Gallery

Contacts

405 W. Greenlawn Ave Lansing, Michigan 48910

contact@techjacksolutions.com

+1-616-320-4064

Twitter Facebook-f Pinterest-p Instagram
Tech Jacks Solutions Education & Training
Mockup Preview

Learn AI, from the page up

The AI Knowledge HubUnderstand it. Then go deeper.

Every concept here is a self-contained lesson on the page — a plain-language explainer, an interactive you can actually play with, and a quick check. When you're ready, each one opens into a full guide. Go from novice to intermediate by scrolling.

Start learning Jump to a topic ↓
14
Concepts
3
Learning tracks
2
Skill levels
~25min
To intermediate
Interactive lessons
0 / 4
Level
Jump to
Track 01

Foundations

Start here. What "AI" actually means, and how the pieces — machine learning, deep learning, neural networks — fit inside each other.

01
Novice 3 min Learned

Is it AI, machine learning, or deep learning?

The three terms people use interchangeably — and why they're actually nested inside each other.

They're not synonyms — they're circles inside circles. Artificial intelligence is the broad goal: machines doing things that normally need human smarts. Machine learning is one way to get there: instead of hand-coding rules, you let a system learn patterns from examples. Deep learning is a powerful kind of machine learning that uses many-layered neural networks — and it's what powers today's generative AI.

  • All deep learning is machine learning; all machine learning is AI — but not the reverse.
  • Old-school AI could be just hand-written rules (no "learning" at all).
  • The recent leap — ChatGPT, image generators — is deep learning at massive scale.
Interactive Tap or focus a ring to explore
Artificial Intelligence
Machine Learning
Deep Learning
Generative
AI
Outermost circle

Artificial Intelligence

Any technique that gets a machine to do something we'd call "intelligent" — including plain hand-written rules, with no learning at all.

Example: a thermostat's logic, a chess engine, a spam filter.

Which statement is true?

Exactly — it's circles within circles. Deep learning ⊂ machine learning ⊂ AI.
Not quite — picture nested circles. Deep learning sits inside machine learning, which sits inside AI.
Open the full lesson: AI, ML & Deep Learning
Glossary: ML Glossary: Deep Learning
IntermediateInteractive lesson ✓

Neural networks, demystified

Neurons, weights, and layers — an animated forward pass showing how a network turns inputs into a prediction.

Diagram + forward passOpen lesson →
NoviceModule coming

Supervised vs. unsupervised

The two main ways machines learn — with labeled answers, or by finding patterns on their own. Sortable example cards.

Sorting interactiveComing soon
IntermediateModule coming

Why training data is everything

Garbage in, garbage out — how data quality, bias, and volume shape what a model can and can't do.

Bias explorerComing soon
Track 02

Language & Generation

How large language models read text, predict what comes next, and sometimes confidently make things up.

02
Novice 4 min Learned

How a large language model actually works

It doesn't "know" things — it breaks text into tokens and predicts the most likely next one.

An LLM reads by chopping your text into tokens — chunks roughly the size of a word or word-piece. It then does one thing astonishingly well: predict the next token, over and over, to build a response. There's no database of facts being looked up; it's pattern-completion learned from enormous amounts of text.

  • Tokens are the model's unit of cost and context — billing and limits are measured in tokens, not words.
  • "Generation" is just next-token prediction repeated until the answer is done.
  • Because it predicts plausibility, a model can sound certain while being wrong.
InteractiveType to see tokens
≈ Tokens: 0 Characters: 0 Words: 0

Simplified illustration. Real models use learned sub-word pieces (e.g. BPE) that don't line up with letters or word-length — so treat this as "roughly how text gets chopped up," not an exact token count.

Predict the next token: “The weather today is very ___”
hot
38%
cold
27%
nice
19%
unpredictable
9%

Illustrative probabilities, for intuition only — to show that the model ranks candidates rather than "knowing" one answer.

What is an LLM fundamentally doing when it writes a reply?

Right. It's next-token prediction all the way down — which is why grounding (see RAG) matters.
Not quite — there's no lookup. It predicts the next token by plausibility, learned from training text.
Open the full lesson: How LLMs Work
Glossary: Token Glossary: LLM
NoviceInteractive lesson ✓

Generative AI vs. everything else

Text, images, audio, code — what "generative" means and where it differs from older predictive AI.

Discriminative vs generativeOpen lesson →
NoviceInteractive lesson ✓

Prompt engineering basics

Why how you ask changes what you get — plus chain-of-thought, few-shot & zero-shot deep dives.

Builder + techniquesOpen lesson →
IntermediateInteractive lesson ✓

Hallucinations & confabulation

Why confident, wrong answers happen — and how to spot them. A "real or made-up?" spotter game.

Spotter gameOpen lesson →
IntermediateInteractive lesson ✓

How transformers work

Attention in plain terms — click a word and watch which other words it attends to. The architecture behind every modern LLM.

Attention visualizerOpen lesson →
IntermediateInteractive lesson ✓

Embeddings & vector search

How meaning becomes numbers — explore a 2-D map where similar words cluster, and see the search behind RAG.

Embedding-space explorerOpen lesson →
NoviceInteractive lesson ✓

Tokens & context windows

What a token really is and why context limits matter — type into a live tokenizer and watch the window fill.

Live tokenizerOpen lesson →
IntermediateInteractive lesson ✓

Diffusion & image generation

How text becomes a picture — step through a denoising animation from noise to image, and see how the prompt steers it.

Denoising visualizerOpen lesson →
Track 03

Applied & Agentic

Where models become useful systems — grounded with your data (RAG), given tools and goals (agents), and connected through open standards (MCP).

03
Intermediate 5 min Learned

RAG: giving a model your sources

Retrieval-Augmented Generation — how to make a model answer from real documents instead of memory.

Retrieval-Augmented Generation fixes the LLM's biggest weakness — answering from fuzzy memory — by fetching relevant documents first and handing them to the model alongside your question. The model then answers grounded in those sources, so it can cite, stay current, and make things up far less often.

  • Your question is matched against a knowledge base to retrieve the most relevant chunks.
  • Those chunks are added to the prompt — the model reads them before answering.
  • Result: current, source-backed answers without retraining the model.
InteractiveRun the pipeline
Your question
"What's our refund policy?"
Search sources
Match against your docs
Retrieve chunks
Top relevant passages
Augment prompt
Question + sources
Grounded answer
Source-backed reply

✕ Without RAG

The model answers from memory. If your policy changed last week, it won't know — and may invent a confident, wrong answer.

✓ With RAG

The model reads your actual policy doc first, then answers — current, specific, and traceable to the source.

What problem does RAG mainly solve?

Correct — retrieve first, then generate. Grounding is how you cut hallucinations and stay current.
Not quite — RAG doesn't retrain anything. It retrieves relevant docs and feeds them in at answer time.
Open the full lesson: RAG
Glossary: RAG Glossary: Embeddings
04
Intermediate 5 min Learned

AI agents & the think-act-observe loop

What turns a chatbot into an agent: a goal, tools, and a loop that decides its own next step.

A plain chatbot answers once. An agent is given a goal and a set of tools, then runs a loop: it thinks about what to do, acts by calling a tool, observes the result, and repeats until the goal is met. Agentic AI is this pattern scaled up — multiple steps, sometimes multiple cooperating agents, with far less hand-holding.

  • Tools are how an agent affects the world: search, send email, query a database, run code.
  • The loop is what makes it autonomous — it chooses its own next action from each observation.
  • Autonomy raises the stakes: guardrails and human approval gates matter for risky actions.
InteractiveStep through a task
Think
Act
Observe
0steps
Goal: "Schedule a 30-min intro call with a new lead next week."
ThinkI need the lead's availability before I can book anything.
ActCall calendar tool → find open 30-min slots next week.
ObserveTue 2:00pm and Wed 10:30am are free.
ActSend email tool → propose both times to the lead.
ObserveLead replied: "Wednesday works."
DoneBooked Wed 10:30am and sent the invite. Goal met.

What most distinguishes an agent from a regular chatbot?

Exactly — goal + tools + a think-act-observe loop. That autonomy is why guardrails matter.
Not quite — it's the loop and tool use toward a goal, not just a bigger model, that makes it an agent.
Open the full lesson: AI agents & agentic AI
Glossary: AI Agent Glossary: Agentic AI
IntermediateInteractive lesson ✓

MCP: the "USB-C for AI"

The Model Context Protocol — plug agents into tools & data through one connector. Full interactive lesson built.

Connector + handshake + wizardOpen lesson →
NoviceInteractive lesson ✓

Chatbots, then and now

From scripted decision trees to LLM-powered assistants — what changed, and what each is good for.

Then/now + anatomyOpen lesson →
IntermediateInteractive lesson ✓

Reading a model card

How to evaluate a model's intended use, limits, and risks before you trust it. Annotated sample card.

Annotated cardOpen lesson →
IntermediateInteractive lesson ✓

Multi-agent systems

How specialized agents collaborate — supervisor, pipeline & mesh topologies, and the controls that keep them safe.

Topologies + step-throughOpen lesson →
IntermediateInteractive lesson ✓

Fine-tuning & customizing models

Prompt, RAG, or fine-tune? Pick a goal and the decision explorer shows which method fits — and what each costs.

Decision explorerOpen lesson →
Track 04

Governance & Safety

Using AI responsibly — the policies, frameworks, and controls that keep it safe, accountable, and compliant.

IntermediateInteractive lesson ✓

AI governance basics

What governance is, the framework landscape (EU AI Act, ISO 42001, NIST AI RMF) and the core practices teams actually do.

Framework stack + functionsOpen lesson →
IntermediateModule coming

EU AI Act, in plain English

How the risk-tiered law classifies AI systems and what each tier means for builders and buyers.

Risk-tier explorerComing soon
IntermediateModule coming

Build an AI inventory & risk register

The first practical governance step — find where AI is used, score the risk, assign an owner.

Interactive templateComing soon
IntermediateInteractive lesson ✓

AI evaluation & benchmarks

What MMLU, SWE-bench & LMArena really measure — explore the benchmarks and see how contamination inflates a score.

Benchmark explorerOpen lesson →

Hit a term you don't know?

The AI Glossary has the rest

Every concept on this page links into a searchable glossary of AI terms — plain-language definitions, difficulty tags, and related reading.

Open the AI Glossary

Where to go from here

You've got the concepts. Now apply them — with deeper guides, hands-on tools, and structured training.

The AI Tools Hub

Vendor-by-vendor breakdowns, comparisons, and how-to guides — ChatGPT, Gemini, Copilot, Claude and more, every claim sourced.

Browse the hub →

Certifications & study guides

Take it further with structured courses and practice questions across security, AI governance, cloud, and risk.

See training →

Talk to a practitioner

Need help applying AI safely in your org? Our consulting team works in governance, risk, and secure deployment.

Explore solutions →