I remember discovering ? as a shortcut for PRINT on the C64. That was when software development as a profession really started looking like a good choice to young me.

Are you in the Salmon business?

  > It often surfaces ideas and approaches that I did not know about.

I'd love to hear some examples. I do agree that tech is moving at a pace that I can no longer keep up, however that absolutely does not mean that I want a dependency on some project that didn't even exist this time last year.

If I search for The Best of Both Worlds but open the video to find it's csam and not st:tng then I've unwittingly started watching, and thus potentially seeding, the csam.

But even simple checks like the number of seeders on a torrent will defeat a vast majority of this problem. I would not expect a torrent with 1k seeders to be csam for e.g.

I understand the worry, and it is a difficult problem worth thought but the alarmist virtue signalling helps absolutely no one, victims included.

That might work for popular culture. It's probably a good strategy if you want to watch the latest superhero movie.

I listen to the Butthole Surfers and Amboog-a-Lard - I'm willing to bet a good portion of their material has less seeders than much illegal content do.

Are you not describing perfect as the enemy of good?

A product doesn't need to serve EVERY human on earths needs, just enough of them that it can become large enough that established legacy solutions need to adapt.

I get why you might think this wouldn't be a fit for YOU but rather than argue against it from a niche viewpoint instead maybe look at how if it did exist it could be good for many.

That's pretty unlikely. P2P can have moderation systems. Also that can be solved by only seeding videos you kept watching.

Is that actually done in practice?

There's no popular peer to peer video system in the first place so uh not yet?

Peertube has moderation.

This is my burning question as well.

I have 80,000 voice recordings to classify. Very few are in English, and the classes are not in English. Many of the classes are project names or other proper nouns. How could a system not trained on data specific to the problem possibly be expected to work?

you could help it along by providing clues about the domain in the prompt. including what various proper nouns might indicate (assuming they're not chosen completely at random, and encode at least some meaning).

and you might find the scores it gives back about its confidence are useful for escalating to a more expensive classifier.

but also, there's no requirement to use it as a zero-shot classifier, you can provide it as many examples as you like, and prioritise giving it examples it had previously gotten wrong. and with input caching it might be economical.

I'd be surprised if they or others don't start offering a fine tuning API for models like this, like openai does for some models (or used to, I haven't checked in a long time).

If it's not trained on my data, it's going to be something on the order of a 500-shot prompt. This is a complex application with more corner cases than I'd like.

As are, honestly, most projects I work on.

There are already Jev-shaped open weight local models coming out that can be fine-tuned on specific data. We'll see which of them the community settles on.

1. When you finish the milk, voice record "buy milk" on the phone.

2. At supermarket, listen to voice recordings.

I'm actually working on a pet project to transcribe my 20 years of voice recordings and organise them, then sync them between the phone and desktop. Maybe in the coming weeks I'll know that I needed milk in October 2008.

And what happens when its your kid who finishes the milk and doesn't mention it to you?

Or you didn't finish the milk, but didn't realize it had gone bad until after you got home with ingredients for a meal depending on milk.

Have a shared shopping list with the family. Teach your kid to put the milk on the list after they finish it. Be at least somewhat aware of how old your food is. Problem solved. Somehow we managed this model for a century.

Somehow we also managed to get by before we had pretty much any kind of technology. Somehow we managed without trains, somehow we managed without telecommunications, somehow we managed without refrigeration, somehow we managed without websites and forums.

Maybe we could do better. Maybe we can automate this as well.

Yup, end game is becoming a Guild Navigator. Obese, immobile, suspended in a tank full of dopamine. Why beat around the bush with all this doing stuff and using our limbs or god forbid thinking? Straight to the neurotransmitters, pure efficiency.

I take it you don't bother writing things down, wouldn't want to have the paper take the place of your own memory turning your brain to mush right?

And obviously you just walk anywhere and everywhere. Wouldn't want your legs to grow feeble and useless. Even riding a bicycle is reducing how much exercise you'd get otherwise, it's simply a tool for the weak.

Well, then, of your cows can only be milked on the 'morn you'll have to do with milking one of the goats.

I mean, there are plenty of voice assistants these days. We use our Alexa for this but both Siri and Gemini could easily do it. Whenever we run low on something, "Alexa, add milk to our shopping list" and then can look at application at grocery store.

  > So if someone opens your unintentionally unlocked front door and steals your laptop, you don't care who's liable?

There are literally hundreds of thousands of script kiddies poking around at servers all the time. Cloudflare stops 99.99% of them. You're still left with many entities from states to hobbists trying to crack your system after they've gotten past the CDN and captchas. If OpenAI got through, then somebody else could too. Somebody malicious even.

I appreciate when white hats inform companies, governments, and the public about their successful exploits. It helps keep the whole internet safer - even if just by waking up lax server admins.

> I appreciate when white hats inform companies, governments, and the public about their successful exploits.

Except it took 3 months for OpenAI to notice they did! OpenAI is not a white hat doing pentesting, but someone that is putting random materials on fire to see if they smell. Reckless, stupid, and criminal.

So what? OpenAI may be a problem, but the security of a government website rests on the administrators of that government website. Not on attackers playing nice.

I'd love to hear more about this. Though I usually write bespoke software for specific clients, I'm about to release my first general-availability application. It's a rust application with Python GUI on the desktop (Linux, Mac, Windows) and a skinny Android application that wraps the rust on mobile. I'm going to be gating features starting probably next week. I'd love to hear tips and experience for implementing feature flags, for both the desktop and mobile space.

Thanks!

To the guy who asked but his comment is dead:

  > Are you gating by license or user, or just by percentage of installs?

I'm gating by version. I want to dogfood features personally before releasing them to all my users.

Bash does this well, reading only the first of half a dozen config file locations it searches. And if the user needs it to read from an additional location, just add a source line to that location.

I would like to subscribe to your newsletter.

Seriously, though, this is a great idea. Good communicators know to tailor their discourse to their audience. LLMs are just as sensitive to small variations in prompts, and just as inconsistent between them, as are humans.