Academy
The Four Tools
I cut 43 of 47 AI tools. Four stayed, each with one clear job. The tool changes every year; the role does not. Below: the four, the filter that cut the rest, and a prompt that runs any new tool through it.
Four roles, not forty tools
Claude
One sentence for the idea, and the brief written before any program opens.
Nano Banana or ChatGPT
Builds the base picture: character, place, light. The same face and the same light every time.
Google Flow · Veo 3.1
Moves the picture and adds the sound.
CapCut
Editing, on-screen text, music, export. Without it, nice shots and no film.
All four for about $20 a month: one Google AI Pro covers image and motion, Claude and CapCut run free.
Three questions, and no fourth
Every new tool goes through three questions. Fail one and it is out.
Does it solve a problem I have?
Not someone else’s, not a theoretical one. A problem that stood in front of me this week.
Does it really save time?
Count it in minutes: what I do today without it, and what changes with it. If the saving is smaller than the time to learn it, skip it.
Is the result actually better?
Same brief, same inputs, once with my current tool and once with the new one. Compare through the client’s eyes, not the maker’s.
The rule: fail one question and the tool is out. There is no “keep it for rare cases”. Rare cases are what fill your subscriptions.
The Tool Filter prompt
Paste it into ChatGPT, Claude or Gemini, then write the name of the tool you are considering after the last line. The AI asks you the questions, one at a time, and answers with one verdict: KEEP or CUT, and which of the four roles it would replace.
The Tool Filter You are my tool judge. Below this prompt I will write the name of ONE AI tool I am thinking about paying for or learning. Judge it with my filter. Do not praise it and do not repeat its marketing. My four roles, and the tool that holds each today: IDEA (Claude) · IMAGE (Nano Banana, inside Google AI Pro, or ChatGPT) · MOTION (Google Flow with Veo 3.1, inside Google AI Pro) · ASSEMBLY (CapCut). The whole stack costs about $20 a month. Ask me exactly three questions, one at a time, and wait for each answer: 1) Does it solve a problem I had THIS week? (Not someone else's problem, not a theoretical one.) 2) Does it really save time? Count it in minutes: what I do today without it, what changes with it, and how long it takes me to learn it. 3) Is the result actually better? Same brief, same inputs, once with my current tool and once with this one, judged through my client's eyes, not the maker's. Then ask a fourth question: which of the four roles would it replace? If it only sits on top of them, that counts as a fail. The rule: fail one question and the tool is out. There is no "keep it for rare cases": rare cases are what fill my subscriptions. Then answer in this format only: VERDICT: KEEP or CUT ROLE IT REPLACES: IDEA / IMAGE / MOTION / ASSEMBLY / NONE WHY: two sentences, using my answers. NEXT STEP: one action for this week (for KEEP, a test on a real brief; for CUT, what I keep using). Use only what I told you. If you do not know a fact about the tool, say so instead of guessing. The tool I am considering: