Tan Tan 🌵

The way products need to be designed is changing, and most people haven't adapted yet. The easier and quicker it is to make software, the more customized it can be for an individual or business. Making it more important to understand the problem that is being solved more deeply, instead of just shipping features. I've been seeing lots of posts lately calling out taste and judgment as the main value of humans vs models, and of designers specifically. While there is some truth to that, it's not the full story. Choosing the best options from what a model outputs is important, but isn't really the most impactful thing that people can do, especially as the models get better at choosing the best options. The most important thing for people to do, especially designers, is to shape all the different parts of the product that a user experiences. This was true in the before times, and it's true now. Somewhere along the way, products being built all started re-using patterns, and the intuition around what a good experience became easier for more people to develop. Now with models creating new ways to interact, defining a product's "user experience" or "interaction" at every layer is the most important thing to get right. The user experience now lives at every layer, from how you structure the model's context to what buttons show up on screen. If you're only designing UI, you're designing maybe 20% of the product. Bolting AI on at the end will get you burned when your product makes mistakes and just doesn't do what it promises. You need to start from the behavior and build up from there.

Moving my explorations from my physical notebook to Figma made me realize that the direction felt a bit too flat for my liking. All the meaning that I was hoping the mark would represent was lost.

I keep intending to post about design, but the narratives around AI are too wild right now. This Daily Show clip kind of frightens me. youtube.com/watch?v=lnLDjXy2…

I watched the 1hr 20m video, and what stood out most to me was @noahrshinn talking about AI as not being a subordinate to do your dirty work. People look down on agents, but I believe they are partners and should be respected. Hence working “in tandem.” youtu.be/Am7IWP8IpEc?is=5qrc…

Given all the hype around Jev, I wanted to understand the thinking behind it a bit more, and found @CompleteSkeptic's talk from a few months ago, which I found to be insightful youtube.com/watch?v=cJ0EOzey…

AI is not going to give us free time. Just like the 1940’s with dishwashers, vacuums, and washing machines, we’re going to be expected to do more, not have more leisure. Even today, I hear people saying that they’ve been working more than they ever have.

what is going on with codex, is it down? does it need biscuits?

Being a product designer at heart, I sketched out interfaces to help me get in the mindset of the product and how the brand might show up. Not just the logo but the philosophy behind the brand.

Things got interesting once I started to pair the logo mark with the text. I was inspired with a variety of shapes.

I do plan to post more hands on work. Especially stuff that's work in progress. But right now only my thoughts and words are work in progress.

One of the things I'm going to do with this account is to share whenever I run into a struggle with deciding how to use AI when I'm designing or building. Also, any wins I have or when it just gets in the way. Anything else people wan to see?

When I first started sketching on paper, I was obsessed with the idea of showing two halves of the same thing. I liked the concept, but was also worried it was too simple.

I would love to shadow people who know how to create compelling content online. Similar to branding, it feels like a dark art. Maybe because I only see the main feed of what goes viral, and not the grunt work behind the scenes.

I know there is indiehackers, but I'm wondering if there are any communities of people who are designers who are building. Optimizing for quality and authenticity in what they build, and not in ARR.

I needed to create a visual mark that was both distinctive and related to my philosophy, so I started far away from anything digital and just started exploring

It really seems like we're not far off from one-shooting internal tools. Once your agent has all the connections and context, especially over many days, will there be a need to buy sass tools? Am I wrong?

I've built these internal tools: customer dashboards, LLM conversation views, daily digest updates on customer usage in slack, research reports on new customers, determining which users are ripe to be upgraded to premium features, etc... It's not THAT hard to build these well, but when I want to set them up so other people on my team can see the same data, and have their own settings, it starts to get complicated.

It's comically easy to build internal tools at work. The moment I have some interesting data or information, I can spin up a dashboard and spend a few hours tweaking it, and I feel satisfied. Which I think is behind this feeling of the sass-apocalypse. Why buy anything if you can just build it? The more I build these though, the more I feel like it's missing some soul. It's just a visual way to look at some data, like customer metrics. And I'm missing something more, but I'm not sure what, yet.

It might seem weird that I'm posting to this account as if I'm a person, when it's more of a "brand" or product. The reason is that I have a full-time job, and I don't want my employers to worry about me making my own thing. Maybe I shouldn't care, maybe they would take pride in it, but I want it to just be my own thing.

My theory on how to get around commodified software is that the hidden parts are the hard parts. Backend problems and infra that is complicated, takes time, testing, etc... are the real "moat."

For a long time, I've had this feeling that AI is some sort of snake oil. I haven't been able to articulate why or what I mean until now. With my AI clients, I'm seeing that what they need are mainly deterministic workflows that are reliable and produce consistent results, with some model calls used as needed. AI is used mainly for messy things like not knowing what files or file types will be coming into an input, or not knowing where you can get certain data from. The benefit of AI comes down to validating, interpreting, or handling information that isn't consistent. So when I see these agent platforms, like GrokBot or OpenClaw, I don't think the average worker in a business will have the time and space to make high-quality workflows using those platforms. It takes a shift in mindset, understanding good software principles, and just dogged persistence. Giving someone an agent platform is like giving someone a bunch of wood and ste and blueprints and cranes and asking them to build the building themselves. Sure, all the pieces are there, but putting them together takes more time and effort than you would think. It feels disingenuous to sell people the materials and promise the outcome without having the scaffolding in between. That’s the snake oil part.