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.
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
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
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.
Honest question, maybe a blind spot of mine. Touch typing is so integrated into my daily experience it feels like driving or riding a bike. I mostly learned to touch type in the 90s just chatting with friends on AOL instant messenger. I think of touch typing as something nearly everyone picks up just as a side effect of living with computers.
Even in previous generations, most self-taught people get fast at hunt&peck rather than learning proper touch-typing. It is not a natural skill in any way, you need a conscious effort to stop looking and to limit your main fingers wandering.
I generally tried to keep my kids away from excessive screen usage, but I motivated them to touch-type anyway, because I always wished I'd learned it earlier than I did (in my early 30s). I see them reaping benefits already in their teenage years, knocking out school assignments very quickly and being able to focus on the content more than the typing.
I'm also confused by this. I taught myself touch typing in the 90's. I also had a required semester-long class that covered only typing my freshman year of high school (1999). Neither of my parents learned it, but I figured everyone younger than me knew how. Pretty shocking to find out that's not the case.
I can't imagine not being able to touch-type. It's such second-nature that I can hold a conversation with someone while typing out separate thoughts I'm having about the conversation on a keyboard.
The average American types around 40wpm, so definitely not touch typing. People definitely get by without learning it.
I work in a huge variety of fields and interact with people from all places in US society. My guess would be maybe 25-35% of people I've worked with use touch typing. Everyone chicken pecks.
Most people use phones nowadays and rarely use a physical keyboard. It just isn't that important to most people. They can get by without it.
It's pretty difficult to pick it up naturally when you only use a touchscreen and never a keyboard, since there aren't any physical keys to stabilize your hand position. It's becoming more common for people to only use their phones or tablet and not a desktop or laptop.
Yeah this is mind boggling to me as a millennial. I didn't set out to learn touch typing either. Hell, my sister who isn't a techie learned to do it just be spending all afternoon on LiveJournal and AIM chat. I don't understand how one could be an avant reader of hn and be interested in an article about this like... you don't? You can't? Whaaaa?
Honestly I'm consistently surprised - I've worked at Amazon and seen many engineers, product people, etc type with incorrect techniques.
I've seen interns looking for symbols on their keyboard for a second or two (the tilde "~" or the pipe symbols "|") when I asked them to type in a certain shell command.
Since I started building this website, many of my friends and family learned touch typing because of the site never even heard of proper touch typing technique until I started talking about what I was working on.
I think it's due to poor education - there's no institutionalized course that teaches this. A couple schools maybe, but nothing on a big scale.
Kind of mind boggling given that almost every desk job uses a keyboard
I think with proper cycling posture you can take the pressure off of your wrists.
Use your back and shoulder muscles to support your upper body and arms, so that you can ride with elbows bent. If you can bend your elbows, you can avoid pushing on the handlebars for support. That will make for more comfortable wrists with or without a keyboard.
Funny you referenced Sheldon.
He's like Knuth of cycling.
The industry will keep shoving down your throat their new "advancements" every year, but at the end of the day, bikes hasn't changed a lot, and the common knowledge will remain the same.
I don’t know that I completely agree. Yes, many elements are similar, but a carbon road bike with disc brakes, electronic shifting, deep wheels, radar, a computer and tubeless 28mm or 30mm tyres is an actual wonder to ride. I’m justifying a truly silly spend, but when I get on an old bike or a rental, it feels like stepping back in time.
Yeah, I have an old Panasonic Sport that I'll take out with some older fellas who live near me in rural Kentucky. They have nice carbon bikes with all the bells and whistles. It's night and day how much harder I have to work to keep up with them
Unsurprisingly, core strength is a big part of cycling. It's a good test to lean down in a more aero position and lift your hands slightly off the bars and see how long you can hold it.
so less pressure on your wrists while typing would be quite the core workout over an extended period of time.
Yeah I learned this lesson the hard way when after a century ride, my left hand ring finger and pinky went numb for almost two days. Started making sure I had my elbows bent and used my core to keep myself more upright.
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.