Still confused
about AI?
Ten straight answers.

September 2026 edition

The ten things everyone's confused about right now — which tool, what agents actually are, whether it's coming for your job. Answered straight, with the sources. Free, fifteen pages, no course at the end of it.

The September Ten

Free. Where should we send it?

Send me the ten answers — free →

$0 — no cardInstant downloadOne-click unsubscribeWe never sell your address

01 — What's inside

The ten questions. Answered today, dated on the cover.

Almost everything written about AI is out of date within a season and almost nobody admits it. Every figure in here was checked this month, every source is listed, and when it stops being true there'll be a new edition.

01What even is this, and why not just Google it?One retrieves pages. One produces work that didn't exist. Different jobs.
02Which one? ChatGPT, Claude, Gemini, Copilot, PerplexityThey're all about $20. Pick by where your work already lives.
03Cowork — what is it, and does it help me?The difference between one that explains it and one that does it.
04AI agents — what are they, actually?Genuinely useful and wildly oversold, with the numbers on both.
05Avatars — do I need one?No. And on a page about money, one will cost you.
06Can I build an app without knowing how to code?You can build one today. Keeping it running is another question.
07Connectors and MCP — what are they?Why the acronym is suddenly everywhere, and what to be careful with.
08Is it coming for my job?Answered with actual labour data, not vibes. It's stranger than both camps say.
09How do I make it sound like me?The three settings nobody switches on — and where your data goes.
10How do I know when it's lying to me?Stanford says 22–94%. Here's what that number actually measures.

02 — Here's one right now

No email required to read this.

Question 10, condensed. If it's not useful, don't give us your address — that seems fair.

10 — HOW DO I KNOW WHEN IT'S LYING TO ME?

THE SHORT ANSWER → You don't. Not from reading it.
Fluency and accuracy are unrelated, and that is the
entire problem.

Stanford's AI Index 2026 reports hallucination rates
from 22% to 94% across 26 leading models — but that
measures how often a model guesses instead of admitting
it doesn't know, not how often it's wrong overall.

Supply the document and the best models sit near 13.6%.

So the lesson isn't "AI is unreliable." It's this:
give it the source material and it gets dramatically
better. Ask it to recall from memory and it invents,
confidently, in complete sentences.

Then the obvious question

Thirty Hustles You Can Start Now.

Thirty real services you can sell using what you just learned. For each: what it genuinely pays (sourced rate data), how crowded it already is, which AI does the heavy part, and how long until a first client — including the eight I'd tell you to skip.

$9 See what's inside →

03 — Fair questions

Asked honestly.

What's the catch?

There's a $9 guide to thirty ways to earn with this. That's the entire business model, stated plainly. The September Ten is genuinely free and genuinely useful — if it weren't, you'd have no reason to believe the paid one is.

Will this make me money?

No idea. Tools don't make money; delivering work people want does. What these do is compress how long that work takes. Anyone promising you a figure without knowing your skills, hours, market and network is guessing.

Do I need to be technical?

No. It's written in plain English for someone starting from zero, and every tool it mentions has a free tier you can start on today for nothing.

How many emails will you send me?

Roughly one a week. The first three have nothing to click. Unsubscribe works instantly and nobody chases you.

Ten answers.
Zero dollars.

Fifteen pages, every number sourced, updated every month. Decide later whether the thirty hustles are worth nine dollars.