01 · ASK
Messy words arrive.
Pidgin, no punctuation, a typo. Real customers write like this — and a good agent must still understand.
Where it breaks: it misreads the question and answers something else.
Welcome from the video — you are in the right place.
Lesson 00 · Free Pilot · Cohort 1
Stop justusing AI.Build it.
This website is your first lesson. In the next ten minutes you will trace how an AI agent thinks, test one yourself, write your first agent brief and score your readiness — then decide if you want the full free cohort.
Chapter 01 · The journey
That instant answer
That instant answer
took a journey.
Follow one real WhatsApp message to a fabric shop in Onitsha as it moves through the six moves every agent is made of. Only one of the six is the AI. The other five are engineering — and that is where things break.
Live traceMove 1 / 6
“abeg how far, una dey open for Sunday?”
Teaching modelAI involved: no
02 · READ
Fetch the facts.
“Mon–Sat 8am–6pm. Closed Sunday.” This comes from the brief the owner wrote — not from the model's memory.
Where it breaks: the fact is missing or out of date, so the model guesses.
03 · THINK
Decide what to do.
The model maps messy words to a clear intent: opening hours, for Sunday. This is the only move that is “AI”.
Where it breaks: it guesses instead of admitting it is unsure.
04 · REACH
Use a tool — or don't.
This question needs no tool. Good agents only reach for a database, calendar or API when they must.
Where it breaks: a slow tool, a network error, a missing permission.
05 · DO
Change the world.
Here, the only action is sending: “We're closed on Sunday. Open Mon–Sat, 8am–6pm.” Bookings and refunds would need a human gate.
Where it breaks: it promises something it cannot check.
06 · CHECK
Prove what happened.
Log the question, the answer and the source used. Tomorrow, when someone asks “what did it say?”, you have evidence.
Where it breaks: no log — and no way to defend your system.
Chapter 02 · The gap
Users hope.
Users hope.
Builders know.
Drag the handle. On the left is how most people experience AI. On the right is how the people who get hired, paid and trusted see the same thing.
AI user
Sees the answer.
- Asks and copies
- Accepts what sounds right
- “The AI is broken”
- Waits for the next tool
AI builder
Traces the system.
- Writes the brief and the limits
- Tests what it must refuse
- “Step 4 timed out — fix the tool”
- Ships, measures, gets paid
← drag or use arrow keys →
Chapter 03 · Live lab
Test an agent.
Test an agent.
Then break it.
Pick a business, send a message the way real customers type, and watch the six moves light up. Then switch the brief off and see what happens to the same question.
Simulation · rule-based, no live AI
ASK
READ
THINK
REACH
DO
CHECK
MN
Mama Nkechi FabricsAssistant · replies instantly
> system ready · pick a message
Chapter 04 · Build your first agent
The whole agent,
The whole agent,
on one page.
Fill this in for any real business — yours, your church, your school, a relative's shop. Copy the result into ChatGPT, Claude, Gemini or any assistant. You will have a working agent in four minutes.
Your brief · live0 / 7 checks
Next: open a new chat in your assistant, paste the brief, add “Reply only as this assistant from now on.” Then run the nine tests from the lab.
Chapter 05 · Engineering judgement
Should this even
Should this even
be an AI agent?
Think of one task you want to automate. Answer four questions. Most ideas stop at the first or second — and saying no quickly is a skill clients pay for.
Q1
Is the rule you follow writable in one sentence?
Q2
Does it involve messy human language, or judging something?
Q3
Would a wrong answer be embarrassing, expensive or dangerous?
Q4
Can you describe what “done well” looks like in one sentence?
Chapter 06 · Money
Do the sum
Do the sum
before the pitch.
You are not selling an agent. You are selling back somebody's time. Move the sliders for a real business you know.
6
₦2,500
1
Ask the client for these numbers — never guess. If they cannot tell you, that is your first finding.
The size of the problem, per year
₦780,000
312 hours of human time
Set-up fee · ~10–15%₦78,000+
Monthly support · ~2–3%₦15,600+
Illustrative rule of thumb for teaching, not financial advice. Price as a share of the problem — never as an hourly rate for your typing.
Chapter 07 · Your scorecard
AI user or
AI user or
AI builder?
Fifteen quick questions across five pillars. Get your AI Builder Readiness Score, see exactly where you are strong and where you are exposed — and what to do next.
Superintelligence: why the next 5 months could change work
Chapter 08 · Why now
The work is
The work is
changing fast.
A 60-second briefing from our channel. The point is not fear — it is timing. The people who understand what is underneath these systems will design how they are used here.
Chapter 09 · The curriculum
AI is not a black box.
AI is not a black box.
It is a stack.
Tap each layer. You will learn them in order, from the machine underneath to the evidence that proves your system worked — building one continuous project, Project Atlas, the whole way.
WEEK 1 · WHY
The engineer
- The AI economy and your place in it
- What an AI Systems Engineer does
- Your first agent brief
WEEK 2 · HOW
The system
- The machine, the network
- APIs, REST and JSON
- Data: SQL, NoSQL, vectors
WEEK 3 · BUILD
The builder
- Python from zero
- Calling a model from code
- Tools and structured output
WEEK 4 · SHIP
The shipper
- Git and GitHub
- Deploying to the web
- Measuring and presenting
MONTH 1
Foundations
- Computers, networks, APIs, data
- Python and Git
- Deploy your first app
MONTH 2
Models & knowledge
- How LLMs work — OpenAI, Claude, Gemini, Llama, DeepSeek, Qwen, Mistral
- Prompting as specification
- RAG, embeddings and evaluation
MONTH 3
Agents in production
- Tools, MCP, LangGraph, CrewAI, n8n
- Guardrails, security, human approval
- Ship to WhatsApp, web and email
MACHINE
NETWORK
API
DATA
MODEL
TOOLS
AGENT
DEPLOY
Project Atlas · one system you grow from lesson one to launch
THE AI AGENT STARTER KIT
Chapter 10 · Free book
Start before
Start before
class.
The AI Agent Starter Kit — Cohort 1 Edition. 32 pages. Two agents to build, a brief worksheet, a test record, ten agents businesses will pay for, and a 7-day plan. Yours to keep and share.
- Why I am giving this away
- Three rooms this changes
- What an AI agent actually is
- The six moves
- Trace one message
- The five-step method
- Should this even be an agent?
- Build one: the everyday agent
- The one-page brief
- Nine test questions
- Before and after
- Four prompt patterns
- Worksheets: brief & test record
- Build two: the document agent
- Script first, model second
- Six ways agents fail
- Limits, privacy and safety
- Do the sum before the pitch
- Ten agents businesses pay for
- Your first seven days
- The 90-day path
- Plain meanings
- Evidence beats certificates
PDF · 32 pages · works offline · no sign-up needed
Chapter 11 · Who belongs here
Done watching
Done watching
from outside.
You do not need a computer science degree. You need one real problem you are tired of doing by hand, and about thirty minutes a day. Tap a card to see the “after”.
The shopkeeper
Loses an hour a day
Answers the same six questions on WhatsApp — price, hours, location, delivery.
Tap for after →After
Gets the hour back
A brief-driven assistant answers the six questions and hands real orders to her.
The office staffer
Retypes everything
Copies the same information into three systems every morning.
Tap for after →After
Becomes the fixer
Designs a script-first workflow and becomes the person the company relies on.
The graduate
Uses, can't explain
Can use the tools, but cannot yet explain what is underneath them in an interview.
Tap for after →After
Shows a portfolio
Walks in with a brief, a working agent, a test record and a value sum.
The founder
Tired of demos
Knows AI matters, has seen the hype, wants something that actually runs.
Tap for after →After
Ships a product
Prototypes around a real pain, prices it by value and knows where it breaks.
Chapter 12 · Why this exists
A note from the founder
I work in rooms with hospitals, insurers and operations teams almost every week. I see brilliant people doing work a machine could do, and organisations paying for software nobody understands well enough to fix. People who can only use AI are at the mercy of the tool. People who understand what is underneath it can build, fix, judge — and sell. This pilot is how I close that gap.
Azuka
Founder · Deary Tech Enterprise
How we teach
- Problem first, tool second
- Systems, not prompt tricks
- We break things on purpose
- Every lesson leaves evidence
- Platform-neutral: OpenAI, Claude, Gemini and open models
- Real cases: clinics, schools, churches, SMEs
Chapter 13 · Quick answersBefore you
Before you
join.
Is Cohort 1 really free?
Yes. The pilot costs nothing. I want real learners to test the curriculum, find the confusing parts and help make it stronger before the full programme launches.
Do I need to know how to code?
No. Weeks 1 and 2 need no code at all. Python starts in Week 3, from zero, one step at a time. Curiosity and one real problem are enough.
How much time do I need each week?
Plan for about 5–7 hours a week: a guided lesson, a hands-on lab and a short assignment. You can go faster if you have more time.
How is the pilot delivered?
Online, through guided lessons, practical labs and assignments like the ones on this page. The exact class schedule is shared with registered learners on WhatsApp and email.
Do I need a laptop?
A laptop is strongly recommended from Week 3, when we start coding and deploying. Weeks 1 and 2 are manageable on a phone.
Which AI tools will I learn?
The programme is platform-neutral. You will learn the ideas that transfer across OpenAI, Claude, Gemini, Llama, DeepSeek, Qwen, Mistral and open-source models, plus tools like n8n, LangGraph, CrewAI and MCP later in the path.
What will I have at the end?
Evidence: a written agent brief, a working agent, a test record, a value calculation for a real business, and your Project Atlas work — a portfolio that is more persuasive than a certificate.
Is the agent on this page real AI?
No — it is a rule-based teaching simulation, clearly labelled. It shows exactly how a brief changes an agent's behaviour, without sending your words to any AI service.
What happens to my details?
They are used only to organise the cohort and contact you about it. We never sell your data, and we follow the spirit of the Nigeria Data Protection Act 2023.
When does registration close?
Cohort 1 registration closes 30 November 2026, 11:59pm (WAT). Places are limited because the pilot is hands-on.
Registration open
Reserve your
Prefer WhatsApp? Chat with Azuka
Reserve your
free seat.
Free Pilot — Cohort 1. Four short steps. We use your answers to shape the cohort around real learners.
--Days
--Hours
--Mins
--Secs
Seat reserved
You are in motion.
1. Save Azuka's WhatsApp so you don't miss cohort updates.
2. Download the Starter Kit and do Day 1 tonight.
3. Watch your email and WhatsApp for your welcome message.
You just finished Lesson 00
Stop just using AI.