August 12, 2026
Six bots, one founder, zero chill
My Agent Setup
One founder built six AI helpers, and the comments instantly turned into a group therapy session
TLDR: A founder built six AI helpers to run parts of a business and nonprofit, but even they admitted the setup can create more work than it saves. Commenters were split between impressed curiosity and full-on exhaustion, with many saying the idea is fun but the payoff still isn’t there.
A founder showed off a homegrown team of six artificial intelligence helpers meant to run products and even support a nonprofit with fewer humans. There’s an assistant for scheduling, one for operations, one for coding, one for marketing, one for research, and one that basically babysits the server. On paper, it sounds like the dream: a tiny digital staff living in a cheap cloud computer, chatting through an open-source work app, and helping one person do the work of many. In reality? The author’s own confession stole the show: after letting an AI edit the post, they spent 10 times longer undoing its changes. Brutal.
And the comments pounced. The biggest mood was: cool idea, questionable payoff. One reader summed it up with surgical precision: it was worth it “for the journey,” but not for the return on investment. Another praised the setup while admitting they’d tried similar experiments and found the results still a little underwhelming in real life, especially for tasks like email and software planning where humans still have to steer the ship. In other words, the crowd loves the sci-fi fantasy, but they’re not exactly ready to fire the office.
Then came the drama. One commenter applauded the author for experimenting without “AI psychosis,” which is both a compliment and a devastating review of the wider hype cycle. Another dropped the nuclear take, calling the whole scene a “bikeshedding maximization machine.” Ouch. Even the more practical replies had that “I’m fascinated, but also deeply tired” energy. So yes, the digital worker squad is here — but in the court of public opinion, the real product might be the comment section meltdown.
Key Points
- •The author is experimenting with a six-agent system to help operate multiple products and a nonprofit with fewer staff and volunteers.
- •The six agents are assigned separate roles across administration, operations, development, marketing, research, and infrastructure management.
- •The current stack uses Hermes agents, shared documentation in Obsidian, Mnemosyne memory banks, and OpenAI GPT-5.6 Sol with GPT-5.6 Terra for subagents.
- •The setup runs on a DigitalOcean Basic Droplet with 4 vCPUs, 8 GB RAM, and 160 GB storage, secured with Tailscale and no public ports.
- •The author still does most coding through Claude Code with Fable, uses Codex as backup, and communicates with agents through Buzz by Block.