Nine weeks. You start by understanding how these systems actually work, and you finish running real growth experiments on LotsTech products.
How it works
A lesson unlocks here each morning (IST). I'll post in #announcements when it does. Read it, do the task at the bottom, and post three lines in #daily-log.
Ask in #ask-anything, publicly, any time. Not in my DMs — if you're wondering about something, three other people are too. I'm in there most evenings around 7:00 PM IST.
No compulsory meetings. We're spread across India, Pakistan and Taiwan and there's no hour that suits everyone, so this runs on writing. Roughly 15–25 hours a week; this first week is lighter while everyone settles in.
Nothing is graded. You're not expected to know any of this already — nobody in this batch does. Ask the basic question. In Week 1 there is genuinely no such thing as too basic.
Week 1 — What these systems actually are
1–6 SeptemberFour days on the fundamentals: what an AI is, how you talk to it, what makes it more than a chat window, and what changes about your work when you use it properly.
Reference — open these before any task on LotsSocial.
LotsSocial marketing context — the current nine notes. Read note 8 first: it lists what marketing may not claim. The copy the lessons used is at /lotssocial-sep.
The marketing context checklist — the method behind those notes, usable on any product.
Brewline — the practice company used in Weeks 1–2.
What an AI actually is
What an LLM is in plain language, why it sounds confident when it's wrong, and which of ChatGPT, Claude, Gemini and Grok to reach for. Includes your tool setup.
Prompts and context
Why most prompt-engineering advice is now out of date, what context actually is, and how to make the AI interview you before it answers.
Tools and connectors
How an assistant stops being a chat box and starts doing real work on your real accounts — what a connector is, where to find it in all five tools, and what it changes for marketing.
What "AI-native" actually means
How the work itself changes when you stop asking "how do I do this" and start asking "how do I get AI to do this" — and the 100 articles I deleted finding out.
Workflows
The five patterns almost every AI workflow is built from, and how to tell which one you need — plus context blocks, acceptance criteria and approval gates. Includes my real 19-item review system as a worked example.
Agents
What an agent actually is, the four parts every one has, when they are the right answer, and why 88% of agent projects never reach production. Plus how I plan with several models and make them argue.
Research
Where the context comes from. Six kinds of research, the source tiers that decide what you can act on, where AI research goes wrong — and our own research agent, Scout.
ICP and positioning
The two decisions everything else rests on. Dunford’s five components, the Moore statement, STP, and jobs to be done — made from yesterday’s evidence, not from a blank page.
The channel map
Where the words go — content, search, social, outreach and community, what each is genuinely for, and the specialist agent that owns each one.
Pages that convert
Where every channel delivers. Message match, leading with the problem, outcomes over features, and the claim inventory — with Herald, the agent that owns conversion copy.
The review gate
Deciding whether work should ship, and why the person who wrote it cannot. With Editor — and the checklist you will use for the rest of the program.
Content that earns attention
The asset that compounds, and the top-ten test that decides whether it does. Topics from real questions, the brief before the draft — with LotsBlog.
Search
Being found, being the answer, and being named by a machine. Intent, choosing winnable queries, and why 84% of AI citations come from earned media — with Ranker.
Social
Distribution and familiarity, not publication. Platform fit, one idea in five shapes, and why commenting beats posting when you have no audience.
The one channel that does not depend on an algorithm. Lifecycle over broadcast, the three questions every email must answer, and restraint as the job.
Outreach
Writing to strangers — the channel that works before you have an audience, and the one with the most ways to do real damage. Research only.
Community and directories
The recommendation you do not make yourself. Reading the rules first, judging fit honestly, and the reply worth posting if the product did not exist.
Measurement
A number with no decision attached is decoration. Instrumentation, the ladder, and the weekly review that forces a choice.
The first 30 days
Everything turned into one dated plan — sequenced, owned, measured, with stop rules. The last lesson.
What's coming after that
Published as we reach each week, so you always know what's next without being overwhelmed by what's far off.
Week 2 — Getting real work out of it
7–13 SeptemberJudging output and knowing exactly what's wrong with it. Research with AI. Turning one-off chats into workflows you can repeat.
Week 3 — Marketing with AI
14–20 SeptemberContent and social. SEO. Outreach. Running experiments and actually measuring them. Agents and approval gates.
Weeks 4–8 — Real LotsTech work
21 September – 25 OctoberActual products, actual growth problems. What you work on depends on your interests and on what you show me in the first three weeks.
Week 9 — Final project
26–30 OctoberYou document and present what you built, what you tried, what failed, and what you'd do next.