I just put up a new page: the Daily Support Goal. It's the (very, very) long version of the little bar on the news page, and it's the honest account of what it costs to run Subeta in 2026 and how close to the line we actually are. If you've ever wondered how a site like this keeps the lights on, that page is me answering it. Please read it!
And to support the goals on the page, two updates went up alongside it. Cash Shop Collections has its own page now, so you can browse every monthly collection going back to 2006 and either buy the sealed set or pick the one item you want out of it. For this week only you can browse and choose from every week, when the sale is over it goes back to being just the ones released in the current month.
And Billboards got rebuilt with a banner maker, so you can design your very own drag and drop 480x90 ad right in the browser instead of uploading one. Please, make something weird and fun!
For one week, both are on sale. Every collection is open right now, all the way back to 2006 👨🦳, and all but June 2026 are 100 CSC each. Billboards are 75% off on every size. The sale ends Sunday, and the prices and the collection selection go back on their own.
If the goal page makes the case to you, the sale is the easy way to help but there are lots of other suggestions on the page. Bump your old bug reports (they're getting fixed!) and make those suggestion threads.
me! in 2026, still clutching that Kumos plushie 💗
And yes, Masquerade is this month - more toward the middle. Hopefully you understand from the page why dates have been more flexible the last few years.
This is probably the neatest page on the site. I really like transparency like this.
An aside, it's pretty annoying how machine learning has been folded into genAI in the public eye...they are related technologies but the former uses constrained datasets for a specific purpose while the latter hoovers up everything to try to do everything, which is where all of the plagiarism and godawful energy and environmental use comes from. Machine learning (and related technologies such as machine vision) have existed for decades, I literally have my name on a scientific paper using it from 2013, and if you own a car with adaptive cruise control, it's using machine vision with a camera, trained on images of cars, obstacles, and pedestrians (and, if your manufacturer is responsible, alongside a radar for extra safety). I don't think anyone would argue that your Honda using ACC is the same as ChatGPT telling someone to kill themselves. It's been used for years in medical imaging and protein folding etc etc, but everyone is calling it "AI" now because it gets eyeballs and makes investors excited. The ethical problems with genAI are the aforementioned environmental usage and plagiarism aspects; I feel that the technology as implemented is unethical but the actual underpinnings of it have existed without harming anyone for years and years, and those ethical aspects are satisified when it is used as originally intended: on limited datasets for a very specific purpose.