Free Live Lessons
Short, hands-on sessions on shipping AI you can actually trust. Free to join, hosted on Maven, and recorded so you can catch the replay if you can't make it live.
The AI Agent Kill-Switch — Who Can Stop It, and How
Three ways to halt a running AI agent before it retries forever, stalls, or ships the thing you can't undo.
- Circuit breaker, dead-man switch, and state machine — which halt fits which failure
- Deterministic caps (wall-clock, iteration) as a normal return, not a crash you hope fires
- Who holds the kill-switch — making 'stop' reachable by a human before you need it
Mon, Nov 2 · 9:00 PM ET
Register HereWhen NOT to Use AI — a Product Leader's Filter
A four-question filter for the calls where the honest answer is 'not a model' — and how to defend it in the room.
- The three failure patterns where forcing AI reliably backfires
- The filter: reversibility, error cost, data availability, human-trust need
- A worksheet you can run on your own roadmap before you commit to a model
Mon, Nov 30 · 9:00 PM ET
Register HereA Team of AI Agents That Won't Step on Each Other
Split work across multiple AI agents so they hand off cleanly instead of clobbering each other's output.
- Ownership boundaries and contracts between agents
- Deterministic handoffs that don't depend on the model behaving
- Where multi-agent actually helps — and where it just adds failure modes
Mon, Dec 21 · 9:00 PM ET
Register HereAI Output That Won't Lie to You
Evals and output checks so an LLM feature fails loudly instead of confidently returning something wrong.
- Why a confident wrong answer is the failure that costs you trust
- A quality gate and spot-check harness you can wrap around any output
- Catching regressions before your users do
Mon, Jan 11 · 9:00 PM ET
Register HereAI Costs That Won't Wreck Your Margins
Cut your LLM bill without cutting quality — route, cache, trim, and a hard cap gated on a quality floor.
- The four drivers of your token spend
- Route / cache / trim — with a quality floor so cheaper doesn't mean worse
- A hard per-session cap so a runaway loop can't run up the bill
Mon, Feb 1 · 9:00 PM ET
Register HereA Content Pipeline That Won't Ship Junk
Quality gates that stop weak drafts from shipping — a scored loop, not a vibe.
- Turn “the output feels off” into a number you can gate on
- A scored refine loop: accept, iterate, or regenerate
- Caps and a human gate so nothing sub-threshold ships
Mon, Oct 5 · 9:00 PM ET
View on MavenA Team Brain That Won't Overshare
Turn your team's notes into an agent that remembers — with access control that actually holds.
- Personal vs. team brain: the who-vs-what split
- Retrieval plus access control that scopes memory by role
- The part most RAG builds skip — and why it breaks in production
Mon, Sep 14 · 9:00 PM ET
View on MavenAI Agents That Won't Wreck Production
Stop an AI coding agent from doing the one thing you can't undo — with a harness it can't argue with. Live.
- Why architecture beats a sharper prompt
- A deterministic guardrail that blocks a destructive command before it runs
- A starter template you can drop into your own setup
Mon, Aug 17 · 9:00 PM ET
Watch the replayEvery lesson is free and recorded. Register on Maven and you'll get the live link plus the replay — no charge, no catch.