· Eric Koslow
Re:VStream
Explaining what has led to VStream's revival and where we're going from here
VStream's announcement that it would be coming back has generated a lot of reactions. A lot of them have been very positive (thank you for those!), but it has generated a lot of questions and concerns as well. It's not every day you see a service like VStream come back from the dead like this. I wanted to take the time to explain what has led to this point, how we got here, and what's the plan moving forward. Spoilers, things will be different this time around.
VStream's history
I originally had an admittedly obnoxiously long history section here detailing the original rise and fall of VStream. But I realized that most people coming to this article care a lot more about the future of VStream than its past. You can still read the article on this blog here: A not so brief history of VStream. But unless you really care for some reason, feel free to skip it.
If you've never heard of VStream before but somehow are reading this article, the tl;dr is VStream was a small live streaming and video hosting site dedicated to VTubers and their communities. It was started by me in May 2022 and it shut down in April 2024 after running out of funding. Some people really liked it, some people hated it, and most people never tried it.
Why relaunch VStream?
Other than shock, this is probably the first reaction everyone has when they hear that VStream is coming back. And it's a good question. Does the world need another streaming site? Won't it fail again the exact same way? The honest answer is: I don't know. What happened is that a friend linked me to Hazumi Aileen's tweet about VStream and I saw a bunch of replies to the post saying that they missed VStream. I remembered how painful it was to shut down the service and how I felt like I betrayed the community I cared for. Many of our old community members felt like VStream had become their home and safe space. When the service was shut down, that home and community was taken from them. I realized that I had the free time, the skills to bring it back, the rights to the IP, and most importantly the old code still sitting around on my hard drive.
The plan was to rebuild VStream, but this time architect everything for keeping costs as low as possible. Originally, VStream was designed with a "billion dollar outcome or bust" mentality. We operated out of four datacenters across the world and duplicated our data across all four to keep latency as low as possible. We used databases that could scale horizontally across the globe to handle massive amounts of traffic. We even wrote and operated our own transcoding servers to handle many thousands of streams at once. And did a bunch of other dumb things like that to design ourselves to handle the traffic that never came. This time VStream would be built on simpler stuff. One datacenter on the West Coast, a couple web servers, a SQL DB, and rely on third-party on-demand transcoding and video hosting. I also realized I could do this all myself this time around (more on this below). For comparison, by the end of VStream we were spending close to $200k/month just in salaries (the highest source of spend by far even with all the datacenters and servers).
I knew a relaunched VStream would probably never turn a profit. But if I could keep costs under $2-5k a month I would be happy to maintain and support the site for those old creators and their communities. Plus anyone else who doesn't like the status quo with Twitch and YouTube and doesn't mind a smaller, more intimate community. Also, I'm a builder at heart. There were so many things that I didn't have a chance to build during the original run of VStream that I've been thinking about for years. I like the community, I needed something to tinker on, and I felt like it wouldn't bankrupt me. So why not?
What will be different this time around?
First off, the mentality, goals, and pressures put on VStream. Before, VStream was a venture capital backed startup. I was pitching to investors that VStream could represent the future of online entertainment and that it could be worth a billion dollars or more one day. That goal changed how we operated the company. The EV math said it was fine to blow millions of dollars on expensive San Francisco engineering salaries if it meant there was a tiny chance of a massive outcome. This time, the assumption is that VStream will never make it "big" or even ever be profitable. I'm applying a fun project mentality to VStream rather than a startup one. If by some strange miracle VStream somehow gets widely popular this time around I'll revisit this policy. But 99% chance that's not going to happen.
Secondly, who VStream is for will be changing. One of the chief objections about VStream I kept hearing last time is that, "A site only for VTubers will never work. As an audience member you often watch a mix of IRL streamers and VTubers. Having to go to one site for one of those things and a different one for another is too big of a pain." I personally don't know if that's true, but what I will say is that when I look back at all of the things we built for VStream, 5% of our features were things that made life better for VTubers. But 95% of the features we built just created a better platform for creators and fans. So, while we never had a hard-rule against non-VTuber creators in the past, this time around VStream will be explicitly open to all types of creators. Anyone who doesn't think the status quo is working for them is welcome to use VStream. This doesn't mean I'm abandoning the VTuber community, far from it. But I do want to open the aperture to anyone that feels like VStream could be a good place for them and their communities.
Thirdly, there will unfortunately be more cost-saving measures in place this time around. Currently, the one big one is that VODs will now automatically be deleted after 90 days. You'll be able to download them to upload to a site like YouTube for long term preservation, but this time I'll be trying to keep our long-term costs down as much as possible.
AI and VStream
Development
To be very upfront and transparent, yes this rebuild of VStream was done with the help of agentic (aka AI) development. To be honest, without AI VStream would have no ability to come back today. Even with the old code around, last time it took half a dozen engineers with full-time positions to maintain and build out VStream. As explained, that's just not sustainable for a niche product like ours. But with coding agents I was able to rebuild VStream all by myself.
If that by itself is a deal-breaker for you, I completely understand. If you can believe it, I too have deep reservations about AI. Feel free to stop here and never use or think about VStream ever again. But if you're down to keep reading or would like an explanation here it is:
Back in 2024 and for much of 2025 I was a huge AI hater. While friends of mine in San Francisco were using AI assisted coding tools I stubbornly refused to. Not only did I dislike that they were using a ton of energy, water, and increasing the cost of memory and CPUs, they also just produced crap code in my opinion. Every time I would try one out it would make poor decisions, produce unreadable slop, and ultimately I felt like I had to spend more time fixing its mistakes than it was saving me on just typing the correct thing from the beginning.
But my feelings really started to change with the release of Claude Opus 4.5 in November 2025. That was the first time I felt, "Wow, this code actually isn't the worst thing ever. I may be able to use this." From there I used agentic coding more and more. And as the models have continued to improve I've found my need to hand-write code has decreased along the way. You'll hear this same story over and over again from other professional programmers. And as much as it hurts me as a programmer of 20 years, I am way more productive today using agentic coding than I ever was. So while I'm still extremely anti-AI when it comes to anything related to art or replacing artists, I can't deny that it's become integral to my workflow as a developer.
Usage in VStream itself
Currently the VStream product doesn't use any AI. I have created some private MCP endpoints so that my coding agents can pull statistics like the number of current streams and the load on the servers on my behalf to make running the site easier for me. But that's it. No chat messages, streams, videos, posts, or anything else are sent to AI for processing or moderation. That said, I'm not going to commit to continuing to keep it that way. If I think the community would be safer or better served by introducing AI somewhere I'm going to do that. For example, the automatic detection of child abuse material being uploaded to our servers. But here are some things I will commit to:
VStream will never train AI on anything you upload to VStream or sell your data to anyone who would (or would ever sell user data in general)
VStream will never use an AI provider that trains on any data we send it
No moderation decision will ever be made by an AI. That will always come from human judgment
What's next
If you read the above and are comfortable with the changes that are coming to VStream and how VStream is actively developed, great! There's a rough timeline of things to come in the FAQ below, but the goal is to get everything live by end of the year or early January. For now, please follow us on Twitter (X) or Bluesky for updates. I'm hoping to have our first Q&A live stream on VStream after American Thanksgiving in December. I'll also try to write a lot more blog posts from here to continue to further explain my thinking about VStream and its future. I much prefer a long-form medium like this compared to a bunch of 280-character messages on places like Twitter. So I apologize if I don't respond much there.
FAQ
What's the timeline from here?
Nov 22nd - Dec 11th: I'll be spending time with family and friends for American Thanksgiving. Knowing me, I'll be doing some work while I'm away but I won't be able to schedule any streams
Weekend of Dec 12th: Hopefully the very first stream using the new tech. I'll be hosting a live Q&A
Throughout December: I'll be inviting some beta users to try out streaming on VStream.com
Late Dec or early January: Open up the platform to everyone!
Should I move to streaming exclusively to VStream?
Depends on what your goals are. If you're comfortable with a smaller audience you're more than welcome to! But if your goals are to become a full-time content creator or have content creation be a real source of income for yourself you're better off with Twitch or YouTube still. That said, you're always welcome to multi-stream to VStream if you'd like.
Will I be able to multi-stream to VStream?
Yes! (see above) In fact that's how I assume most people will try out VStream.
Does this mean VStream was vibe-coded?
I personally read every single line of code that my agents produce. I don't merge anything until it goes through human review. And I'm still catching mistakes the AIs are making! I personally don't think we've gotten to the place where you could just vibe-code a product as complicated as VStream yet. It still takes years of experience to direct the agents properly to get good results and to catch mistakes before they happen. But if you think any usage of AI development means something is vibe-coded then we're just going to have to agree to disagree on the definition.
I'm still worried about the stability of VStream. Doesn't it cost a ton of money to run a streaming site?
You're right to be worried. As far as I'm aware, even Twitch still struggles to be profitable within Amazon. And while it takes a lot of money to run a streaming service at Twitch's scale, it's not too bad when there's only a small community using it. I'm going to publish another article that breaks down all of the upcoming costs to see if it alleviates anyone's fears.
Who will be paying for VStream's server costs?
To be explicit, I'll be paying for everything out of pocket myself. I'm not be seeking any outside funding so everything will be 100% owned and operated by me.