Weekly Live AI Workshops
Every week Jason and the AI Builder Club community go live to share new AI tools, agent workflows, real-world builds, and answer your questions in interactive sessions. Can't make it live? The complete archive of past workshops and recaps is below, so you can catch up anytime and reuse the techniques in your own projects.
Created by
Jason Zhou · Founder, AI Builder Club
Course Outline
How the Live Sessions Work
Each week splits into project & knowledge sharing followed by an open Q&A.
09/13 - Realistic AI UGC: Characters That Don't Look AI
Video of a real person is easy now. A brand new person who looks real is the hard part, so this session builds one: a locked JSON character prompt that fights the image model's defaults, the model that draws the most believable face, and Seedance 2.5 with a real audio reference.
09/06 - Making High Quality Launch Videos with HyperFrames
Jason builds a 78-second launch video with HyperFrames from scratch. Why HTML gives an agent a timeline it can check frame by frame, and why the first few compositions need hand direction before the rest can copy them.
08/30 - Chain of Custody for AI Work, and a Week on Grok Bot
Member session on proving AI-made work is real when anyone can generate a portfolio in a day: a voice clone grounded in cited facts, and a phone number an agent can actually receive verification codes on. Then Jason on a week of Grok Bot as his daily driver: local and cloud in one agent, and a Cursor VM per task.
08/16 - Ara: Cloud Coding, Verified PRs, and Token Maxing
Ara founder Adi Singh on moving a codebase to the cloud with its secrets and CLIs intact, then shipping without a code review bottleneck: every PR carries a before and after session link as proof, and each merged fix spawns a follow-up check against real usage data.
08/02 - Graph Engineering Without a Framework
One term covering three different problems, and the two ways to actually enforce a graph: control flow written as code, or an SOP written into a skill the model runs. Plus the design patterns for cutting a job into agent nodes, and three graphs running in production.
07/26 - Vibe Coding From Your Phone, and a Loop on Reddit
Member session: an Android client that drives Claude Code on a personal VPS by voice, so software gets built in the gaps a real life leaves. Plus Jason's Reddit-karma loop, on a channel he had failed at manually.
07/19 - Running a Fleet of Parallel Agents with Orca
Orca founder Jing on running a fleet of parallel coding agents: cross-agent orchestration through the Orca CLI, taming token cost with multi-model routing, and a live demo of parallel worktrees, design mode, and automated PR review.
07/12 - Anatomy of a Good Loop + Hermes Personal Automation
A member demos Hermes, a personal executive-assistant agent, then Jason breaks down what makes a loop trustworthy: the contract, four trigger types, the verifier, and the self-improving evolve step.
07/05 - Pi Agent Deep-Dive: Customizable Agent Harnesses
Jason's deep-dive on Pi Agent (pygen): a minimal coding harness you fully customize through extensions, where the agent can write its own extension and hot-reload the harness it is running in.
06/28 - Codebase Memory MCP + Postia Teardown
A pure-programmatic code graph that cuts agent token cost ~50% via a grep hook, plus a full teardown of Postia, an agent that builds and runs a whole business.
06/21 - Crabbox: Isolated Sandboxes for Parallel Agent Testing
Give every parallel agent its own isolated cloud sandbox that syncs only the dirty diff, tests in seconds, and returns video evidence to the PR.
06/14 - Loop Engineering Best Practices
Replacing manual prompting with self-running agent loops across support, SEO, growth, and ads - Claude Code on local machines with cron triggers.
06/07 - Harness Engineering: Legible, Executable, Verifiable Repos
Make your repo legible, executable, and verifiable before running serious agentic workloads.
05/31 - Voice Agent Error Rates & Production Iteration
Building a calendar-booking voice agent with Dogra + Claude MCP, and taming voice AI error rates in production.
05/24 - EachLabs CTO: From Model Chaos to Unified Workflows
How EachLabs unifies 600+ AI models behind a single API, and why multi-model workflows persist.
05/17 - AI Wiki Toolkit: Eliminating Agent Memory Loss
A folder-based system for durable agent knowledge that compounds across sessions and repos.
05/10 - AI CTO Architecture: Orchestrating Agents Across Isolated VPS
Anicet (ascii.dev) demos an AI CTO that delegates to coding agents running on isolated cloud desktops.
05/03 - OpenAI Symphony + Automated Testing Setup
OpenAI's Symphony open-source scheduler for coding agents, plus a harness-ready testing setup.
04/26 - Open Organizational AI with PKA and Departmental Agents
Simon Wallace-Jones's production system coordinating 36 codebases and 17+ dev tasks/week autonomously.
04/12 - Voice-to-Letter Pipeline: Agent Memory + Physical Mail
Anders Kitson's voice app that turns spoken words into physical mailed letters, rebuilt in 2 days.
03/29 - OpenClaw Production Guide: Real Business Workflows
Advanced OpenClaw for real business workflows, beyond basic tutorials.
03/22 - How I Use OpenClaw Internally
An inside look at how OpenClaw is used internally for day-to-day AI-powered workflows.
03/15 - From Probabilistic to Production: Multi-AI Agent Workflows
Learn how to move multi-AI agent systems from experimental to production-grade reliability.
03/01 - Building Context Infrastructure for Agent Collaboration
Explore how to build robust context infrastructure that enables multiple agents to collaborate effectively.
02/22 - Toolhouse.AI: No-Code Agent Building
Build powerful AI agents without writing code using Toolhouse.AI's no-code platform.
02/15 - GitContextController, OneContext, File-Based Memory
Master advanced context management tools including GitContextController, OneContext, and file-based memory systems.
02/08 - Production-Scale Code AI: AugmentCode Deep-Dive
A comprehensive deep-dive into AugmentCode for production-scale AI-assisted software development.
01/25 - Building ChatCut: Video Editor Agent Systems
Learn how to build ChatCut, an AI-powered video editor using agent-based systems.
01/18 - Long-Running Coding Agents & MCP Context Optimization
Techniques for building long-running coding agents and optimizing context with MCP.
01/11 - Context Engineering: Learning from Manus
Explore context engineering best practices and lessons learned from building the Manus agent.