Overview of Syntax Episode 1027: The Rise of the Design Engineer
This potluck episode of Syntax focuses heavily on the growing importance of the design engineer: someone who can both design and build polished, usable products. Wes and Scott also dig into why AI-generated interfaces often lack “taste,” why testing is more important than ever in the age of coding agents, whether AI is a bubble, how to talk to kids about AI, and how to use coding agents efficiently in large codebases. The episode also includes an update on Scott’s Linux journey, a guest segment with psychologist Dr. Courtney Tolinski on kids and AI, and a pair of useful “sick picks.”
The Rise of the Design Engineer
The central theme is the rise of the design engineer—a hybrid role that blends aesthetic judgment, UX understanding, and technical implementation skills.
Main takeaways
- Wes and Scott see this as a major role for the future.
- The best products increasingly require people who understand:
- visual design
- interaction design
- usability
- implementation details
- performance and debugging
- They argue the old “design handoff to engineering” model is less effective for modern web apps.
- Good design is framed as more than appearance—it’s about how the product behaves and feels in use.
Why this matters now
- AI can generate “okay-looking” UIs quickly, but often fails at:
- interaction quality
- consistent UX
- thoughtful copy
- meaningful product flow
- A strong design engineer can catch the subtle issues that make an app feel polished and trustworthy.
Design Bias, Credibility, and Hybrid Skill Sets
A listener asked whether people discount hybrid design/dev expertise, especially when it comes from a design background.
Wes and Scott’s response
- They agree there can be bias against design-heavy backgrounds in technical circles.
- Scott suggests some of it is simple jealousy: engineers may undervalue design because they themselves lack that skill.
- They also acknowledge the gender bias women often face in the industry.
- Both hosts strongly believe that being excellent at both design and engineering creates a powerful career advantage.
AI Has No Taste: Why So Many Interfaces Feel the Same
Another listener comment led into a broader discussion of AI’s lack of taste or character.
Their argument
- AI can produce visually acceptable work, but it often lacks:
- originality
- character
- subtle usability decisions
- product-specific judgment
- They compare this to:
- early-2000s generic websites
- Bootstrap/Foundation sameness
- modern AI-generated “slop” UIs
Key insight
The problem is not just that AI-generated products look similar—it’s that they often behave similarly in bland or frustrating ways too.
Examples discussed
- Social apps that are hard to distinguish visually
- Threads desktop feeling less polished than Twitter/X in actual usage
- Apple Screen Time’s confusing and frustrating UX
- McMansion-style houses as a metaphor for over-designed, impractical products
Testing Is More Important Than Ever
A listener asked how people are writing tests in the AI era and whether AI should generate tests too.
Core message
Tests matter more now because AI makes it easy to:
- generate more code faster
- introduce more regressions
- push changes without deep review
Their recommendations
- Unit tests for complex logic:
- Scott gave the example of parsing social media URLs and testing every known format.
- End-to-end tests are especially valuable in web apps:
- use tools like Playwright or Cypress
- click through the app as a user would
- make sure the app still works, not just that functions compile
- Visual testing can be helpful in larger systems, though they caution that style changes can make it noisy.
Practical advice
- Make tests deterministic.
- Use AI to assist, but don’t rely on it blindly.
- If possible, have agents interact with the browser in a controlled, test-driven way rather than ad hoc prompting.
Is AI a Bubble?
The hosts discuss whether the AI boom is a bubble and whether it could burst.
Their view
- The financial hype around AI absolutely has bubble-like qualities.
- The massive spending on GPUs, models, and enterprise subscriptions may not be sustainable.
- But AI itself is not going away.
Their take on the future
- The internet didn’t disappear after the dot-com crash; AI won’t either.
- What may burst is:
- unrealistic valuation
- inflated vendor promises
- the idea that AI will replace everything
- The field will likely consolidate around a few major players and/or cheaper open-weight models.
Bottom line
- The bubble may pop in terms of money and hype.
- The underlying tooling and workflow changes are permanent.
Data Centers and Public Backlash
They also address the growing public frustration around AI data centers.
Points raised
- Data centers are seen as a necessary part of AI infrastructure.
- But communities often bear the costs:
- noise
- water usage
- heat
- energy demand
- Wes and Scott are sympathetic to local opposition, especially in places with water scarcity.
Nuance in the discussion
- Scott notes some protestors don’t fully understand what they’re opposing.
- There’s also a broader anti-AI sentiment tied to fear of job loss and AI hype.
- They emphasize that the public should have a say in whether these facilities are built in their communities.
Scott’s Linux Update
A listener asked for an update on Scott’s move toward Linux.
Current status
- Scott is back on his MacBook Pro full-time.
- His Linux machine is older and not ideal for his media-heavy workflow.
- He still thinks Linux could work for him as a developer, but not for:
- photo/video editing
- DaVinci Resolve workflows
- day-to-day media production
What still appeals to him
- Upgradability
- Customization
- Better hardware control
- Framework-style modularity
Linux use today
- He’s still using Linux on a Steam machine.
- He’s planning a Linux-based cyberdeck project, which he’s excited to share later.
Kids, AI, and Parenting: Guest Segment with Dr. Courtney Tolinski
The episode brings in Dr. Courtney Tolinski, a licensed clinical and school psychologist and Scott’s wife, to answer listener questions about teaching kids about AI.
Her credentials
- Licensed clinical and school psychologist
- Over 15 years working with kids, families, and schools
- Co-host of the child development podcast phases.fm
Main advice on kids and AI
- The right approach depends on:
- age
- maturity
- developmental stage
- For older kids and teens:
- teach critical thinking
- explain that AI is not always correct
- encourage fact-checking and skepticism
- For younger kids:
- use AI together with supervision
- avoid giving them unrestricted access
Important themes
- AI should be framed as a tool, not an authority.
- Kids need to understand it can be:
- highly agreeable
- inaccurate
- manipulative in tone
- Boredom, struggle, and working through problems still matter for development.
Concerns about image tools
Courtney also warned about AI image transformation tools for kids:
- They can encourage unrealistic self-comparison.
- They may reinforce insecurity about appearance.
Healthy exposure
- They recommend guided use, not total avoidance.
- Teach children how AI works and where its limits are.
Getting Better at Using AI for Development
A listener asked how someone can become effective at using AI tools for development after being turned away for lacking AI experience.
Their answer
- AI fluency is becoming an expected part of the job.
- Like learning Git, testing, or frontend frameworks, this is now a core skill.
- Developers need to learn:
- how to prompt effectively
- how to structure tasks
- how to control cost
- how to review and guide outputs
Important note on cost
- One hidden skill is understanding token and context costs.
- Continuing huge chats can become expensive fast, especially if the model must reprocess the full conversation.
Keeping Coding Agents Efficient in Big Codebases
The final major topic is how to keep tools like Claude Code, Cursor, and other agents efficient, accurate, and maintainable.
Best practices they recommend
- Keep code files small.
- Use strong linting and static analysis.
- Store important decisions in ADRs rather than bloating
agents.md. - Use hooks and deterministic checks wherever possible.
- Avoid overloading agent instructions with unnecessary detail.
Scott’s practical approach
- Make code more findable with:
- descriptive names
- modular functions
- reduced duplication
- Prefer explicit rules and guardrails over long prose instructions.
- Let agents do less guessing and more controlled execution.
Core philosophy
- If something can be made deterministic, do that instead of leaving it to the AI.
- Smaller, cleaner, more explicit codebases are easier for both humans and agents to work with.
Sick Picks
Wes’s pick: magnet fishing set
- Wes bought a powerful magnet fishing kit to recover a snapped dock chain.
- It worked, and they found the chain underwater.
- He recommends it as a fun outdoors project.
- Caution: the magnet is extremely strong.
Scott’s pick: a kitchen brush
- Scott recommends a $10 kitchen scrub brush with:
- a hanging notch for the sink
- a scraper head
- strong cleaning power for stainless pans
- His point: small usability details make everyday tools a delight.
Final Takeaway
This episode argues that the future of product work belongs to people who combine design taste, engineering skill, and practical judgment. AI can accelerate output, but it cannot replace taste, critical thinking, or careful product design. The hosts’ broader message is clear: learn the tools, but don’t surrender your judgment to them.