The 100th Take on Gen AI You Didn't Ask For
There's an epic scene in The Matrix where Trinity learns how to fly a helicopter by downloading a module to her brain in seconds. I thought about this recently when I rapidly built software that would have taken a lot longer.
Like most knowledge workers in 2024, I use LLMs often, mostly to write first drafts, summarize content and for more efficient online searches. But recently, I started coding with help from Claude, and it's been a gamechanger. So far, I have built several things that would have taken me a lot longer: a case summarizer for a friend that saved her weeks of manual work, a podcast editing tool that removes host segments and replaces them with summaries in my voice, and a children's app that generates customized stories in the style of Beatrix Potter - all using a variety of different APIs. Each of these would have taken me weeks to figure out independently - and I never cared enough to spend hours learning syntax and debugging code*. * I still need to spend a decent amount of time checking the algorithm and debugging, but I think that will also improve soon.
This experience made me reconsider how we think about skilling today. We're at one of those rare technological inflection points and this is my take on what this means.
Any "skill" that is deterministic will soon become redundant. If you can describe a task to someone and expect a predictable outcome, assume that an LLM will take over it. This isn't limited to coding – it applies to design, data analysis, copywriting and more. The implication is that being "technical" in the traditional sense will matter less and less, becoming as quaint as memorizing phone numbers.
I think what remains uniquely human distills to three qualities:
1. Judgment: What are important, meaningful questions to explore?
2. Intuition: What are efficient ways to approach and solve problems?
3. Taste: What does "good" look like and what is your unique style?
Our education systems do not prioritize these abilities. Those who managed to pick up these skills have usually abstracted general principles while learning traditional subjects like Math and History - usually by accident and rarely deliberate.
What does this mean for us today? It's time to upskill, but not in the way we usually think about it. Instead of learning specific technical skills, we need to find ways to improve these abilities. Some ways I have been trying to do this:
1. Judgment: Exploring different idea spaces and understand their big questions. Build a baseline understanding to understand what their canons have and what they miss. I've found that reading and listening to podcasts gets me a decent distance.
2. Intuition: Improving how I think about systems, learning the basics of different mental models and analytical techniques. This is a helpful recent discovery
3. Taste: Creating more and developing taste by looking at more creations. Work with people who have high standards - it used to first annoy me, but then I internalized these standards and the quality of my creations improved. The TRIZ framing of creativity (which I heavily endorse) gets at the same idea.
Education has already changed in our lifetimes - with modern abstractions hiding skills that were previously considered important. From memorizing verses to learning long-form division to writing code – each generation had to adapt to new tools. Even modern software engineers do not know how to code in assembly (unless you're Jeff Dean). We're living through a similar transition, but at a much faster pace. How useful were horse riders when motor vehicles came along?
What will soon matter most is not the ability to code the best or design the prettiest interfaces. What