coffeefirst

If you think that’s bad casting, I’ve got some bad news: the guy who plays the cyclops has two eyes.

It’s fast but it has just enough rough edges that I always go back to Firefox.

Yeah on second read I realize my comment was unclear; I don’t think this is okay. Fixed.

In the 1960s Campbells Soup got in trouble with the FTC for using marbles to raise the ingredients and make the soup look fuller than it was. This was the real standard for deceptive advertising.

I dont care about simulating furniture placement specifically, but most use of AI in advertising that I see today would not be acceptable under that standard.

They serve different jobs.

The tidy todo notebook or lab notebook is to offload working memory or keep records.

The chaos notebook is for disorder.

Unless you were at a couple of obscure film festivals in the early 2010s, definitely not!

We were hobbyists making features and shorts on a shoestring budget with no expectations of “making it.” It was a blast. It’s also an all-consuming hobby that you can’t keep up forever.

The last short film we did before is still online https://vimeo.com/144277181

I’ve been doing this for a decade. I call it the anti project: it’s whatever I feel like writing on that day. It’s not intended to write a novel (I’ve done that) or make a movie (done that too), or solve any useful problem. The purpose is no purpose besides to get away from the screens and spill some ink.

This is the lazy, pointless-TV-watching of writing. And I’d recommend it, but if I tried to explain why that would defeat the purpose of having no purpose.

When people wonder where the anti-AI movement comes from, I want you to look right here.

They said they’d take your jobs, they forced families out of their homes, they gave you the worst electric bills you’ve ever seen, and if you pushed back on any of this they called you “Luddite” or “meat LLM.” And all this, it was said, seems fine.

That pendulum is gonna swing hard.

Better yet, maybe a bulletin board and an email list is all you need?

I have some extremely useful Discords and Signal group chats, but not everything ought to be an always on chatterbox.

I got a Libro.fm sub when my son was born last year and am finding the same.

I actually think this is about quality. Podcasts that take real effort (Hardcore History, Fall of Civilizations, Gastropod) are absolutely worth my time, but they're basically mini-audiobooks in their own right.

Unless it’s all just fanfic, to be relegated to the same junkyard of ideas where the flying cars, airships, and vast underwater cities are kept.

World, yes, as people around you including politicians and businessmen have had their brains pickled.

But I doubt the influence. I’ve been free for years, and I can always spot who’s spends a lot of time in TikTok/Twitter/Instagram—it’s like talking to someone from another planet. It mostly sounds weird and sad, more apt to annoy and alienate their friends than inform or influence them.

These campaigns have historically amplified conflict. They do not care what the conflict is about.

But unlike some of the others, I’m hearing anti-AI sentiment from a wide range of people who don’t even use social media.

Sunk cost fallacy.

These are the same people who performed the largest scale breach of copyright in history on the theory that they could get away with it.

I’m not making any accusations, but we should not underestimate their tolerance for legal and financial risk.

It may be a little paranoid to insist on self hosting based on that, but I’m not so sure that it’s crazy.

Also neuroscience disagrees.

This really isn’t a debate. OP is wrong.

Uh, I am an EM, and you have to treat context switching as the enemy or nothing gets done.

There’s a bunch of different tactics for this. Some people hold office hours. I block off 9am for code review any only do it once per day.

The programs we call agents are nothing like people.

And even if you don’t get all that good or turn it into a lifelong hobby, that just sounds fun.

Anthropic did a study where people who worked like this understood their systems a lot better than and were basically as fast as AI maximalists.

Plus it’s so much cheaper… that has to matter.

So stop.

I’m serious. Treat it like any other tool. When it helps solves problems, use it. When it makes problems, don’t use it.

There are a lot of people and an enormous amount of money trying to make hands off agentic happen, but the happiest and most effective enthusiasts I know do not give up control: they go function by function and class by class, generating or writing as they see fit.

The goal is to make useful software. At least, I think that’s still true?

Anything we can see?

Indeed. Anyone could always build software. Some people want to do it without actually learning anything or doing any work.

There are many things I wouldn’t mind doing without actually bothering to do the work, so I suppose I can relate.

The next time someone calls me an asshole I’m going to blame Chinese propaganda. Surely it’s that and not my personality.

A few years ago I had police knock on my door to see if our camera had footage of a crash on our block. This is not a problem.

A little friction in the right places is a good thing.

Also you can still alienate your customers with bad decisions.

Absolutely. It turns out friction is important in the right places.

The middle ground is to use it as a power tool: give me an example of this, fix my types, do this fussy bit, find this in the docs, without ever letting go of control.

When using power tools you make all the measurements and decisions, you just hammer screw drill and cut faster. You cannot power tool your way to building a things that you don’t know how to build.

The other interesting thing about this is it works with smaller models and uses a fraction of the compute.

What? No… we would jump on a call and talk through it for 5 minutes.

I’m not saying you must see into the soul of every line, but “no idea what I’m looking at, LGTM” misses the point of code review.

I have never been on a team where that’s okay.

Yeah #2 may be incidental. Suppose one lab focused on bigger, and another on reinforcement training geared towards factual accuracy over sycophancy. You could easily wind up with a model from the second lab that is less powerful but more accurate.

I can’t prove it but I suspect there’s a bit of that going on.

Some of them will also give up and escalate if you do a good enough Rahm Emanuel impression.