stravant

Because LLMs are changing so fast that nobody wants to spend time carefully engineering an efficient setup for them which might just get thrown away in 6 months when the modality changes again.

That's not how it works.

The reason you see "dumb" refusals that "should clearly be allowed" is that they're using more traditional deterministic methods to deny prompts rather than just relying on the random LLM which you could bypass by luck.

Embarrassing for HN to have a huge thread over a Feature Flag.

https://martinfowler.com/articles/feature-toggles.html

That's just how big companies roll out software changes for software that auto-updates.

It's much preferable to be able to instantly fix it if the rollout of a new feature goes wrong than have everyone who installed the broken version bring stuck with problems until the company realizes the issue and rolls forwards with a fixed version.

https://martinfowler.com/articles/feature-toggles.html

I'm pretty sure that skip ahead is deliberately broken to satisfy advertisers.

Sponsorblock has zero issue jumping to the precise moment after the sponsor vs jump ahead always leaves just enough time to still hear the name of the sponsor.

That's a silly reply.

Imagine telling a kid to chisel what they want out of the marble line by line with a straight face when they can clearly see people appearing entire programs into existence in minutes.

The system cards are effectively a data dump for researchers to sift through.

They're not meant for normal consumers who just want to use the model for work.

People don't want to make games, they want to make multiplayer games.

Very few kids are going to go though the incredible difficulty of making multiplayer work in something like Godot.

Don't ignore the transient.

The industrial revolution created a hell on Earth for workers for the better part of a century.

Feels like a poor philosophy to me. 10 years feels like plenty of value to get out of something before switching.

Thousands of years?

We've only had the tech to be able to research this in some technical depth for a few decades (both scale of computation and genetics / imaging techniques).

People who made tools that generate roads using circular arc spines unite: https://devforum.roblox.com/t/new-geomtools-plugin-quickroad...

This would not fund the people you want it to fund.

Bad or borderline actors would be so much better at creating whatever metrics you're basing things off of that the actual value creators wouldn't stand a chance.

As for me that's a risk I'm willing to accept in return for the freedom of the code.

I'm not going to deliberately write code that's LIKELY to do more harm than good, but crippling the potential positive impact just because of some largely hypothetical risk? That feels almost selfish, what would I really be trying to avoid, personally running into a feel-bad outcome?

You can easily print bricks that work. They will just require more force to assemble than normal because you have to make them slightly undersized to make up for the lower tolerance.

Just think of how many 3d prints you've seen that consist of multiple parts friction for together.

Desperately lacking a sound elevator pitch, eh?

One way I like to write comments is describing what would break if the block / line of code weren't there.

I find that ends up being succinct and useful in the future.

Why wouldn't you want it to be a computer? Then it can be connected to your devices AND also do the job itself in a situation where it's awkward to connect to a device.

If already needs a computer in it to drive menus / modern display protocols. Having that computer be powerful enough to also decode content is barely an extra cost.

The website you're using right now is possibly the single least want of a CDN out of any of the sites I regularly visit.

What other popular site has zero images or video to speak of?

The key piece of the stack it gives you is sensible cross-platform component behaviors to put your styling / branding on to of.

Cameras in the hands is a pretty killer idea, why do things on hard-mode when you can throw extra data at the problem?

Why are people assuming they did store it after the process was completed?

With the relatively low number leaked here it could have been information collected actively during an ongoing breach, not a dump of some permanent database.

That just means that owning the walled gardens and network effects will become yet more important.

Ever heard of invasive species?

That's the problem. Mirror life may not be a problem for the biosphere in the long run but it could totally be a problem for humans in the short run.

People realized that past phishing attempts were quite badly constructed and a well constructed one is actually really easy to fall for.

That assumes he would have stopped with the shenanigans, which is a pretty big if.

There's literally hundreds of such settings. When you get into the combinations there's such a multitude of scenarios that you certainly can't have dedicated code for everything.

I suppose you could have it attempt to run a full forward-looking flight simulation to predict but part of the reason for there being so many controls is to deal with situations where the plane isn't acting like it should be, situations which would invalidate the simulation.

It seems like a flop not in user count but in that I'm not seeing it originate any valuable content.

I've seen several interesting posts from Bluesky referenced elsewhere already, but literally nothing from Threads despite it having had more users for longer.

Yes, and what does preceding research do? Get followed by more research building on it.

Probably something related to LLMs. The LLMs picking up outdated info, them trying to hoard the data. Who knows, but it smells of something LLM related.