Before pods form
Ask who has Claude Code, Codex, ChatGPT, Gemini, DeepSeek, or only browser access. Route people into mixed pods.
Workshop operating guide
Build something real before the room leaves.
This is a hands-on build day, not a lecture. You will join a pod, pick a real problem, plan it with AI, and ship a visible artifact: a website, a workflow, a tool, or a working demo. Here is what the day includes: pod formation and role election, an AI model primer, guided setup on your own machine, two build sprints, and a three-minute demo from every pod.
Guide Index
Open the page that matches the moment.
Join the DiscordStep one: join the workshop server, create a Discord account if you need one, and find your pod room.
Model choicesWhich LLM to use and why: Claude, ChatGPT/Codex, DeepSeek, Gemini, and NotebookLM.
Plan modeWhen to plan, when to execute, and when to switch to a faster model.
Simple setupPlain-language setup for Claude Code, Codex, Terminal, PowerShell, WSL, and fallback roles.
Claude Code setupTerminal setup, model selection, auto mode, and plan-first prompts.
Codex and ChatGPTUse /plan, /goal, /status, and goal-driven work without overcomplicating the build.
Hermes fallbackA practical path for attendees who do not have Claude Code or paid tool access.
Pod rolesHow groups elect roles so coders and non-coders both contribute.
Prompt bankCopy-paste prompts for ideation, planning, building, debugging, presenting, and next steps.
Presentation prepWebsite/repo-first demo flow, with NotebookLM or deck tools as optional support.
Support the workshopDonation and sponsorship links for food, tokens, prizes, and future events.
Room Flow
Run the day in short, visible cycles.
0:00 Arrival and access check
Everyone joins the Discord, opens this guide, confirms one AI tool, and asks for help before the room starts moving.
0:15 Intro and outcome
Set the measurable outcome: every pod leaves with a visible artifact, a short demo, and one next action.
0:30 Model and workflow primer
Show why stronger models plan better and faster models are better for bounded execution after the plan is clear.
0:45 Pod creation and role election
Build mixed-skill groups. Elect a facilitator, builder, researcher, writer, tester, and presenter so non-coders own real work.
1:00 Ideation and scoring
Each pod lists three ideas and scores them by usefulness, feasibility today, demo clarity, and available inputs.
1:20 Plan-mode pass
Pods ask AI to make the smallest plan that can ship today, then cut everything that does not support the demo.
1:40 Tool-track setup
Claude Code, Codex/ChatGPT, and Hermes/fallback users split into setup tracks. Do not block the whole room for one machine.
2:05 Build sprint one
Build the minimum working artifact. Send one practical tip at a time instead of dumping every instruction at once.
2:45 Midpoint review
Each pod answers: what works, what is blocked, what is the demo, and what should be cut?
3:05 Build sprint two
Finish the artifact and test the demo path. Facilitators help unblock, but pods keep ownership of the work.
3:45 Demo prep
Prepare a three-minute story: problem, build, proof, next step. Website, repo, workflow, or output comes before slides.
4:15 Demos and close
Each pod presents, captures next steps, and shares the support link for anyone who wants to fund the next room.
Facilitator Intervals
Give the room one useful nudge at a time.
After role election
Drop the role prompt and make every pod name its facilitator, builder, tester, and presenter.
After planning
Ask each pod to cut scope, pick the first visible artifact, and switch to a cheaper or faster model for bounded execution.
During build sprint one
Send one tip about setup, one tip about prompts, and one reminder that blocked machines should move to fallback roles.
At midpoint
Review what works, what is blocked, what the demo is, and what should be removed from today.
At close
Ask for demos first, then share contribution and sponsor links after people have seen the output.