Daily Support Goal
This page is the longer version of the little bar on the right of the news page. It explains what the daily goal is, what it actually pays for, where the trend is going, what I am already doing about it, and what would help most.
A note from Keith
Subeta has been running for 22 years. The team is a ragtag set of illustrators, user administrators, content managers, a business manager (our COO, who keeps the rest of us honest), and me, the only programmer, and the person who owns the LLC. Most have been here a decade or more. We haven't had an open position in ages, which is mostly a good thing and a little bit a bad thing, because it means we have not grown.
I should be straight about my part of this. Subeta is not my day job. For the last ten years I've worked in civic tech, helping governments deliver services to people who deserve better ones. Subeta is my hobby. I haven't taken a paycheck from it in years. (Timeline here, if you want it.) I'm not trying to grow Subeta into a venture-backed anything. I'm trying to keep a clubhouse open, and to keep paying the people who help me keep it open.
Here is the part I have been avoiding writing for a while. We are not "sleep better with a bigger reserve" thin. We are closer to the line than that. Cash shop revenue has been trending down month over month for a long time. Active users have been dwindling alongside it. The Daily Goal above is the honest number: what we need to bring in each day from the cash shop and subscriptions to cover what running Subeta actually costs. Today we are at 12%. Last 30 days we hit it 5 times out of 29. The trend is going the wrong way, and it has been going the wrong way long enough that I have stopped telling myself it will work itself out.
Two months ago we canceled Project Kumos, the ground-up Laravel rewrite we'd been chasing in one form or another since 2022. I learned more from canceling it than I did from doing it. Every day since, I've been investing directly in the legacy site, because that's where Subeta actually is, and where the next 20 years are going to be built. The team and I have shipped more in the two months since canceling Kumos than in the last two years of trying to chase Kumos. There is a list further down this page. It is longer than I expected.
Some of what is on that list: a complete redesign of user hovers, a redesigned pet spotlight with community moderation, two new ways to dress and decorate your avatar (Wardrobe Contests and Wardrobe Effects), a Gardens overhaul, an Item Directory rebuild, cosmetic prizes drawn from the user-submitted Studio, the first major change to the battle system in ten years, a mobile beta, and a lot of plumbing. Two months. One programmer.
We have plot ideas we can't afford to ship right now, and we'd love to ship them. The difference between a thin month and a full one is the difference between keeping the lights on and being able to make something new again.
I'll keep going as long as you all keep showing up. Thanks for being here.
The trend, year by year
One bar per year. Each bar is sized relative to our best year, which was 2017. Specific dollar amounts intentionally left off, but the shape of it is the story.
- 2008
- 2009
- 2010
- 2011
- 2012
- 2013
- 2014
- 2015
- 2016
- 2017 peak
- 2018
- 2019
- 2020
- 2021
- 2022
- 2023
- 2024
- 2025
- 2026 YTD
The line went up steadily until 2017, plateaued, and has been drifting down ever since. The drop you see in 2026 is partly that it's only five months in, and partly the same trend.
Where the money goes
Rough breakdown of where the money goes, averaged across the last year. These aren't exact numbers, but they're honest. If something here surprises you, it's probably "lawyers and accountants."
- Team wages 52%
- Mostly part-time. Seven illustrators, four user administrators, one content manager, and one business manager who keeps the rest of us honest. I'm the programmer, and I don't draw a wage (more on that in the note above). Everyone else is paid a living wage for their hours. The team has barely changed in a decade. An open position on this list would be a real event.
- Servers, CDN, email 16%
- Bigger than you'd expect. The image CDN serves around 30 million pet and item sprite requests a day. Email deliverability is its own line item now thanks to Gmail's 2024 sender rules.
- AI scraper load and defense 6%
- New line item, and growing. LLM training crawlers hammer the site disguised as regular browsers, rotating through millions of residential IP addresses to slip past rate limits. They eat bandwidth, CPU, and a lot of my time writing the heuristics to keep them off. They can spike a sleepy Tuesday into a half-down site. Other small communities are quietly folding under it. Weird Gloop wrote up the gory details if you want a longer read.
- Legal, accounting, tax 12%
- Required when you are an LLC accepting payments in many jurisdictions. The line item that used to be "an accountant and a year-end tax return" now also includes ongoing review of age-verification and ID laws as they pass in different countries and states (more on that just below).
- Payment processing 8%
- Stripe and PayPal fees on cash shop purchases. The price you see is what we charge. The fee comes out of our side.
- Tools, software, monitoring 4%
- Source hosting, google accounts, error tracking, staff chat system, the boring stuff the site runs on and we have to pay for monthly.
- Reserve, domains, the rainy day 2%
- Domain renewals (we own six), year-end tax bill, and the buffer that keeps the lights on if a month goes sideways.
Why we are talking about this in 2026
We are not a nonprofit. We are a small LLC. We are also not "raising a round" or "exploring strategic options." We have run on cash-shop revenue and subscriptions since 2004, and that is the model we plan to die with.
The reason we are talking about this directly, in 2026, is that the small, independent web is genuinely harder to run than it used to be. Three of the things that are making it harder are below. None of them is unique to us. All of them cost money the cash shop used to comfortably cover.
Cohost is the recent one most of you know
Cohost shut down in September 2024. Tens of thousands of regular users, a thoughtful team, an anti-big-tech manifesto, doing everything the way you'd want a small independent site to do it. They still ran out of money. Their stated reasons were "lack of funding and burnout." Both of those things are recognizable from this side of the screen.
A 250,000-user forum shut down rather than comply
The UK Online Safety Act took effect in March 2025. It covers any user-to-user service with "links to the UK," it has no exemption for small non-commercial sites, and it makes the site operator personally liable for what users post. Dee Kitchen, who had run the LFGSS London cycling forum and the wider Microcosm platform (around 300 sub-communities, roughly 250,000 monthly users) alone for nearly three decades, announced she would delete the servers. Her line: "The risk to me personally is too high." A handful of other small UK forums went with her, from a hamster-care board to a local Cotswolds community site.
Age-verification laws in roughly two dozen US states and parallel bills in the EU and Australia do the same thing at smaller scale: per-jurisdiction rules, geolocating every visitor, and rebuilding the flows when the next bill passes or is challenged in court. Real engineering work, real legal work, and a real chunk of what's now in the legal-and-accounting line above.
The bill for keeping bots off the site keeps going up
LLM training crawlers are now a meaningful chunk of internet traffic, and they hit the small, public, open parts of the web hardest. GNOME's GitLab measured about 97% of incoming requests as bots. Wikimedia reports bots driving about 65% of its most expensive traffic and put it plainly: "Our content is free, our infrastructure is not." SourceHut's maintainer wrote that he was spending "20-100% of any given week mitigating hyper-aggressive LLM crawlers." Read the Docs cut its bandwidth bill in half by blocking AI bots, one of which had downloaded 73 terabytes of their content in a single month.
We are running the same kinds of mitigations they are, and paying the same kind of bill. The Weird Gloop writeup below is the closest mirror to our own situation.
Subeta has more users than Cohost did, a 22-year cushion, and a model that has worked for two decades. None of that makes us immune. The trend lines for the last year have not been kind, which is why we are saying this directly now, while there is plenty of room to fix it, rather than later, when there isn't.
Building the next 20 years
Worth being clear about this: the money is not only going to keeping the lights on. The two months since we canceled Kumos have been the largest single technical investment in Subeta in years, most of it under the floorboards, most of it the unsexy work that makes a 22-year-old codebase last another 20.
The dependency cull
Every PHP package on the site is now on a modern (or the newest) version. Months of work, mostly silent, exactly the kind that keeps a small site from waking up one Sunday to a critical CVE. We scrapped tens of thousands of lines of CSS along with it, including two entire versions of Bootstrap that were both being loaded at once for legacy reasons. The .card class was being touched by five cascading design systems stacked on top of each other. It is now touched by one. Each removed dependency is a potential security hole gone and one less thing anyone has to reason about the next time something has to change.
Every action, finally logged
The site now writes a real audit trail of who did what: items moved, currency spent, quests turned in, battles fought, comments posted. When a user opens a ticket about a missing item or a weird interaction, the admin who picks it up can pull a clean timeline instead of guessing. The user administrators have been asking for this for years. It also feeds future work: better fraud detection, smarter mod tools, and the kind of activity feeds that used to take a week of staff DM-archaeology to put together.
The admin and ticket tools, finally usable
The whole admin panel got rebuilt around what the team actually does day to day, rather than the 2008-era forms they were working through. The ticket system backend was overhauled at the same time, which means mods are answering faster, with more context per ticket, with less manual lookup. None of this is user-facing, but every minute a UA isn't fighting their own tools is a minute they get to spend helping people.
The battle system, finally
I just re-launched the perk system as Battle Spheres, the first real change to battles in ten years. Beyond shipping new content, the part that matters is this: I can now watch battles in flight and pull real data on what is working and what isn't. We have never had that. We have been making balance changes on intuition for a decade. Now we can make them on what people actually do.
The first tests Subeta has ever had
While we were rebuilding all of this I wrote the first automated tests in the site's history. There are now over 400 of them, covering the things that should never break: moving an item from your inventory to your vault, the math behind battle damage, currency transactions, login flows. Every one of them runs on every change. The site is significantly more secure than it has ever been, and the gap is still widening.
The messy part, honestly
You have probably noticed pages looking wonky for a day or two, or a feature changing twice before it settles. The new shop search thread is a fair example: I took a swing based on ten years of accumulated feedback, the data we now have, and my own gut, and some of it landed and some of it had to be walked back. That is going to keep happening. It is hard to know what to ignore in ten years of feedback. I am going to get things wrong. The upside of being one person on the code is that I am also very willing to change it back, fast, when I do.
What's been shipping, since we canceled Kumos
Two months of work, in roughly the order it shipped. None of this would have happened if we were still trying to build a parallel Subeta in a different framework.
Pet Spotlight, now community-moderated
User hovers, completely redesigned
Wardrobe Contests
Wardrobe Effects
Item Directory, rebuilt
Gardens 2.0
Studio art powers cosmetic prizes
Battle Spheres
Mobile mode
The fix spree
What's coming up next
Not promises. Things I'm actively working on or starting on. Dates are not yet a thing.
Pet Profiles V2 (opt-in)
User Profiles V3 (also opt-in)
More battle
Quest refresh
Each of these has had the benefit of being started up at some point in the past, and now we can build on top of that work.
What you can do, in plain order
These are listed in the order that helps us most, but anything on the list helps. Posting a bug helps. Bumping an old thread helps. Telling a friend who used to play helps a lot. The single biggest thing that gets us through a thin month is people deciding to renew their subscription one month earlier than they would have.
-
01
Go to cash shop →
Support directly via the cash shop
Every cash shop purchase counts toward today's goal automatically. Items are fun. Buying them keeps the lights on.
-
02
Subscriptions →
Subscribe, or renew your subscription
Monthly recurring support, which gives us a much greater ability to budget. Renewing one month earlier than you would have makes more difference than you'd think.
-
03
Bugs forum →
Post the bug you've been quietly hating
We are on a fixing spree this quarter. If a thing has bugged you since 2019 and you assumed nobody would ever fix it, post it now. We are reading. Bumping old threads is encouraged.
-
04
Suggestions forum →
Post the feature you've wanted
The suggestions forum has been a graveyard for a decade and we owe it an apology. We are reading it again, sorting by recent activity. If your idea got ignored, bump it.
-
05
Feedback forum →
Add your weight to the quality-of-life threads
QoL threads are the ones where everyone agrees something is annoying and then nothing happens. Pick one. Reply with what you would change. That signal is now actually being used.
-
06
Preferences →
Try the new mobile mode (beta)
A real mobile layout, the first one we are not embarrassed by, is in beta right now. Opt in from your preferences. Feedback goes in the mobile feedback thread.
The team I'm trying to keep paying
A few of the people behind everything above. Hover for the usual card.
