What I'm Doing Now

Posted on Jan 11, 2026 • 3 min read

What I’m Doing Now

Updated: January 11, 2026

This is a now page, inspired by Derek Sivers. It’s what I’d tell a friend I haven’t seen in a year.

January 2026: Shipping, Not Building

I’m done with endless prototypes. Time to ship production work.

Active Projects

1. The Red Corner: Moscow 1956 🎭

Status: 100% production-ready, event TBD

12-player murder mystery set in 1956 Soviet underground. Complete with 12 character dossiers (30+ pages each), interactive dashboard (132 relationships), 480-line GM script, voting system, and venue setup guide.

What makes it special: Built with unconventional tools (Groq, DSPy, Typst, Vis.js, custom ontology) that enable 3-hour generation time instead of 3 months.

Live demo: Interactive Dashboard

Explore The Red Corner →

2. Artefactum: Mystery Game Generator

Status: Production-ready, commercially viable

Automated pipeline that generates complete 12-player murder mystery games in 3 hours. Multi-stage generation with graph-based validation, custom ontologies, and professional PDF typesetting.

What it generates: 12 dossiers, 14 clues, 36 objectives, 44 props, 132 validated relationships, interactive dashboard, GM materials, deployment package.

Innovation: DSPy structured programs (not prompts), Groq 0.3s inference, Typst 2s compilation, domain-specific ontologies.

Commercial angle: Exploring product launch, licensing, or pre-generated game sales.

Read about the generator →

3. EventFlow

Status: Feature-complete, production-ready

Full-stack event planning platform with Caroline AI (47 type-safe tools, 87.5% accuracy on messy meeting notes). Built for chaotic planners with creative vision but scattered execution.

Using it for: An actual wedding (survival mode = best motivator)

Next: PostgreSQL migration, production deployment prep

Read the EventFlow story →

4. Orfanato (Next Game)

Status: In development (80% complete)

Portuguese orphanage mystery. Second production game from Artefactum generator.

Currently Learning

  • DSPy - Structured LLM programs (now production-ready in my stack)
  • Typst - Modern typesetting (2s compile time, production-proven)
  • Graph validation - Relationship consistency checking for complex systems
  • Custom ontologies - Domain-specific AI outputs that feel handcrafted
  • Groq inference - Sub-second LLM generation (0.3s vs 5-10s)
  • Commercial viability - Turning tools into products people pay for

Tech Stack (Production-Proven)

Backend: FastAPI, Pydantic, SQLAlchemy AI: DSPy, Groq (llama-3.3-70b), Claude Code Typesetting: Typst (not LaTeX) Visualization: Vis.js (3D network graphs) Database: PostgreSQL, SQLite Deployment: Netlify, GitHub Pages, Hetzner + Tailscale

Q1 2026 Goals

  1. Generate 5-10 mystery games - Different themes, validate commercial viability
  2. Run The Red Corner event - Real-world playtesting with players
  3. Explore commercialization - Sell games? License generator? Both?
  4. Document unconventional tools - Blog posts about Groq, DSPy, Typst, custom ontologies
  5. Complete Orfanato - Second production game (80% done)

What I’m NOT Doing

  • Tutorial hell (if I’m not building something real, I’m wasting time)
  • Freelancing or consulting (not interested right now)
  • Traditional job hunting (unless the perfect AI Creative Producer role appears)
  • Building things nobody wants (validate commercial viability first)

Philosophy Shift

I used to build to learn. Now I’m learning to build products people pay for. The skills are the same, but the mindset is different.

Want to Connect?

This page gets updated whenever something major changes.

Last update: January 11, 2026

Latest: The Red Corner is 100% production-ready. Artefactum generator proven viable. Shifting focus from prototypes to products.