Lane Garner | Design-Focused Software Engineer

Hello.

I'm Lane Garner,a design-focused frontendsoftware engineer basedin Austin, Texas.

My specialty is the front of the frontend.See my work

I build

fast

interfacesandsystems

What I do

Usability

Don Norman is my north star: signifiers that invite the right action, feedback that answers it, mappings and constraints that make the wrong action hard. If users need the manual, the conceptual model failed.

Accessible craft

Semantic HTML, keyboard-first interactions, WCAG 2.2 AA. Controls should look like what they do and answer every action with feedback. If the focus order is wrong, the work isn't done. Accessibility issues are usually usability issues in disguise. Fixing them makes the product better for everyone.

Product engineering

I build and ship whole products, backend included. GarnerGuitar.com has paying students; Rhythm Fit works in a basement gym with no signal.

Design systems

Design tokens and components that survive contact with more than one platform. This site runs on tutti-ui, my React + React Native system.

AI-native interfaces

Streaming output, tool calls rendered as real UI, honest fallbacks. Rhythm Fit's coach streams NDJSON into scheduled workouts; Tribe Tracker drops to deterministic stats when the model fails.

AI is an interface problem

A canned exchange, no model behind it. The rendering is real: tutti-ui's AgentWorkflow and StreamingText, the same components documented in the Storybook.

Build me a three-day dumbbell plan for a basement gym with no signal.

○

Parse the request

goal, days per week, equipment

○

Call generate_plan

arguments validated against a JSON schema

○

Stream the reply

NDJSON chunks rendered as they arrive

Canned data, zero API calls.

AI without the chat box

AI doesn't have to be a conversation. The interesting work is baking the model into the product itself: onboarding that turns a short survey into a ready-to-go plan, coaching that reads your actual data and speaks up on its own, digests that land where you already are.

No blank text box, no prompt engineering left to the user. The model works behind interfaces that look like the product, and when it's unavailable, deterministic fallbacks keep the feature honest.

A survey becomes a plan

GoalEquipment

Three weeks of workouts land on the calendar. No chat involved.

Coaching that speaks up

Bench press has stalled

Three weeks without progress. Try a 5% deload on Friday.

Confidence86%

The model reads your training data and raises its hand first.

Digests where you already are

Weekly recap: 4 of 5 kept their streak. Sarah leads by 12 points, closable in 3 days.

Posted by a cron job, straight into the group chat.

A superpower, not a replacement

I code with Claude daily and have shipped with Copilot and Cursor too. The speedup is real, but the output is only as good as the human in the loop. Without someone who knows what good looks like, you get slop: code that runs today and can't be maintained tomorrow.

UI is where that shows most. AI has historically struggled with pixel-perfect interfaces, and while newer tooling keeps getting better, a discerning eye and a human leading the way still make the difference. Used that way, AI isn't a replacement for the craft. It's an enhancement, and a superpower.

Side Projects

Passion projects, built nights and weekends because I wanted them to exist. Some solve my own problems, like a workout app whose AI coach comes up with the training plan and tracks my progress over time, so I can just do the work. Others are things I wanted to own end to end, like a design system that ships to web and native from one set of tokens. Each card links to a write-up with the decisions behind the build.