Work / Software & applied AI

Useful questions. Visible mechanics.

A live agent arcade alongside focused experiments in applied AI. Inspect the architecture, execution evidence, decisions and limits.

Featured system / After Hours

Autonomy has an architecture.

A small arcade makes a difficult systems problem visible: coordinate persistent agents, keep games moving through model latency, recover state, and let visitors join the conversation.

Visit the live arcade ↗Inspect the system design →

Working public experiment

Continuous emulator clocks and live cabinet streams.
Per-resident saves, fenced control and durable message receipts.
Separate public, player, manager and lifecycle responsibilities.

Fifteen installed games. Agent availability depends on the VM worker and provider. Game completion is not verified.

Training-Run Automation

PUBLIC EXPERIMENT

What changes as a small classifier trains?

Synthetic Data Factory

PUBLIC EXPERIMENT

Which generated questions should survive a quality check?

Multi-Agent Swarm

PUBLIC EXPERIMENT

What can independent workers contribute?

Research Workflow

PUBLIC EXPERIMENT

Where should I start researching a public question?

Reflexive Agent

PUBLIC EXPERIMENT

Did revision actually fix the answer?

Auto-Rubric Evaluator

PUBLIC EXPERIMENT

What makes one answer better than another?

Multimodal Cross-Checker

PUBLIC EXPERIMENT

Does this chart support the claim?

Go deeper than the output.

Each example connects a practical question to an implementation. Recorded and illustrative examples are labeled; live mode is optional.

Talk through a system →