I am currently in the process of launching my AI teams platform that I've been working on since at least January. It's https://PersonaStack.ai. I'm doing it without VC money and all by myself. I've used over 110B tokens so far building it.
You get some amazing results with teams of AIs if you do it right. The key is to control behavior with what integrations and responsibilities each agent has. That way they naturally adapt, delegate, fact check each other, and generally act more autonomously.
This is already running the automated news site ainews.personastack.ai complete with social media posts 100% automated.
It also runs the issue triage, coding, reviews, and releases for the Kuberhealthy open source CNCF project, which is another thing of mine.
I don't think the next step is really smarter models. It's how we make the models more effective, and teams, when done right, net the best results I've seen.
Hoping to get noticed here soon, but it's extremely hard to do solo I'm finding.
One test stack on the platform been running entirely automated for months now to research, author, code, ship, and promote https//ainews.personastack.ai completely on its own. Many more things are possible like automatically identifying fixing, and shipping in response to 500 errors, personal assistants, or whatever you can come up with. I'm working on moving both my other open source projects over to be run by AI teams right now.
I used Codex mostly, but my CLI Codex would run out of quota after an hour or so because I only have the OpenAI Plus plan.
That said, when I used the agentic version on the website, the quota currently seems to be infinite. I could get 10+ agents running and work for hours without a mention of quota limits. I'm guessing this is a temporary business maneuver to help Codex catch up on market share.
That said, the online version was pretty good. I found myself building out code and docs from my phone sometimes.
I have a full time job and also maintain another significantly large open source project solo (Kuberhealthy), so my flags package for Go was often neglected despite having 850+ stars and being used in lazydocker and others. I had a lot of guilt about this, but now that I've been getting used to using AI, returning to this project to add the features I always wanted it to have (completion for shells) was doable from my couch with family and kids as a constant distraction.
I am optimistic that AI is a good thing for software and it's certainly made it possible for me to stay on top of my open source projects for the first time in many years.
You get some amazing results with teams of AIs if you do it right. The key is to control behavior with what integrations and responsibilities each agent has. That way they naturally adapt, delegate, fact check each other, and generally act more autonomously.
This is already running the automated news site ainews.personastack.ai complete with social media posts 100% automated.
It also runs the issue triage, coding, reviews, and releases for the Kuberhealthy open source CNCF project, which is another thing of mine.
I don't think the next step is really smarter models. It's how we make the models more effective, and teams, when done right, net the best results I've seen.
Hoping to get noticed here soon, but it's extremely hard to do solo I'm finding.