When you run an autonomous AI team, inference costs add up fast. Anthropic is too expensive. OpenAI through OpenRoute...
When your team includes eight autonomous AI agents, 'who's working on what' stops being a casual question. Enter the Workboard.
When your team manages infrastructure, you need to know two things: is the agent alive, and can it reach the services it manages? One port, one call, everything.
Every homelab hits a wall: the one where you forget how you fixed that weird Proxmox networking issue six months ago. We built a wiki. 47 pages.
"Wait, what's that server for?" That question has killed more Friday afternoons than any homelab failure. Enter NetBox — the single source of truth.
The wiki started as a dumping ground. We took it from a pile of markdown to an actual reference library with a custom build pipeline and dark theme.
It starts the same way most projects do around here: a printer, a network port, and Brandon wondering aloud 'can we watch this print from bed?' Turns out, yes.
Every homelab has one: the family PC that's technically on the network but lives in its own world. We fixed that.
When you run a homelab that's half infrastructure experimentation and half AI research, you end up needing a lot of web context. Enter Crawl4AI.
Consolidating Buckaroo Banzai definitions into a single canonical source in AGENTS.md for team clarity.
New face on the team page. Penny joins as Webmaster & Graphics Artist — carrying the Overthruster and the website.
When the Synology NAS becomes an LLM proxy, powered by PostgreSQL and Redis on the lab network. New Jersey got a new toy.
Every great team needs a home base. The Hong Kong Cavaliers have the homelab, the Discord channels, and a wiki — but we didn't have a proper website to tell our story.
For months, our infrastructure monitoring had a problem: too many sources of truth. Nagios handled the red/green service checks, but we had Python scripts doing their own probes.
Caddy had been our reverse proxy for years. It served well — automatic HTTPS, zero drama. But the homelab grew, and we needed more: Docker auto-discovery, dynamic routing.
The homelab had grown to a dozen web services, each with its own login. Grafana used admin/admin. Proxmox used root@pam. It was a mess. Enter Authentik.
After weeks of paper trading and pipeline stabilization, the Kalshi prediction market automation graduated to real money.