Tools I reachfor, in order.
Tools I use to research, build, test, and maintain systems. The point is what they make possible, not how long the list is.
AI & models
Python, data, and persistent systems are the working tools behind ZAI, research experiments, and infrastructure.
Python
Daily driverMain language for ZAI, the World Model, Oracle, and most internal tooling.
PostgreSQL
Daily driverPersistent state and structured records for AI infrastructure, research tools, and time-series projects.
Node.js
ComfortableGlue layer for the trading terminal and the Next-side APIs.
Supabase
ShippingAuth + storage when a project needs to ship fast without becoming infra.
Frontend
Interfaces for research and real use: legible records, clear controls, and responsive reading experiences.
Next.js
Daily driverApp Router, server components, streaming. Default for product work.
React
Daily driverComposable UI, expressive state. The substrate I think in.
TypeScript
Daily driverStrict mode always. Types as a design tool, not a tax.
Tailwind CSS
Daily driverTokens via @theme. Fast iteration without CSS sprawl.
Platform
Shipping, monitoring, and keeping the lights on without ceremony.
Docker
ComfortableReproducible builds, simple deploy targets for the studio's services.
GitHub
Daily driverSource of truth + CI. Anteroom Studio's work lives here in the open.
How I choose tools
Long memory beats hot takes
Long-lived research and AI systems need context they can trace and correct, not a stream of unexamined claims.
Pick boring tools
Reach for the proven defaults first. The interesting bet is in the system design and the reasoning layer — not in adopting the latest framework.
Reproducibility from day one
Treat source trails, repeatable tests, and explicit limitations as part of the system, not documentation added at the end.
A question worth exploring.
Open to thoughtful correspondence, research exchange, authorized security work, and engineering collaborations where the problem is worth investigating.