If you're training this on your own timing patterns, do you worry that eventually captchas will pick up on this and you yourself will no longer be able to prove you are a human?

Not particularly. I'm not yet convinced people's mouse movements are unique enough to our identity that they're useful as a fingerprint, whereas it's very easy to classify whether something looks bezier or looks human.

Eventually I'm hoping to collect enough data here to train a biased decoding model, so you could input some randomized personality vector (which implicitly encodes slow movement, jerky motion, trackpad, mouse, etc) and have that impact the RNN generation. So in theory there would be infinite combinations from the larger subspace we're sampling from.

I think the way to do it is to think of this as your own browser that can also be used by agents (with granular permissions). I use the browser for 5h today and my patterns then inform another 12h of agent use

Try from inside the signal app itself instead of system settings? On android Signal has an option at hamburger menu > Settings > Notifications > Notifications (toggle switch)

Oh... hmm, two toggles actually. One at Settings > Notifications > Calls > Notifications toggle, and the other at Settings > Notifications > Messages > Notifications toggle

Self-describing API endpoints... is the server side for this basically just HATEOAS?

Or SOAP with WSDL.

> Published in Proceedings of the Royal Society Bee

I think the core idea in "masking" is to provide adjacent pixel art tiles as part of the input when rendering a new tile from photo reference. So part of the input is literal boundary conditions on the output for the new tile.

Reference image from the article: https://cannoneyed.com/img/projects/isometric-nyc/training_d...

You have to zoom in, but here the inputs on the left are mixed pixel art / photo textures. The outputs on the right are seamless pixel art.

Later on he talks about 2x2 squares of four tiles each as input and having trouble automating input selection to avoid seams. So with his 512x512 tiles, he's actually sending in 1024x1024 inputs. You can avoid seams if every new tile can "see" all its already-generated neighbors.

You get a seam if you generate a new tile next to an old tile but that old tile is not input to the infill agorithm. The new tile can't see that boundary, and the style will probably not match.

That’s exactly right - the fine tuned Qwen model was able to generate seamless pixels most of the time, but you can find lots of places around the map where it failed.

More interestingly, not even the biggest smartest image models can tell if a seam exists or not (likely due to the way they represent image tokens internally)

I'm curious why you didn't do something like generate new tiles one at a time, but just expand the input area on the sides with already-generated neighbors. Looks like your infill model doesn't really care about tile sizes, and I doubt it really needs full adjacent tiles to match style. Why 2x2 tile inputs rather than say... generate new tiles one at a time, but add 50px of bordering tile on each side that already has a pixel art neighbor?

Yeah I actually did that quite a bit too. I didn't want to get too bogged down in the nitty gritty of the tiling algorithm because it's actually quite difficult to communicate via writing (which probably contributed to it being hard to get AI to implement).

The issue is that the overall style was not consistent from tile to tile, so you'd see some drift, particularly in the color - and you can see it in quite a few places on the map because of this.

Oh that makes sense, thanks for explaining! And thanks for sharing your process and result! Interesting to see your process, and looking at the map really tickles my nostalgia

Have you tried restraining the color palette by post-processing?

There would have to be some tiles which don't have all four neighbors generated yet.

Did you ever consider using something like https://github.com/jenissimo/unfake.js/ in your process, to make it more proper-pixel-art?

Maybe to process the Nano-Banana generated dataset before fine-tuning, and then also to fix the generated Qwen output?

Thanks for this - I was confused as well. Makes perfect sense now.

I don't know why California's electricity costs so much, but the gas prices are high due to regulation distorting the market. California has special California gas produced only at in-state refineries. It's for a good cause--California's gas, "CARB gas" is cleaner. But the gas market in California is segregated from the wider US market

> When cats walk or climb on your keyboard, they can enter random commands and data, damage your files, and even crash your computer.

And they might turn you into the Freakazoid

That you Dexter?

From the title I thought this was going to be a variation on the bear and swimmer puzzle.

https://www.quantamagazine.org/can-math-help-you-escape-a-hu...

Import one from Australia or the UK? Someplace where they drive on the left?

Or just learn not to be a lefty! So easy.

This sounds like a footgun.

At some point some application developer will introduce a bug where they're not sending utc.

Without the time zone, the wrong times will end up in the database, bad data mixed in with good. This will be a nightmare to fix.

With the time zone, I don't think this class of bugs is possible. Most application developers will wrap the time in an object that allows them to do time operations without needing to directly handle time zone in most cases.