Taste: The last frontier of software development

In the age where ai tools can build just about anything you can imagine, the last bottleneck for businesses is ironically the first one: creativity. Humans have always been able to achieve any goal they set their mind to. We harnessed fire, water, and the wind. And somehow we got to using tiny things we cant see that we break apart to release an insane amount of energy which (depending on your local grid) powers the device that you are reading or listening on right now. Tell a person 100 years ago that everyone would have all the knowledge in the world in the palm of their hands they would call you a crazy person. 400 years ago they would burn you at the stake. And 3000 years ago they’d say: ”𓅓𓂋𓏏𓏭𓏏𓏤𓏥”. What these examples all have in common is that they thought it impossible, but here we are.

The Bannister effect

There is a phenomenon known as the “Bannister effect” in which the amount of people who achieve a goal rapidly increases once a pioneer breaks a barrier. Everest used to be unclimbable, now there are lines of people waiting for their turn to take pictures at the top and get that satisfying Strava screen for their instagram stories for the rest of the month. Or doing an Ironman used to be the pinnacle achievement for a multi sport athlete but now it seems like everyones boss, co-worker, or sister that has no business being that cracked has done one or more. The Bannister effect is active proof that us humans will fill the mold that we set ourselves. Now there are two types of people which respond very differently when they see someone do something they previously thought impossible. One thinks “I could never do that.” And the other thinks “Why not me?”

“Why not me?”

Why not? Really? Have you asked yourself that? If your reasoning boils down to “I’m just not someone who says ‘why not.’” Then allow me to get real meta here: why not try saying “why not?” Our minds are like water, we fill the mold we give ourselves so why not just give yourself the biggest, dopest, most amazing mold possible? You should take this same principle to how you build software. AI tools allow us to write any software in any stack and in a fraction of the time. It is like if Einstein spoke 100 languages, had 6mg zyns, and access to Huberman’s nootropic stack. You can make a personal tool that no one in the world will use but it replaces a 5 dollar monthly budgeting tool; or you can build an enterprise grade B2B software that lands you $40k/month contracts for a couple months until one of your clients builds the same thing 1% better and puts you out of business. You can quite literally build anything you can conjure up.

Taste, creativity, and AI-driven development.

I know it is tempting to ask your favorite LLM “Hey what can I make look better about this site” or “how can i make this blog post more engaging” but please I beg you for the sake of yourself and your audience: DO NOT OUTSOURCE YOUR CREATIVITY AND TASTE. I repeat: I know it is tempting but your creativity and taste is one of the things that makes you you . I am guilty of it in the past and I always end up getting frustrated or just being unhappy with the results. It might not happen in the moment but I will come back to a project a month later and say “Wow this looks like shit. Who the hell made this?” Mostly everyone has had their first wow moment with the /frontend-design skill and if you haven’t yet I recommend you try it out and just throw a landing page at it to get comfortable with the fact that ai can make beautiful designs without any sort of help from you. Now imagine what it can do with some help.

If you have a subscription and some extra usage there is no reason you shouldn’t spend some time guiding your agent to create a truly beautiful UI. It is not that hard, it just takes time, some creativity, and knowing what you want. If you want to make a background with 400 different svgs on your portfolio site you can! You want a liquid glass nav bar? That will be 20k tokens please!. How about an interactive website experience that lets each visitor change the layout of the site for them? You can have it done in a couple sessions! Because it is no longer you doing it, it is the agent implementing the changes you want the ceiling of creation is whatever you say it is. I am a little upset that I did not learn how to be a designer earlier in my career because now is the time for designers to shine—especially those with taste. (Notice the double hyphen rather than the em dash which has now been put on a no fly list due to AI abuse).

I mentioned the /frontend-design skill but I wanted to reiterate that this is a starting point, not the finish line. These ai skills created by people have the peoples opinions baked into them. A designers preferred /design command might be different from mine so I recommend that everyone reading this makes their own skill rather than pulling one off of X or the skills marketplace and plugging it in and expecting it to make designs perfectly how you want them. I try to deopinionate my skills that I use and I try to focus on the negative space rather than the postive (what not to do rather than what to do). My slop-tells skill is an example of this if you would like to get some creative juices flowing.

You don’t need to let a lack of design knowledge hold you back. If you want something on your site you can go to gemini, generate an image of what you were thinking about, then have an agent iterate on it until it matches the original. The bear on my site originally looked like this:

But after a couple iterations of draw, compare, review, draw, I ended up with this:

Pretty solid right? If you can think of it you can make it. And it is cheaper and easier to do so than ever. So please please please if you have ideas that you think would be cool to add to a project you are working on just do it! Don’t say you are running up against the limits of AI’s talent because I promise you you are not. Obviously claude didnt cook up that bear and wizard pair in one go with one prompt saying “do it no mistakes.” No, it was at least 6-7 iterations before I had something worth putting out there but I’d say that is a solid deal for something that would’ve probably taken me a week to do in figma on my own.

So go out there. Get creative, explore and see what is possible. Spice up your landing page, make your login as ostentatious as possible, make the navbar the most liquid liquid glass of all time. You might end up with a gem.