Trend Tuesday: Teaching AI Skills Changes Collaboration
Art by @basilonmypizza: https://lnkd.in/eF8FkWzN - https://basilhefti.ch/
Remember how you learned to ride a bike? With the help of your parents or a friend, you acquired a skill through repetition and feedback (well, falling, actually ;).
Skills are also the AI trend this week. Since Claude introduced skills in October, the concept is now appearing elsewhere. This week, skills showed up in ChatGPT.
So what are skills?
At a technical level, they are text files. Cookbooks. They read like prompts. For example: “Ensure Excel formulas are calculated without errors and cached before providing the spreadsheet to the user.” The trick: these instructions are stored, named, discoverable, and reusable. Given a specific problem, an LLM can select the relevant cookbook and follow it. Until now, the user had to provide such instructions each time. A good skill library expands what an LLM can handle. It modularizes AI behavior.
This highlights an important distinction. Knowledge is knowing how something works. Skill, for humans, is reliable execution learned through training. Riding a bike is the classic example. LLM skills don’t learn this way. They approximate the procedural part by applying the defined steps.
In that sense, skills encode domain knowledge into executable procedures.
This also reinforces a broader point: English can function as a programming layer. Not in the traditional sense (nothing is compiled), but a way to express instructions and requirements that are interpreted and eventually result in executable code.
In other words, we move from one-off requests (“do this”) toward structured delegation. Skills turn expertise into reusable, executable behavior. That enables a different way of organizing our work.
What skill took you longer to learn than you expected?
• Art: https://lnkd.in/eF8FkWzN https://basilhefti.ch/
