Building In Public: Week 1

No pitch deck. No corporate comms. Just the truth about week one.

The stack.

Everything is self-hosted on a VPS I own. The UI is Next.js + React. The AI infrastructure is a Hetzner server running Qwen models locally, with RunPod for GPU tasks, free API tiers for non-critical work, and Claude as a last resort.

The whole thing costs me about $150/month to run at current scale. At 100 users, it'll be $400/month. At 1,000 users: $1,200/month. The unit economics work.

The numbers (honest edition).

Current waitlist: Single digits. I launched the waitlist page 48 hours ago.

Time spent building: ~6 months of nights and weekends. The platform actually exists. This isn't a landing page with a "we're building this" promise. You can talk to the team right now.

Features working: Chat with all C-Suite members, the full Creative Studio (avatars, voice, video generation), three domains, The Architect coordination layer, session memory, the onboarding flow.

Features still building: Email integration, automations engine, social media dashboard, mobile app.

What went wrong this week.

Better-sqlite3 native module wasn't compiling correctly in the Docker build. Spent 2 hours debugging before finding the cause (host node_modules overwriting Alpine Linux binaries in the container). Fixed with a one-line .dockerignore entry.

Domain routing for athelm.ai through Cloudflare had mixed HTTP/HTTPS behaviors. Resolved.

The OG image meta tag was pointing to a file that didn't exist in production. Created a proper PNG.

What's actually hard.

Not the code. The code is solvable — there's always a path forward with code.

The hard part is building alone with high standards. Every line, every design decision, every word on every page — one person deciding. No QA, no second opinion, no "does this make sense to you?"

The Architect is the obvious irony here. I'm building a product that gives people a team — while doing it alone.

What's next.

This week: finishing the blog system, docs, support pages. Next week: refining the onboarding flow (the Board Meeting sequence). Week after: email integration.

The goal for month one: 100 people on the waitlist. Not 10,000. Not viral. 100 people who actually see what we're building and want to be founding members.

If you're one of them: hello. You're early. That matters.

— DarkHorse ♞