Hà Phan

I’ve always considered my experience leading teams to build a Search platform from scratch, one of the most valuable experiences of my career. It gave me a platform view on how data products powered by algorithms are built and scaled. It also made me much more empathetic to an eng perspective.

Regardless of all the noise about vibe coding, I think there are actually different kinds of designers needed:
1. Solution designers (Deterministic/known constraints)
2. Design Systems Designers
3. Service Designers/Value Chain Architects
4. Frontier Designers (Design as inquiry)
5. Brand Designers

If you’re applying to a service design job, make sure you show the service design in your portfolio, not just beautiful pixel pushing. That is all.

Best portfolio I saw this week was on Google slides. Masterful storytelling, clear visualization of a highly complex system & workflows, and the final execution of the experience was crisp. We have a similar Human-in-the-loop operation so I understood the complexity.

The best account I saw on TikTok was this guy who bought a crab and some prawns at an Asian food store and tried to raise them as pets. Not all this AI bullshit.

It never ceases to amaze me what prototyping 3D experiences reveals. You can never fully understand spatial experiences without dog-fooding the interaction and movement. There was an insight I had about a behavior and to see the team exclaim delight seeing it come to life is an amazing feeling.

Painfully true. Design systems is the hill they’ll always die on.

There’s a difference between shaping what the emergence tech can do and should do vs receiving the constraints downstream.

Designers understand systems thinking in terms of patterns and scaling. But systems thinking in terms of experiences with automation is about understanding the ecosystem, the ranges of inputs and outputs.

Something I learned from working on products with probabilistic model is that part of the process is to run the model against a range of samples or use cases. We do this in cycles and we can segment the kinds of problems. Also you have to build feedback into the operations of AI products.

UX for AI is cannot be designed or derived by traditional design methods. Because traditional design methods cannot identify the variable ranges of behaviors, nuance and outcome. The principles still apply but the explorations and lean experimentation must be a cross functional effort.

I received this “karma” from a team mate this week. It beautifully encapsulates how I work that I saved it for my portfolio. Her comment about “playing with the medium” here is about me figuring out ways to play with Gaussian Splats, a method to scan environments into 3D models.

I have this principle about “need.” I try not to need anything. I am with someone bc I want to be, not bc I need to be. When I have a job, I try to continually explore other options or have sufficient FU money, so that I don’t have to need the job. I can walk away anytime.

That’s right. The ability to recognize that there is critical thought behind designing an activity for a thing you’re trying to prove evades this person.

I saw this post on LI where a designer recommends not running comparative research bc the result is muddy where users will prefer some things from both versions. This is due to poor design of the proof - he didn’t narrow the thing he wants to test. Unfortunately this how a lot of design teams think.

What you get from experience is the integrative brain. For me, it’s R&D, science, and experience. So people tend to hire me for very specific problems. IME, UX for non-deterministic problems requires an integrative brain.

A lot of design managers wouldn’t even know where to start with emergent tech. That’s bc they haven’t any experience with that level of ambiguity. So they try to apply the design framework into a much more massive R&D problem. It’s like bringing a knife to a gun fight.

Kinda awesome when you work on a proof of concept and someone is like, “Looks like patents need to be filed.”

Designing for probabilistic models.

Last week an engineering manager said, “You have a highly specialized skill.” She was referring to designing & R&D in rich media and 3D experiences. It’s funny bc I’d always considered myself a generalist. It just so happens that I have experience that space, but I can do a lot of other things.

I work across the data pipeline for automation in Rich Media Experiences. The design work here is really thinking about what spatial intelligence is needed to power the experience. I also consider the role of Human in the Loop. What could be tuned to make the experience sing.

I was asked to present to the Mentorship Circle yesterday to share my process and how I work. It forced me to distill my principles and differentiate the innovation process vs the product development process.

Presenting to our Mentorship Circle today on Designing UX Automation for Rich Media Experiences. The good thing about this is that it forces me to solidify my principles.

Sometimes when leaders at the top politicize the situation, they don’t realize the trust they destroyed and that is an invaluable commodity that cannot be bought or coerced. Suddenly their values and words are viewed as performative.

This is a hot take but for early stage innovation that yield new class of experience — if Design isn’t embedded within science teams in the early stage then Design is merely production and execution, not invention. But then I wonder if that work is even Design. It’s definitely UX and R&D.

I’ve long established that how leaders behave, who they reward is indicative of their values. It’s easy to be performative.

I started doing this 10 years ago because I use it for talks. So it just became a habit.

I save everything I work on — every ugly prototype, every presentation & research study. The early ugly prototypes are the best, bc you can look back and see evidence of your curiosity. I even save meta jokes about the product. When I was working on Search, the nerdiest joke was, “Did you mean X?”

When I’m uncertain, I deliberately do the least possible. By reducing variables and resisting the urge to overbuild, I can isolate what’s working and what’s not. This helps prevent premature complexity and keeps the prototype clean enough to learn from.

Two years, one pending patent, 1 pivotal research project that led to an acquisition for Virtual Staging, and the release of SkyTour. More to come.