Reading, Writing & Building with AI
What fundamentally changed? Earlier, doing was thinking—
Reading - I read long form content pieces. I get a better grasp of the work. A point/ theme/ character, read in different narrative structures sinks in.
Writing - I write a draft, read it a few times and edit it a few more times. I become a better writer in the process. Is there a continuous natural flow of thought? What sequence of words are working well to communicate the intent behind the piece?
Building - I write code to come up with an app. I become a better programmer in the process. How are the if-else loops, DOM node types, CSS attributes change the end-result (looks, performance, value)?
Now, with AI, doing became faster (and often, better) compared to doing it myself. So what’s the problem? I risk losing the retraining process.
If all I am reading is a summary, I am losing out on repetition, immersion, relatability and nuance.
If I am relying on AI to convert half-baked thoughts into a cohesive structure, I am not learning the skill to get clarity from chaos.
If all I am writing is 10 words to produce a complete app, I am not honing the skill of improving one particular feature better than anyone else.
The good news is that it is not a new problem. Some chains of thought—
Abstractions:
Minimizing control - Earlier, people wrote all the CSS by themselves. Then came Bootstrap. Within couple years, almost every website looked the same. So people went back to writing their own CSS because what mattered was taste.
Replacing control - Earlier, everyone wrote low level code (Assembly/ C++). Then came programming languages (JS) and then little later frameworks (React). These abstractions mostly survived the longevity test because they improved what came before while still giving a playground to customize for variables that matter to the craft - taste and performance. (This is the reason, I think, TailwindCSS will stay longer than it’s predecessors)
Volume related changes:
Transferring weight in the same game - If something is available in new quantities, people value the nicher aspects of the same quantity. Until books were scarce, any book available was a game changer (Secret Seven, Harry Potter were the only ones available to me back in my schooling. So I read whatever was available). But when procuring a book became as easy as tapping a button, I care more about the author, plot and genre. Similar patterns are there in photography (we now care more about how “real” the pics look/ candid shots) and coding (performance of the app/ better UX).
Replacing the game - If something is produceable in greater volumes easily without much difference in any factor of the output, then the game gets completely replaced. If everyone can produce similar rice grain, then there’s no point optimizing for any variable of rice. You instead shift the focus to what kind of food you can cook with the same rice (farming → cooks → restaurants). Similar shift has been happening to one aspect of reading. Reading to accumulate facts. This is the game that is first replaced by AI.
So how do I make sure I still think?
Writing - Don’t use AI to write. Especially the pieces where I write to think. Ex: life planning, product roadmapping, architecture, structured thinking and mental models. Any use of AI here should only help produce neural pathways that didn’t exist before.
Reading - Use summaries with caution. Summaries should help me narrow down to the content pieces that I want to read in detail. Or use them when I would otherwise never read the original piece (translation, technical language pieces).
Building
Think in the new programming language (English). AI coding is a new programming language and a new framework in one. A few lines of prompt would one-shot an app. But it's the same app everyone is one-shotting. To cater to other variables like taste and performance, I will have to write longer prompts. Understanding what word in prompt would create the desired taste/ performance in an existing app are the new neural pathways to learn. So write, read and observe prompts. Practice writing detailed docs.
Learn other skills. Marketing/ Sales/ GTM had mattered always and will likely matter even more. Learn the core aspects of these skills. Mass emailing the same email to 1000 without customization worked when not everyone could send those 1000 emails. Now that everyone could do that (or newer addition, make 1000 AI calls) proving that an actual human took time to write that email/ make that call will matter.
Develop taste/ product thinking. If everyone can produce the same business website. Standing out with taste will matter more than the decade before. Farming → Cooking.
Stick with the nuance. Not everyone needed to write C++ code. But the very few who are really good at it and cared about it, are still highly valued. This will remain true for quite sometime. So, as long as you are really good at existing frameworks and continue to do so, there will be a need for it.
