أربع أنواع من الطموح مع الـ AI: السرعة والارتفاع والعمق والمدىFour kinds of AI ambition: speed, altitude, depth and range

Key terms

Speed

The same work in less time. Felix calls it the bread and butter of AI's economic value and says it needs little creativity, so his article skips it.

Altitude

Going up a level of abstraction. You ask for the outcome you want, not the single task you guessed would get you there.

Depth

Doing the task the way a top expert in the field would. You go one level down and touch the data, tools and problems under the task.

Range

Drawing on knowledge, tools and techniques from several fields, not just the one the task seems to need on the surface.

Outside judge

A check on the result that is not the model grading itself. Ideally a tool the model can run to get feedback right away.

The steps

1. 1

   Step 1

   Pick one real task and give Claude everything

   Pick one thing you are actually working on this week. A report, a launch, a bug, a pitch.

   Felix says every kind of ambition he describes needs the model to have a computer that can install software and access to the relevant data, through MCP connectors or other means. So before any prompt, give Claude as much raw material as you can: files, exports, logs, links.

   The prompt below puts the task on the table in a form the next steps can build on.

   PromptSet up the task

   Copy

   Here is a task I am working on: [describe the task]. Here is everything I have about it: [paste or attach data, files, links]. Before doing anything, tell me in five lines: how you understand the task, what result I most likely want, and what data or access you are missing to do it really well.

   Tip: Only connect data you are allowed to share.

2. 2

   Step 2

   Speed: get the same work done faster

   Most of us already do this one. Same result, less time.

   Felix says it needs little creativity in how you approach AI, so his article skips it.

   Use it as your baseline. Get the plain version done fast, then use the next three kinds to go further.

   PromptSpeed (my prompt)

   Copy

   Do this task exactly the way I would, as fast as possible: [task]. Use my usual format: [format or an example of it]. Ask me nothing unless something blocks you.

3. 3

   Step 3

   Altitude: ask for the outcome, not the task

   Ask yourself what you want to happen next. What is the goal beyond the goal?

   Felix's example: he used to ask Claude to fix individual crashes. Later he asked it to simply keep the apps crash free. To make that work he first let it fetch crash reports, map each app version to its git commit, and reach Windows, macOS and Linux machines so it could reproduce and verify fixes on its own.

   His office example is a quarterly business review. He climbs in three steps: from what changed this quarter and why, to the three most urgent decisions, to designing a better way than a quarterly deck. Those three prompts are in the extras below.

   He says altitude works best when you are after a real outcome, and less well for creative work. A good sign it fits: there will be meetings about this work.

   PromptAltitude (my prompt)

   Copy

   Here is the task I usually give you: [task]. Before doing it, go one level up. Tell me what outcome this task really serves, and what the goal beyond that goal is. Then propose how you would own that outcome instead of the single task: what you would do, what access and data you need from me, and how we will know it is working. Wait for my OK before you start.

4. 4

   Step 4

   Depth: make it work like the best expert

   Depth means pushing the model to use the tools, techniques and knowledge of the best professionals in the field, until even they could learn something from your work.

   Felix's techniques, using a company losing customers as the example. Insist on the real instruments: when Claude reaches for markdown tables, insist on databases and statistical models. Pick goals that look unreasonable and tell the model they are reasonable: need a page in 5 seconds, ask for 200ms. Challenge what you know: give it the raw data and ask it to reproduce your assumption. Ask for new knowledge once it is deep in the data. And use an outside judge: do not ask the model to judge itself.

   This is the prompt he shares in the article, word for word. Paste it after your task.

   PromptDepth (Felix's prompt)

   Copy

   I want a deeper version of this task, not a bigger one. Before rewriting it: (1) say what a top professional in this field would do here that I'm not asking for; (2) name the rawest material they would start from instead of what has already been digested for me; (3) name the specific methods and tools they would use; (4) say what, other than you or me, would show the result is right. Then rewrite my prompt to ask for exactly that.
   
   [your task or current prompt]

   Tip: Models often switch vocabulary without switching methods. Watch the methods.

5. 5

   Step 5

   Range: bring in more fields

   Range means looking at your work through more fields than used to be worth the cost.

   Felix's first trick, from The Art of Innovation: replace the nouns with verbs. A quarterly review is not a deck. It is getting busy people to believe a story about cause and effect, and to decide under uncertainty, in 30 minutes. The verbs show which fields can help.

   His second: assemble the crew. For his deck he would bring an analyst, a journalist and an editor, maybe a motion and audio designer, plus a red team like a short seller and a forensic accountant.

   His warning: use this to find fields and techniques, not to role play. Do not ask the model to write like a short seller. Ask it to do what a short seller does to research a position.

   PromptRange (my prompt)

   Copy

   My task: [task]. Step 1: rewrite it as verbs, what it actually does for people, not what it is called. Step 2: list 5 fields beyond the obvious one whose methods would make the result better, with one specific technique from each. Step 3: name one adversarial field that would try to break the result. Step 4: apply those methods themselves, not the voice of those experts, and show me what changed and why.

6. 6

   Step 6

   Ask three questions about your own work

   This is my short version for daily use. Take one thing you are working on and ask: can I aim higher? Can I go deeper? Can I bring in another angle?

   These three questions are my summary, not a quote from Felix. Each maps to one kind: altitude, depth, range. Speed you already do.

   The prompt below runs all three on a real piece of work.

   PromptThree questions (my prompt)

   Copy

   Here is something I am working on: [paste the work or describe it]. Its current goal: [goal]. Ask three questions about it and answer each with specifics:
   1. Can I aim higher? What outcome sits above this task, and what would it take to own it?
   2. Can I go deeper? What would a top professional in this field do here, from which raw material, with which methods, and what outside check would prove the result right?
   3. Can I bring in another angle? Which two fields outside mine would change the result, and which technique from each?
   End with the one change I should make today, and the first prompt to start it.

7. 7

   Step 7

   Run all four kinds in one prompt

   When you want everything at once, use this. It walks one task through speed, altitude, depth and range, and gives you a plan before it builds anything.

   I wrote this prompt, built on the four kinds in Felix's article.

   PromptAll four (my prompt)

   Copy

   Task: [task]. Data and access you have: [list]. Work through four kinds of ambition, then wait for my pick.
   Speed: how to finish the task as it is in half the time.
   Altitude: the outcome above this task, and how you would own it.
   Depth: what a top expert would do differently, the rawest data they would start from, their methods and tools, and an outside check that is not you judging yourself.
   Range: the task rewritten as verbs, and three fields beyond the obvious one with one technique each.
   For each kind write one paragraph and the first concrete step. Then recommend which one to run first and why.

Ready prompts to copy

Tell me what changed this quarter, why, and what we should do about it.

Work out which three decisions are most urgent in the face of everything you've learned, and start working on potential directions with for/against arguments.

The review exists so leadership catches problems early and moves money and people to the right places. Design a better way to do that than a quarterly deck, show from our own history that it would have caught things sooner, and run it beside the current one for a quarter.

We believe churn is an onboarding problem. Try to reproduce that from the raw data.

Build a simulation of our customers that reproduces the last year from our logs. Then find the earliest signal of churn we don't know about yet, and show it predicts on data it has not seen.