I've built an i18n system that I started coding last year. Engineering it was cool : it's an i18n systems with no keys (because I hate looking at my codebase with keys everywhere) and designed in a way that, by design, there can not be any hole in translation. Of course I use it for all my projects, I start i18n from scratch now whereas before I would wait the last moment, or I would leave it in English, which is too bad because i18n is a growth boost. Anyway, I liked it so much that I thought I would make a Saas out of it. A little subscription, very cheap because it costs me almost nothing. I thought there was a market because i18n was such a pita to setup, and i18n Saas are insanely expensive. Then 2026 arrived and no developer code anymore. Every dev build their own Saas replacement. I'm a vibe coder myself too now, like everyone else, so I took the opportunity to port open sourced implementation libraries from node+react to all kinda languages and frameworks. But vibe coding makes i18n not the same pain as before, it has just become a "setup i18n in this codebase" prompt. So maybe I don't have a product anymore. Still, my engineering is also a better solution for agents than the old key system, where you have missing translations, really bad UX with keys rendered on clients, sometimes crazy ci pipelines to get the system working. And UGC is not even covered by traditional systems. So my engineering does have some value: either you can take 5 minutes and use my system for which I've put some engineering thought (and many, many tokens already), either your agent will have to do the same, and experience bugs throughout the iterations, too bad (and many, many tokens+hours to spend). So I have something to share to the world. Not the devs world, but the agents world: hard to sell anything to developers now, better sell to agents. I setup my whole project towards this outcome, and at some point, I had to make my backend open sourced too so that it could be seriously audited by agents. But then... subscription model makes no sense for open source, why wouldn't you steal the whole system from me instead, and self-host your own free version of my software? Sounds a better idea at vibe coding age, cheaper in tokens. Good idea, but my business model is dead now: I've spent days engineering a very good and useful system, which is a good enough reason to try to sell it, I've made it open source for getting agents green light, so any agent can steal it (humans were doing it also). How to make money? First I thought I would licence the self-hosted version, with one project free and the other gated, a bit like https://eggerapps.at/postico2/ is doing, but I thought meah, people will still the code and hack the licence stuff. Then I thought about Sentry, because I'm actually kinda like them now. A bit smaller though, yet. And I decided to go the following way: make everything open source and free, as long as it is self-hosted and you provide your own AI translation API key. Push for donations like https://cyberduck.io/ did with me for years (and I finally donated). Do marketing around that solution, and maybe because it's free and MUCH better than existing solutions, people will use it. Maybe they will donate? let's see. But you know, it still requires some effort from the developer to self host the solution, you need to deploy a VPS and one or two commands (I use https://once.com system so it really is that easy). So maybe lazy devs would prefer a cheap subscription more than hosting their own stuff? So here are my expectations: as it is free, if I do good marketing, devs (50 millions of them) will get interested, they will make their agent use it on their own VPS, and it will become notorious. Afterwards if my software is that good, the part of the people who are lazy will pay for it through subscriptions instead of self hosting, like people do for Sentry. What do you think?