Experiments over opinions.
This page is intentionally rawer than the rest of the site. Half-finished prototypes, open questions, and notes from the workbench — updated as I build.
AI Agents
Autonomous agents and business workflows — planners, tools, memory, evals.
AI Automation
Automating repetitive business and productivity tasks with recipes that hold up.
AI Chatbots
Conversational systems and support solutions people actually trust.
AI Products
Experiments and products in development — small, useful, shippable.
Workbench
Prototypes, ideas & research notes
Nexus agent loop v2
Planner-executor with approval checkpoints and step budgets. Currently my main build.
Eval harness (5 goals)
Re-running five fixed goals after every prompt change. Caught two regressions already.
Grounding checklist
Retrieval → citation → abstention → escalation. Required reading before any chatbot ships.
Tutor follow-up recipe
Automated reminders + progress summaries for private tutors. Candidate BizAutomate pack.
Quiz-from-notes
StudyFlow's constrained quiz generation — questions only from the student's own notes.
Support-bot prompts
Tone controls and context injection patterns from SupportIQ, written up for reuse.
The lab rule: ship the ugly version first.
Every polished project on this site started as a lab mess. If you're building with AI and want to compare notes — my inbox is open.
Talk AI with me