Your winner log is evidence. The toolkit is policy — what you'll actually open on a Monday morning.
Map lab tasks → real jobs
| Your Lesson 10 task | Job type (toolkit row) |
|---|---|
| 1 Social posts | Quick marketing copy |
| 2 Flash-day table | Structured planning / tables |
| 3 Late-payment email | Customer tone / sensitive comms |
| 4 Image (optional) | Images / visuals |
For each of those rows, your log already gave you a Winner and a one-line why. That's your go-to app to start from.
Add the rows the lab didn't cover
Most people need these too, even if you didn't test them:
| Job type | Typical approach |
|---|---|
| Deep research (multi-source report) | Research / Deep Research mode on whichever app won you on hard tasks — Lesson 7 map |
| Explain / learn | Fast chat on whichever app you already like |
| Word / Excel / Outlook | Microsoft 365 Copilot, if you pay for M365 — not the same job as ChatGPT in a browser |
| Red-tier private data | No cloud app — Lesson 8 — build on your own PC later (Off the Grid onward) |
"Which AI when" in one sentence
Open the app that already won that job type in your lab — on the mode you noted — and fall back to a backup when it's down, capped, or just wrong.
If every task ended up with a different winner, that's completely normal. Subscriptions stack because the bundles genuinely differ, not because you're bad at choosing.
I badly wanted one hero app — pure ego. The log never once cooperated. Different winners per task turned out to be the useful result: it's the thing that tells you what to cancel later and what's worth keeping.
Continue — fill the card.