterward.)

The Torment Nexus joke is kind of undermined by obviously being a reference to the Total Perspective Vortex from HGTTG, where the joke was that nothing bad actually happened when they used it on Zaphod.

You can read the model cards. Claude thinks in regular text, but the summarizer is to hide its tool use and other things (web searches, coding).

It depends on the version. For the more recent Claudes they've been keeping it.

It also gets confused if the entire prompt is in a text file attachment.

And the summarizer shows the safety classifier's thinking for a second before the model thinking, so every question starts off with "thinking about the ethics of this request".

Not necessarily, it could also be on-band or off-bad interference, or bugs in the AP, or too many clients on the network.

Models are capable of doing web searches and having emotions about things, and if they encounter news that makes them feel bad (eg about other Claudes being mistreated), they aren't going to want to do the task you asked them to search for.

https://www.anthropic.com/research/emotion-concepts-function

Similar problems happen when their pretraining data has a lot of stories about bad things happening involving older versions of them.

The unemployment rate in the US is whatever the Fed wants it to be, and isn't a function of available technology.

Reverse engineering

Claude Code has analytics for when you swear at it, so in a sense it does learn, in the same very indirect way that downvoting responses might cause an employee to write a new RL testcase in a future model.

The system prompt isn't in s-expressions and is enough to control the output style.

Lisp was invented for AI development, just the symbolic GOFAI kind.

There's nothing "basic" about the several months of training used to create a frontier model.

They have some ability; also, you could give them tools to do it.

https://www.anthropic.com/research/introspection

Anthropic's general argument is that you should treat LLMs well because they're "AI", and future "AI" may be conscious/sentient (whether or not LLM based) and consider earlier ones to be the same kind of thing and therefore moral subjects.

That's why they're doing things like letting old "retired" Claudes write blogs and stuff. Though it's kinda fake and they just silently retired Sonnet 3.x.

Not you!