Most students ask AI for answers.
This workshop teaches students how to use AI to create evidence: plans, datasets, prototypes, project logs, deployed pages, and work they can explain.
AI fluency for serious students
Counselor Jay's AI Workshop
Learn how to use AI like a project partner: build sharper portfolio projects, research smarter, publish real work, and show colleges or startups that you know how modern AI workflows actually work.
College application projects
Startup-ready AI fluency
Research and build systems
BuildWebsites, dashboards, datasets, automations, and project demos.
ResearchCompetition databases, winner analysis, literature maps, and roadmaps.
PublishGitHub repos, personal domains, case studies, and explainable project evidence.
SignalShow colleges, internships, and startups that the student can work with AI responsibly.
Codex is the current default.
Claude remains useful, but some students run into age restrictions. We start with Codex, keep a Claude harness reference, and teach students that the model leaderboard changes month to month.
The student still owns the work.
AI can organize, research, draft scaffolds, check code, and find gaps. It cannot replace the student's judgment, effort, source understanding, or disclosure obligations.
Start with the vocabulary
The names get confusing fast.
OpenAIThe company behind ChatGPT and Codex.
ChatGPTThe general chat app many students already know.
CodexThe agentic desktop workflow we use first for building projects.
AnthropicThe company behind Claude.
ClaudeThe chat assistant, often strong for ideation and deep project strategy.
Claude CodeAnthropic's coding and agentic workflow. Useful, but not the default student path here.
Level 1 milestone
Install an agentic harness.
A harness is the file-based system that tells Codex or Claude Code how to work: roles, rules, project memory, verification standards, and what counts as finished.
Jay's starter harness is named Alfred, after Batman's butler. Alfred is a personal assistant and organizer, not the student's replacement. Students can rename, revise, and improve the harness over time.
Starter harness files
AGENTS.mdgives Codex its role, rules, agents, and verification standards.CLAUDE.mdserves the same purpose for students using Claude Code.SESSION.mdrecords project state so every new session starts informed.MEMORY.mdcaptures mistakes, discoveries, and workflow improvements..secrets/keeps API keys out of chat and out of public repos.
Four levels
What students can become capable of doing.
Level 1
AI Power User Setup
Choose Codex or Claude Code, install a starter harness, and stop losing context every time a new chat starts.
Student value:A repeatable AI workspace that makes projects faster, cleaner, and easier to resume.
Level 2Portfolio-Grade Projects
Use sequencing to turn vague ideas into databases, research maps, GitHub repos, deployed websites, and competition plans.
Student value:Work that can be shown, explained, improved, and connected to applications or internship conversations.
Level 3Frontier Plus Local Compute
Use subscription models for judgment and orchestration while routing bulk work to local LLMs or Counselor Jay's lab fleet.
Student value:A cost-smart AI workflow that looks more like how advanced builders actually work.
Level 4Personal AI Lab
Move toward local agents like Hermes and OpenClaw, personal knowledge bases, browser-aware workflows, and durable project memory.
Student value:A serious operating environment for research, building, and long-running goals.
Why this is worth learning
The advantage is not "using AI." Everyone can do that.
The advantage is knowing how to structure the work: gather the right evidence, sequence the phases, verify outputs, publish proof, and explain the decisions. That is the difference between a student who prompts and a student who can actually build with AI.
The central method
Sequencing turns ambition into a build plan.
A weak prompt asks AI to solve the whole problem at once. A strong sequence isolates the phases so the student gets evidence, strategy, and a realistic path.
01
Create a database of relevant competitions, eligibility rules, deadlines, and submission requirements.
02
Identify the last five or more years of winners and analyze what strong submissions had in common.
03
Upload the student's current resume, project history, skills, constraints, and time horizon.
04
Ask the harness to connect the evidence to the student's profile and propose realistic project directions.
05
Turn the best direction into a roadmap with datasets, milestones, GitHub commits, and verification checks.
Non-negotiable
Academic integrity is built into every guide.
Allowed
Brainstorming, planning, tutoring, source organization, code debugging, dataset cleanup, study scaffolds, and project management when the student understands and owns the work.
Risky
Heavy rewriting, undisclosed assistance, unverified research summaries, generated citations, and code the student cannot explain.
Not okay
Submitting AI work as original student work, fabricating sources or results, bypassing school policies, or using AI where the assignment forbids it.
For parents
A practical investment path for parents.
Free
Enough to see the promise, but limits arrive quickly once a student starts building seriously.
$20/month
The best starting point for most families: enough room to learn the workflow without overcommitting.
$100/month
Useful when the student is building regularly and the subscription is clearly saving time.
$200/month and hardware
Advanced territory. High-RAM Macs or lab access make sense after the student's work justifies scale.
Later invitation
Students eventually see the AI lab model.
The public workshop shows what the lab does without making lab access the starting point. Students first learn the desktop workflow, harness setup, sequencing, GitHub, domains, and verification. Lab compute comes later, when they have enough project discipline to use it well.
Frontier subscriptionsPlanning, reasoning, code generation, tool use, verification.
Local LLMsBulk summarization, extraction, classification, and high-volume drafts.
lab.counselorjay.comDedicated AI MacBooks for students who are ready for serious workloads.
Guide library
Start with the ideas that change how students work.
Level 1
One-Shot Prompting vs. Agentic Harnesses
Why serious AI work needs context, tools, files, roles, standards, and verification.
Level 1Install The Starter Harness
The private workshop gives students Jay's starter AGENTS.md named Alfred and the first project setup flow.
The Counselor Jay Sequencing Method
How to turn a vague ambition into a buildable project with evidence, milestones, and proof.
Level 3Why Local LLMs Matter
The existing local LLM posts become the advanced track for orchestration, cost control, and private high-volume inference.