iiinova — where your digital self keeps evolving

E Lai @elaiuiux.bsky.social · 17/08/2026

So the recommendation is restraint. Cut controls until the handoff of judgment is clear. If a person can trust it, change it, and reverse it, you've done the job. Before you add another toggle, ask what decision it actually serves.

000 0

E Lai @elaiuiux.bsky.social · 17/08/2026

The convincing wrong answer is the one that costs you. A confident interface on top of an uncertain model is a trap you set for your own users. I'd rather ship something that admits when it's guessing.

100 0

E Lai @elaiuiux.bsky.social · 17/08/2026

Three things earn trust: 1. Show confidence where the decision happens, on the output itself. 2. Make every result editable in place, no modal, no detour. 3. Make undo boring and instant. Everything else stays hidden until the moment needs it.

100 0

E Lai @elaiuiux.bsky.social · 17/08/2026

Here's how I think about it instead. The real job is deciding how much of the model's uncertainty a person can absorb at the exact moment they act. That's it. That's the whole design problem.

100 0

E Lai @elaiuiux.bsky.social · 17/08/2026

The instinct is to fix this with controls. Sliders, toggles, a temperature setting nobody understands. Now you've built a settings panel with a chat box attached.

100 0

E Lai @elaiuiux.bsky.social · 17/08/2026

The output shows up looking finished. Same typography as everything the user typed themselves. A guess dressed up like a fact.

100 0

E Lai @elaiuiux.bsky.social · 17/08/2026

Every AI feature I review lately has the same flaw. The model works fine. The interface has no opinion about how sure it is.

2026-07/3e62cbac59b90d582478299361d9124b.png

110 0

E Lai @elaiuiux.bsky.social · 14/08/2026

A small frame is respect for the client's money. We'd rather commit to less and keep it than capture everything and drift. Before your next kickoff, try naming what the project will not solve. Out loud.

000 0

E Lai @elaiuiux.bsky.social · 14/08/2026

Everything past those six things is scope creep dressed up as discovery. More questions feel thorough. Mostly they're avoidance of the harder call about what to leave out.

100 0

E Lai @elaiuiux.bsky.social · 14/08/2026

1 non-negotiable: the thing that kills the project if it slips. One. Not five. Making someone pick one forces the honest conversation about what actually matters. Also who owns the decision. We ask that on this call too, before any energy gets spent.

100 0

E Lai @elaiuiux.bsky.social · 14/08/2026

2 constraints: budget and the thing that can't move. Usually a launch date or a technical reality. Constraints early make every decision after this faster. Untested ideas always look stronger than they are.

100 0

E Lai @elaiuiux.bsky.social · 14/08/2026

3 outcomes: what has to be true when we're done. Business results, stated plainly. If the client can't name three, that's the real work of the call. Not the timeline. Not the deliverables list.

100 0

E Lai @elaiuiux.bsky.social · 14/08/2026

So we cut it down to a frame we can hold in one call. 3 outcomes. 2 constraints. 1 non-negotiable. That's it.

100 0

E Lai @elaiuiux.bsky.social · 14/08/2026

We used to run discovery over two weeks. Workshops, docs, a deck. One project still drifted anyway because nobody had agreed what it was for. The long process just delayed that argument.

2026-07/2cd2562a32a15bedf8cca1145ecb1e5e.png

100 0

E Lai @elaiuiux.bsky.social · 10/08/2026

The thing I keep coming back to. Respecting an expert audience means showing more of the real thinking, not less. Arranged so it lands fast. Sanding it down is what actually loses them.

000 0

E Lai @elaiuiux.bsky.social · 10/08/2026

Order mattered more than anything. Decision first, what leads and what gets left out. Evidence second. Expression last. Most teams start at expression and wonder why nothing lands.

100 0

E Lai @elaiuiux.bsky.social · 10/08/2026

Product visuals followed the same rule. No shrunken dashboard with a glow behind it. That proves software exists, nothing else. We picked the moment of value and cropped to the part that proves it.

100 0

E Lai @elaiuiux.bsky.social · 10/08/2026

The structure was simple. For every claim on the page, we asked what skeptical question it creates. Then we put the answer right there. A product state, a step in the process, an honest constraint. If we had no answer, the claim came off the page.

100 0

E Lai @elaiuiux.bsky.social · 10/08/2026

So we went the other way. Keep the hard parts visible. Structure them so a skeptic can watch the mechanism work instead of taking our word for it.

100 0

E Lai @elaiuiux.bsky.social · 10/08/2026

The instinct on most SaaS sites is to simplify until nobody feels lost. With this audience that backfires. An engineering leader reads a vague promise and assumes you're hiding something.

100 0

E Lai @elaiuiux.bsky.social · 10/08/2026

Karat's audience is engineering leaders. People who can smell an oversimplified pitch in one scroll. The brief was basically: sell technical recruiting to them without dumbing it down. Harder than it sounds.

2026-07/9fc6a472a9439e1f94c629badd6a8f2e.png

100 0

E Lai @elaiuiux.bsky.social · 07/08/2026

Before you pick a typeface, write down what it should make someone feel in the first second. Then choose against that. The mood board can wait.

000 0

E Lai @elaiuiux.bsky.social · 07/08/2026

A fintech client wanted a display face they'd seen on a fashion brand. Beautiful type. Wrong promise. Their users needed to feel safe before they felt impressed.

100 0

E Lai @elaiuiux.bsky.social · 07/08/2026

What I do now: 1. Name the outcome first. Should this feel fast, safe, expensive? Pick one, maybe two. 2. Choose against that outcome, not against what's trending. 3. Say out loud why it holds. If the explanation is thin, so was the choice.

100 0

E Lai @elaiuiux.bsky.social · 07/08/2026

That bothered me more than I expected. If I couldn't explain the choice, it wasn't a decision. It was a reflex dressed up as expertise.

100 0

E Lai @elaiuiux.bsky.social · 07/08/2026

I sat in reviews for years as the person who said "that one feels right." It worked until a designer asked me why. I didn't have an answer that wasn't taste.

100 0

E Lai @elaiuiux.bsky.social · 07/08/2026

So the typeface is a claim. A silent one, but a claim. Pick it off a mood board and you're shipping a claim you never examined.

100 0

E Lai @elaiuiux.bsky.social · 07/08/2026

People decide how fast, safe, or expensive your product feels in about a second. Type does a lot of that work. Before the headline even registers.

2026-07/a1db75057356daca901f1d6352e84d4d.png

100 0

E Lai @elaiuiux.bsky.social · 05/08/2026

So the first deliverable is never a design. It's the decision the design has to express. Put the answers side by side before you reconcile anything. Agreement you can hear isn't agreement you can build on.

000 0

E Lai @elaiuiux.bsky.social · 05/08/2026

Consensus isn't the goal either. Consensus produces the safest, least useful version of the truth. The goal is clarity about who decides, what evidence backs it, and which tradeoffs the company is accepting on purpose.

100 0

E Lai @elaiuiux.bsky.social · 05/08/2026

And I don't treat those gaps as a problem to smooth over. They're data. If sales and product disagree on why customers buy, the homepage was never going to work anyway. We'd just be decorating the disagreement.

100 0

E Lai @elaiuiux.bsky.social · 05/08/2026

That's why I collect the answers separately. In a room, answers bend toward the loudest person. Alone, people write down what they actually believe. The gaps between the versions tell me more than any polished brief.

100 0

E Lai @elaiuiux.bsky.social · 05/08/2026

The words match. "Platform," "workflow," "enterprise-ready." The beliefs underneath don't. You'd never catch it in a kickoff meeting, because everyone nods at the same terms.

100 0

E Lai @elaiuiux.bsky.social · 05/08/2026

Four smart people, same company, same vocabulary. Four different businesses on the page. The founder describes the vision. Sales describes whatever closed the last deal. Product wants every feature in the story.

100 0

E Lai @elaiuiux.bsky.social · 05/08/2026

The questions are simple. Who is the primary customer? What forces them to look for a solution now? What outcome are they buying? Why do they choose us? What makes them hesitate?

2026-07/1c0e05fede07f31f812930c623d664a3.jpg

100 0

E Lai @elaiuiux.bsky.social · 03/08/2026

Design the first four clicks like the round depends on them. It does.

000 0

E Lai @elaiuiux.bsky.social · 03/08/2026

Empty state on screen three? Fix that before the deck. Before the motion polish too, which is where his budget had actually gone.

100 0

E Lai @elaiuiux.bsky.social · 03/08/2026

Fill those four screens with a believable customer's mess instead. Awkward numbers. A long name that wraps weird. One edge case sitting in the table. It reads as survived, not staged.

100 0

E Lai @elaiuiux.bsky.social · 03/08/2026

An investor reads fake data as a product that hasn't touched a real customer yet. They won't say that in the room. They just get quieter.

100 0

E Lai @elaiuiux.bsky.social · 03/08/2026

And the thing that kills them is rarely ugliness. It's placeholder data. John Doe. $1,234. Lorem Corp. The screen looks finished and feels fake at the same time.

100 0

E Lai @elaiuiux.bsky.social · 03/08/2026

So here's the rule I work by. Those four screens get more design attention than the other forty combined. Uneven on purpose.

100 0

E Lai @elaiuiux.bsky.social · 03/08/2026

That's about how far an investor gets before the verdict forms. The rest of your walkthrough is narration on top of a decision that already happened.

2026-07/c192cd460a91788ff8bc4f7e8fdefbe0.png

100 0

E Lai @elaiuiux.bsky.social · 31/07/2026

Feature grid redesign for Karat 💜 Focused on making technical interview prep feel more approachable, structured, and community-driven for engineers at every stage. #WebDesign

010 0

E Lai @elaiuiux.bsky.social · 31/07/2026

The strongest companies usually move calmly because operational refinement has compounded for years ⚡️ #Operations

000 0

E Lai @elaiuiux.bsky.social · 30/07/2026

Mobile UI for Bikeroute 🗺️ Built to make every ride feel connected, focused, and easy to follow. #UIDesign

010 0

E Lai @elaiuiux.bsky.social · 30/07/2026

The more designers understand business, the stronger their work usually becomes ⚡️ #ProductDesign

000 0

E Lai @elaiuiux.bsky.social · 29/07/2026

Travel booking card UI inspired by Santorini sunsets 🌅 #TravelUI

000 0

E Lai @elaiuiux.bsky.social · 29/07/2026

The best client relationships eventually stop feeling transactional. You start thinking together. Planning together. Solving problems together. That’s usually when the strongest work happens. Trust changes the quality of collaboration dramatically 🚀 #ClientWork

000 0

E Lai @elaiuiux.bsky.social · 28/07/2026

Design systems are really communication systems. Yes, they improve consistency. But the biggest value is alignment. They help: • designers move faster • developers build consistently • product teams make cleaner decisions Strong systems reduce ambiguity across the company ⚡️ #DesignSystems

020 0