Kye

That's become my answer to all "why not ActivityPub?" questions.

AP isn't completely stagnant but there's a reason AT is still holding on to and accelerating that early developer excitement AP had. Maybe it's marketing, maybe it's money, maybe it's some technical thing. Maybe it's the community. Whatever it is, people seem to enjoy developing in the Atmosphere in a way I never saw on AP.

The knot is the bit that causes two canids to get en-tangled after getting frisky.

Furry developers are all professionals and won't have a giggle fit every time they think about it.

I assume you don't mean Tangled is an expert martial artist. Can you translate this to not-a-dev-but-uses-git?

Blender has an actual spreadsheet view. Everything model-related can be quantified in it.

https://docs.blender.org/manual/en/latest/editors/spreadshee...

People surprised by Anthropic getting in on Blender funding obviously never saw any of the Blender/ChatGPT integrations a few years ago. This has been coming for a long time.

https://www.youtube.com/watch?v=xhN7P7ENu4g

That was ancient times in LLM terms. I've seen demos that create whole scenes in a single prompt.

There are lots of non-coding use cases for LLMs that don't burn through compute but are still useful. Anthropic is starting to catch on, and it makes sense to focus there with the compute crunch.

Art generators need to come a long way to completely replace art tools. I dabble, but if I were doing real work with it, there have been times it would have been faster to composite in a 3D model rather than keep trying to prompt an image generator into fixing something.

I even made a better map in a few minutes to prove the point: https://bsky.app/profile/kyefox.com/post/3mkibnvrt3c25

There are still inaccuracies, but I'm also not pretending to care about legacy like Moleskine.

It's incorrect because it was lazily AI-generated. Most modern AI image generators can handle this if you're the least bit thoughtful.

Look at the names on the map. You don't need to be a LOTR superfan to tell something is off.

Furtrack seems to suffer from the same problem a lot of photo sharing sites do: the only people who visit are the subjects and other photographers.

For example: https://www.furtrack.com/index/species:fennec_fox

Tens, hundreds of photos for each suit, but almost no views.

Finally, competition for Clickhole.

That's probably what I was thinking of. I haven't kept up as much on non-text generative AI.

You're thinking of a feed trough, like for pigs. This use of feed comes from news services.

My understanding is music generation is more like stable diffusion. It generates a waveform as an image, then turns it into an audio file.

Consider Trilium if the collaboration stuff people use Notion for isn't important. It's open source, uses SQLite, and does automatic daily and weekly backups.

https://triliumnotes.org/

Opus 4.7 would come out the day before my paid plan ends.

Firefox started at 0% when IE was more dominant than Chrome is today. Nothing is certain.

Classic "is this a feature or a product?" problem. You're going to have a bad time if you spend all your effort on a feature and nothing to set it apart.

404 for me

I think it also points to the problem of implicit assumptions. Fish is meat, right? Except for historical reasons, the grocery store's marketing says "Fish & Meat."

And then there's nut meats. Coconut meat. All the kinds of meat from before meat meant the stuff in animals. The meat of the problem. Meat and potatoes issues.

If you asked that question before I'd picked up those implicit assumptions, or if I never did, I would have to guess.

Was this one of the trips on the good-as-new plane? The video was an interesting watch.

Modern LLMs do a great job of following instructions, especially when it comes to conflict between instructions from the prompter and attempts to hijack it in retrieval. Claude's models will even call out prompt injection attempts.

Right up until it bumps into the context window and compacts. Then it's up to how well the interface manages carrying important context through compaction.

My threat model here focuses on what the provider gets out of the free tier. Cloudflare gets a broad view into activity on the internet for building the models they use for their paid offerings. Free Gmail puts people on a path in to Google's ecosystem with basically zero marginal cost.

They might not enforce it consistently, but they do bother themselves with it enough to have guidelines.

https://www.ftc.gov/business-guidance/resources/disclosures-...

Have a contract that encodes "fuck you, pay me" into the terms. Ideally, have an actual lawyer take care of the contract and the enforcement. There's a lot of law-y stuff out there that won't hold up in reality. Mortgage companies don't take payment in excuses from your clients, so neither should you.

The D5 doesn't have built in GPS, and adding it requires an attachment. I don't know if the smartphone app works on that model, but it is from the same year as my D5600 which does support it. The app provides GPS but also drains the battery fast. I turned airplane mode on after the first dead battery.

GPS might work out there though: https://www.nasa.gov/directorates/somd/space-communications-...

I was under the impression ML-driven trading was already widespread.

Variety! I appreciate that it's not all tech writing from tech blogs from people in tech like almost every blog list/aggregator thing on HN.

I haven't tried a local model in a while. I can only fit E4B in VRAM (8GB), but it's good enough that I can see it replacing Claude.ai for some things.