What Happens After the Course

Ten weeks end, the cohort disperses, and within six weeks most people are back to using AI as a slightly better search engine.

That is the default outcome, and it is not about motivation. It is about having no mechanism. This lesson is the mechanism.

---

1. The one thing

If you do nothing else on this page:

Put a recurring fifteen-minute slot in your calendar, weekly, called "AI review".

Everything else here is optional. That slot is the difference between having done a course and having changed how you work.

---

2. The weekly fifteen minutes

Same slot, every week. Five questions.

WEEK OF [DATE]

WHAT I USED AI FOR
[Including the times it did not help]

WHAT WORKED / WHAT DIDN'T

DID I DO ANYTHING 3+ TIMES?
[Candidates for a saved prompt or an automation]

AUTOMATION CHECK
[Did everything run? Anything misbehaving? Anything to turn off?]

PROMPT FIXES
[Any prompt I corrected the same way twice? Fix it at the source.]

ONE CHANGE FOR NEXT WEEK
[Just one]

The two questions that actually pay:

"Did I do anything three or more times?" — this is how your automation backlog builds itself from evidence rather than from ideas you liked the sound of.

"Any prompt I corrected the same way twice?" — if you keep telling it "shorter" or "no bullet points", that belongs in the prompt or your custom instructions, not in every conversation.

One change. A review that produces five changes produces zero.

---

3. The quarterly hour

Four times a year, not weekly:

• Re-run your prompt test sets. Models change under the same name; your prompts may be silently degrading.
• Review your tool stack and costs. What are you paying for and not using?
• Check the Tool Radar — has anything you depend on been superseded?
• Update your Digital Twin.
• Turn something off. There is always something.
• Re-read your own notes. You have forgotten half of what you learned.

---

4. Staying current without drowning

The problem is volume, not scarcity. AI news is produced faster than anyone can read it and most of it changes nothing.

Follow capabilities, not releases. A new version number is not news. "This is now reliable enough to build on" is news.

The test for any announcement: does this change something I would do differently tomorrow? Almost always no.

Two sources. Not ten. Pick two you will actually read, and unsubscribe from the rest. The marginal value of source three is near zero; the cost is real.

What to ignore

• Benchmarks and leaderboards — they move weekly and rarely change what you should do
• Model version numbers
• Anything predicting a date for AGI
• "10 AI tools that will change everything" — it is the same ten tools
• Anyone announcing that a technique is dead

What to watch

• A capability crossing from unreliable to reliable
• Price changes, which shift what is economic to automate
• Deprecations of something you depend on
• Standards, like MCP, that change how things connect
• Regulation in your jurisdiction
• Failures reported by people doing your kind of work — the most useful and least amplified category

---

5. The experiment log

The only reliable way to know whether something is useful is to try it on your own work.

Date	What I tried	On what task	Better than my current approach?	Keeping it?

One experiment a fortnight is plenty. The discipline is trying it on a real task. A tool that impresses in a demo and fails on your actual work is worse than no tool.

---

6. Protecting your judgement

The risk nobody warns you about: if AI does all your drafting and you only ever review, your ability to judge drafts erodes quietly. You will not notice until it matters.

• Keep doing some of it by hand. Not everything, and not for efficiency — for calibration.
• Periodically do a task both ways and compare. You are checking yourself, not the tool.
• Notice what you automated that you should not have. Some tasks were how you spotted things. Undoing an automation is a legitimate and underused move.

Ask yourself quarterly: could I still do this without the tool, if I had to?

---

7. The anxiety part

At some point you will feel that everyone else is further ahead. A few things that are true:

The gap between knowing about a tool and using it well is large, and most people are on the near side of it. Reading about agents is common; running one in production is rare.

Depth beats breadth. Someone who uses two tools extremely well is more capable than someone who has tried thirty.

The fundamentals barely move. Context, constraints, verification and judgement applied to every model released while this course was written, and will apply to the next ones.

Nobody is keeping up. The people who appear to be are specialising narrowly and ignoring the rest — which is the correct strategy, and it looks like mastery from outside.

---

8. Your setup

My weekly slot (day and time)
My quarterly slot
My two sources
Where my experiment log lives
What I will still do by hand

Put the weekly slot in the calendar before you close this page. A system that depends on remembering is not a system.

---

⚠️ Common Mistakes

• No recurring slot. The single reason this evaporates.
• Following ten sources. You will read none of them.
• Chasing every new tool. Try one a fortnight, on real work.
• Never turning anything off. Automations accumulate, keep costing money, and stop being useful long before anyone notices.
• Reviewing but never changing anything. One change, and actually do it.
• Letting the judgement atrophy. Keep doing some of it by hand.
• Mistaking the feeling of falling behind for evidence of it. Ask what you actually observed.

---

That is the course. You have built something real, measured it, documented it, and you have a mechanism for keeping it going. The certificate names your capstone and its measured result — which is the only part of any certificate that a sensible person finds interesting.