MuggsOf
Computer Science
Fall 2026
One Lab Scene plus the Setup Scene that primes it and the CFU that closes it. All three artifacts travel together; a class block is built by arranging them around the 50-minute lab.
Setup Scene on Markdown specs + SDD. The deck is built to split - about two-thirds delivered before the lab to prime the work, one-third after to land the takeaways.
A 50-minute Lab Scene from Act 1 - Sketch, Spec, Swap. The sub-runnable Scene Script printout (Print to PDF from the toolbar).
Everything a student produces from this triplet: the lecture-aligned CFU plus the four Lab Scene artifacts (sketch, four-section spec, reconstruction of a peer's spec, reflection).
Working notes that don't belong inside any single artifact - scaling questions, model evaluations, follow-ups carried into the next triplet.
What this course is, who itâs for, and what youâll actually do.
Sean Muggivan. I am a Licensed Clinical Social Worker (LCSW) working on full certification in high school math and computer science. My background isnât in corporate IT, but in Community Systems. I view Computer Science not just as a set of job skills, but as a tool for community agency and âSovereign Engineering.â I build and maintain the hardware and software ecosystem we use in this lab.
Iâm building this course as a pathway into computer science for everyone - not just the students who already self-identify as âtech kids.â If your interest is biology, music, journalism, sports, social work, art, history, trade work, or youâre still figuring it out: thereâs a place for you here. Computer Science is becoming the common substrate underneath every field, and the goal of this lab is to make sure no student gets locked out of that substrate because of their background, learning needs, or assumed major. The Markdown-first design, accommodation supports, and PAR projects are all there so the on-ramp meets you where you are.
Markdown is a simple way to add structure to plain text using a few special characters. A # makes a header. A * makes a bullet. **word** makes that word bold. Triple-backticks wrap a code block. Thatâs basically it - it looks like normal writing, but it has hidden structure underneath that any computer can read.
This course uses Markdown for nearly everything: specs, prompts, blueprints, notes, even pages on this site. Three reasons it earns that role:
1. Itâs structured and readable at the same time.
Most formats force you to pick one. JSON and XML are structured but unreadable without practice. Plain prose is readable but has no structure for a machine to grab onto. Markdown is the rare format thatâs both - clean for you, parseable for a computer.
2. AI models follow Markdown structure precisely.
When you use a header, the AI treats that section as its own job. When you use a numbered list, the AI follows the steps in order. When you use a code block, the AI treats it as the exact output format you want. Structure is how you control the AI without writing real code.
3. It makes your spec the actual work.
In Acts 2 and 3, your Markdown spec is the project. The AI translates your structured English into HTML, CSS, and JavaScript. Your writing is your code. Thatâs why we say the spec is the work.
Bonus: Markdown is plain text. No proprietary editor, no subscription, no lock-in. Every text editor on every device handles it. Once you can write one Markdown spec, you can write one anywhere.
Louisiana now requires one credit of Computer Science for graduation (Act 211). The LDOEâs vision is to move students from being passive consumers of technology to active creators who understand how their tools work. This course is the CS I foundation built for that mandate - see âMadness to the Methodâ for the philosophy underneath it.
Participatory Action Research (PAR) is the framework that grounds this course in real needs. Instead of made-up projects, we start by asking: what does this school community actually need? The auditing and building process in Acts 2 and 3 is driven by real problems identified through observation and inquiry. When you build a tool in Act 3, it solves a need that matters to someone in your school.
This course is the CS I foundation that creates the on-ramp to AP Computer Science Principles. Every act is vertically aligned to AP CSP Big Ideas and Computational Thinking Practices. Act 1 builds the systems literacy and structured thinking that AP CSP assumes you already have. Act 2 covers data analysis and algorithmic bias - core AP topics. Act 3âs Prompt-to-App workflow mirrors the AP Create Performance Task process: spec, build, debug, defend. Students who complete this course enter AP CSP ready to perform, not just participate.
A Logic Blueprint is a structured Markdown prompt that makes a local AI behave as a specific tool. It is the core skill of this course. Instead of typing vague questions into a chat box, you write precise instructions using headers, numbered steps, and defined input/output formats.
Why Markdown?
AI models read your prompt once, top to bottom. When you use Markdownâs structure - headers for sections, lists for steps, code blocks for formats - the AI follows your instructions like a recipe. When you write vague paragraphs, the AI guesses. Structure is control.
What Makes a Good Blueprint?
Role: What tool should the AI act as? (e.g., âYou are a vocabulary quiz maker for Biology class.â)
Input: What information will you provide? (e.g., âI will give you a list of 10 terms and definitions.â)
Output: What should the AI produce? (e.g., âGenerate a 10-question multiple-choice quiz with an answer key.â)
The test of a good Logic Blueprint is simple: can someone else use it cold - without any explanation - and get the intended output on the first try?
There are two main ways to use an AI to do a task. One is the way most people use it. The other is the way this lab will train you in. The difference shapes both what you produce and what you keep learning along the way.
The Trap: AI as Data Source
You hand the AI a prompt or a task. The AI generates the data, the answer, or the writing - and you take it as fact. This is the default ChatGPT pattern: students using it for homework or research, asking it to âtell me about Xâ or âwrite me a paragraph on Y.â Two problems make this mode quietly dangerous: the AIâs output is what it predicts is likely to be said, not whatâs actually true; and the AI is doing the thinking, while youâre just receiving.
The Method: AI as Analyst
You bring the data - sources, evidence, your own draft, a piece of code - to the AI along with a structured prompt telling it how to analyze what youâve provided. Now the AIâs job isnât to generate data out of thin air; itâs to read what you brought and help you see patterns, check arguments, or transform the input into a different format. Your sources stay in your hands. The thinking stays partially yours.
This is the precursor to a much bigger distinction youâll meet in AI: scientist AI (you bring the question and the evidence; AI helps you reason) versus agentic AI (you give a goal; AI generates content or takes actions on its own). Both have their place. But if you donât know which one youâre using, youâre being used by it.
Fair question. This course isnât about becoming a programmer. Itâs about staying yourself in a world where AI is everywhere and algorithms increasingly make decisions that affect your life - from loan approvals to job applications to what you see on social media.
1. AI Is Already in Your Life
Whether or not you personally use AI, itâs already in your life. The TikTok and Instagram algorithms picking what shows up in your feed are AI. Your schoolâs essay-grading software probably uses it. The image filters on your phone use it. Your friends use it for homework. So the question isnât really will AI be part of your life - it already is. The question is whether you understand it well enough to recognize it, audit it, and decide for yourself when to use it.
Some students decide they donât want to use AI to do their thinking for them, and thatâs a position worth respecting. This class isnât trying to talk you into being an AI user. What it is doing: making sure you can see what AI actually does, catch its mistakes, and know that the commercial cloud services everyone pitches you (ChatGPT, Gemini, Copilot) arenât the only option - there are local, private, self-hosted versions you can run on hardware you control, with your own data, on your own terms.
2. The Trap (For People Who Do Use AI)
People who let AI write for them notice their essays and even their ideas starting to sound like everyone elseâs: generic, GPT-flavored, smoothed over. The same flattening happens to projects - when AI makes every design choice, every app looks like every other AI app. And the brain muscle for sustained, hard thinking only grows when you actually use it; outsourcing the hard parts feels great in the moment and quietly atrophies that capacity over months.
This lab is built around the opposite habit. You write a structured spec in Markdown - your decisions, your aesthetic, your priorities - and the AI builds what you described. Your voice and your taste stay yours because the spec is the work. When something breaks, you donât ask the AI to fix it; you read the error, audit your spec, and find what your instructions missed. Thatâs where the cognitive stamina actually grows.
3. The Architecture of Power
You may not be interested in coding, but youâre definitely interested in not being manipulated. The skills in this class apply to your life right now:
The Feed: When TikTok or Instagram puts something on your For You page that makes you feel terrible about your body, your friends, or your life - can you recognize that the algorithm picking that video is optimizing for your attention, not your wellbeing? Can you tell when itâs working on you?
The Group Chat: When an AI-generated image, fake screenshot, or âleakedâ clip goes around the group chat, can you spot the giveaways before you forward it? Can you tell when a friendâs essay was actually ChatGPTâs?
The Score: When your schoolâs AI essay-grader marks you down and the teacher canât fully explain why, do you know how to figure out which patterns triggered the score? When your part-time job application gets auto-rejected by a screener, do you understand the filter that blocked you?
4. This Is a Laboratory for Discovery, Not a Coding Boot Camp
This is a space for discovering what you can do with computer science - not just coding, not just programming. The AI does the coding in Act 3; you do the thinking: writing the spec, designing the UI, verifying the output, and defending your choices. But the deeper learning is the lab itself.
We learn how a whole tech stack gets the job done - and how understanding it lets you do more with it. We study the relationship between software and hardware: why an AI model runs differently on a 4090 GPU than on the integrated graphics in a laptop, why those differences matter, and how to choose the right tool for what weâre actually trying to build.
We learn whatâs possible with free and open-source software. Most of what we use - llama.cpp, GGUF model files, the libraries behind every page on this site - costs nothing, is auditable, and is built by people instead of corporations. The paid subscriptions youâve been pitched arenât the only path; often they arenât even the best one.
And we learn how much can be done to keep our data off cloud servers, where every prompt and every photo becomes training material for billion-dollar companies. When the model runs in this room, on hardware we control, our data stays in this room. Thatâs a choice you should know exists.
And yes, this course also satisfies Louisianaâs Act 211 Computer Science graduation requirement (one credit, starting with the class entering 9th grade in 2025-2026). We list that last on purpose - compulsion isnât a good reason to take any class seriously - but if you needed permission to be here, you have it.
Almost none. This course is Markdown-first. The AI does the coding; you do the thinking and designing. Your job is to write clear instructions (Logic Blueprints and Markdown specs), verify the output, and make design decisions about how the final product looks and works.
The Markdown Spec (Your Main Work)
Most of your effort goes into writing structured Markdown: Logic Blueprints in Act 1, verification reports in Act 2, and app specs in Act 3. This is written in plain English with Markdown formatting. You are the Architect, defining what gets built and how it should behave.
The AI Translation
In Act 3, you feed your Markdown spec to the local AI running on the classroom GPUs. The AI writes the HTML, CSS, and JavaScript for you. You arenât cheating - you are managing AI output like an architect managing a contractor.
CSS & Troubleshooting (The Hands-On Part)
You will learn CSS basics to style your tool and F12 basics to diagnose errors. When something breaks, you use the Console and Elements tabs to find the problem. But the fix is usually updating your spec and regenerating, not writing code from scratch.
Most people look at a computer and see a Service. They see a âGreen Dashboardâ that promises to solve their problems if they just click the right buttons and pay the right subscriptions.
We donât do that here. In this lab, we look at a computer and see an Architecture. We see a reaction that can be mastered, a system that can be audited, and a territory that can be defended. Here is the logic behind our âinefficient,â high-friction, and entirely sovereign method:
1. The English Majorâs Revenge
The secret at the heart of modern AI is that Clear Prose is the Ultimate Code. If you canât describe a problem, you canât solve it. We treat the English language as our primary programming language. If your writing is vague, your website stays offline. We donât teach âcodingâ; we teach Instructional Precision.
2. The âOppenheimerâ Protocol
We donât fear failure; we require it. Real science doesnât happen when everything works on the first try - it happens in the âpost-mortemâ of an Out-of-Memory (OOM) error or a collapsed Oculink bridge. We call this âFailing Forward.â Your âRedâ errors are the only data points that actually matter.
3. The Markdown Ticket
No one âvibesâ their way onto the internet in this lab. Your subdomain is a locked gate, and a Validated Technical Spec is the only key. You donât touch the hardware until you have architected the intent in Markdown. We audit the blueprint before we lay a single brick.
4. Sovereign Engineering
We donât outsource our brains to the âCloud.â We host our own models on our own hardware. By bridging local 4090 GPUs with GitHub Codespaces and Cloudflare Tunnels, we prove that you donât need a corporate subscription to have professional power. We arenât âusersâ of the machine; we are its Systems Administrators.
5. The âGPSâ vs. The âArchitectâ
A âServiceâ acts like a GPS - it tells you where to turn, but you never learn the neighborhood. âArchitectureâ forces you to build the map yourself. Our method is intentionally âtediousâ because that tedium is where expertise is built. We arenât producing âsuccessfulâ students who lack resilience; we are producing Sovereign Adults who can look under the hood of any system and find the truth.
The Bottom Line
âWe arenât here to use the AI; we are here to master the reaction. There will be OOM errors. The tokens per second will plummet. The code will break.
Thatâs lab work.
We donât compete with the âGreen Dashboardsâ of polished software; we compete with our own ability to architect the truth.â
Muggs Lab Note
This statement serves as the âhandshakeâ for the lab. It tells the student that their struggle is the point, their writing is their power, and their teacher is a fellow operator who built the stadium theyâre playing in.