Personal Productivity

How Khe Hy Built an Ambition-Upgrade Skill for Prompts and Tasks

Built by Khe Hy · · Source: X

Quick answer

Khe Hy built an "ambition-upgrade" skill that checks any prompt, task or query and suggests how to make it more ambitious. It is based on Felix Rieseberg's four-part framework: speed, altitude, depth and range.

Result: Khe Hy says they've been using the skill a lot. It's available in the LaTour skill library.

What they built

Khe Hy, who runs the AI training firm LaTour, built a skill called ambition-upgrade. It looks at a prompt, task or query and suggests ways to make it more ambitious. The idea comes from Felix Rieseberg's essay How to be more ambitious: A tutorial for humans with models, which splits ambition into four parts: speed, altitude, depth and range.

How it works

The post explains the framework behind the skill but not how the skill is built inside. Here is what it says about each part:

  • Speed: models can code, write and summarize much faster than people.
  • Range: using knowledge from several fields, not just your own.
  • Depth: having the model think like a top expert. It names the raw source material, the methods and tools, and an outside check that would show the result is right. Then it rewrites the prompt to ask for that.
  • Altitude: asking "what's the goal beyond the goal?" Khe Hy suggests adding "so that" to the end of a task to find what actually matters.

Khe Hy says the skill came out of asking Opus how their newsletter could be more ambitious. Opus suggested always including something readers could use right away. They also describe applying altitude to LaTour's Claude skills training. Ideas include a screening tool that finds new skills, mining clients' chat history for workflows that repeat, reusable skill templates built with Opus 5.5, standard rubrics and simple evals. These are presented as possibilities, not finished work.

The result

Khe Hy says they've been using the skill a lot. It's available in the LaTour skill library.

How to build something similar

  • Write a skill that reviews a prompt against a short checklist, like the four types of ambition, before anything runs.
  • Have it ask "so that…?" over and over to find the bigger goal behind a task.
  • Add a depth step: name what an expert would start from and how they would check the result.
  • Have the skill return a rewritten prompt, not just advice, so you can use it right away.

The original post

Be more ambitious. But what does that actually mean? @felixrieseberg broke it down with a 4-part framework. (Oh and I also built a skill to put it into practice) https://x.com/i/article/2113482469173338112 “You need to be more ambitious with AI.” This advice regularly gets dispensed by engineers at the frontier labs. And I hate it. Not only does it make me feel permanently behind.1 But it reveals my own shortcomings. It’s really hard to move from linear thinking to exponential thinking. I’m not leveraging the “country of geniuses in a data center” enough. And I’m only using AI to build a faster horse. So when I hear I’m not being ambitious enough with AI usage, my first reaction is just tell me how. And now we have an answer. (And a skill that you can download.) The 4 types of ambition The advice comes to us from Felix Rieseberg in his new essay How to be more ambitious: A tutorial for humans with models. (Felix is the product manager for Claude Cowork.) He breaks ambition down into four parts: Speed Altitude Depth Range 📷 Speed is the obvious one — LLMs can code, write and synthesize information way faster than humans. Not much to add here. Range leverages the “geniuses in the data center.” It moves you beyond your specific field and “draws on multiple fields” of expertise. This reminds me of the cross-functional collaboration found in innovation consultancies like IDEO. Depth pushes the models to take on the persona of the expert and then go deeper onthe underlying data, tools and problem spaces. Felix gives a prompt to make your tasks go deeper: 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. But the advice that really landed with me was “letting the model operate at a higher abstraction” and giving it more altitude. What’s the goal beyond the goal? I love this question so much (that I geeked out on jobs to be done, milkshakes and Clayton Christensen.) What’s the goal beyond the goal? This is where our functional responsibilities prevent us from zooming out and assessing the big picture. We miss it because we’re too focused on the work in front of us. Or because it’s someone else’s responsibility to “think bigger.” AI can collapse this. Let’s take a simple example of summarizing a quarterly earnings review. What happens when you “move up” the level of abstraction to make this exercise more valuable? A useful approach is to add “so that” to the end of your task to identify what actually moves the needle. You can see this play out in the chart below:2 📷 At the end of the day, the so what for investment firms is to maximize returns. Claude can’t promise that specifically, but it can likely add value by adding more “altitude.” Applying this to my business I’ve been playing with this idea for LaTour and the role of “AI training firms.” We are hired to drive investment results as well. Teaching firms how to use AI is why LaTour gets hired. But the true value is if we enable our clients to perform better at all facets of their jobs. This means bringing them closer to implementation. Over the past few months, LaTour has been hired to help teams learn how to build Claude skills. A relatively straightforward training mandate. But what if we used altitude to deliver 5x more value. We could: Run a screening tool to identify new skills Search their chat history to extract repeatable workflows Use Opus 5.5 to build skill templates that they can reuse Create standardized rubrics that the skills reference Set up simple evals I’m getting dizzy from this altitude ;) In fact, I asked Opus how this newsletter could be more ambitious. It told me to always includ
Khe Hy (@khemaridh) · October 23, 2026 · View on X

Want to build something like this?

Join the Shipping Skool community free. Paid members get live bootcamp calls every week.

Join Shipping Skool — Free

More use cases

← All AI agent use cases