the "build wars" experiment
The setup was dead simple:
45 minutes to create a landing page for an AI robotics company called Pulse (view the design brief).
Brett (10+ years design experience, $145k/month design agency) versus Henrik (on Lovable's marketing team, not a designer).
Both had access to the same assets and copy. Both had to include at least two interactive elements. Pre-built UI components were okay. No full templates allowed.
Henrik w/ Lovable started strong with quick component assembly and an early animated robot.
Brett w/ Webflow began methodically with fundamentals.
By the end, Brett's experience clearly showed in cohesion and quality, but Henrik's AI-powered output was surprisingly competent for a non-designer.
But that's just context. Here's what matters:
7 brutal lessons (and how to apply them now)
There was a lot of discussions happening in real-time (over 4,000 comments an counting):
1. The "Zero-to-Something" Gap is Closing Fast
Henrik went from blank canvas to functional hero section in minutes. This matters enormously. How many projects die in the "getting started" phase? How many good ideas never see the light because the activation energy is too high?
AI tools are demolishing that initial barrier. They're not making everything perfect, but they're making "started" infinitely more accessible than ever before.
What you can do today:
- Try Lovable, Bolt, or Replit for rapid prototyping before diving into detailed design
- Use AI tools to generate 3-5 different layout approaches in minutes before committing
- Start projects by asking "what's the fastest way to get a working prototype?" not "what's the perfect solution?"
2. Experience Can't Be Prompt-Engineered (yet)
Brett's process looked like controlled chaos to the untrained eye, but designers recognized it immediately: the methodical layering of fundamentals before details. He knew exactly which elements deserved focus in a time crunch and which could be handled later.
This judgment comes from thousands of hours solving similar problems. No prompt can replace it.
What you can do today:
- Document your own design decisions and rationale to build a personal "experience database"
- Build a swipe file of effective designs and analyze why they work
- When using AI tools, spend more time refining the outputs rather than generating more options
- Focus on learning design fundamentals that transcend tools: hierarchy, rhythm, contrast, and alignment
3. Design Systems Might Become Non-Negotiable
Neither competitor designed from scratch.
Brett leveraged Webflow's component system; Henrik used 21st.dev components plus AI generation. The era of blank-canvas design is effectively dead for production work.
And while this was certainly a decision made to compete with the comical time-limit, it's also a glimpse into how to make these tools run efficiently.
The future might be about intelligently assembling, customizing, and extending components - whether through visual tools or AI prompts.
What you can do today:
- Set up a simple design system before your next project (even if it's just typography, colors, and spacing)
- Use 21st.dev, Grit UI, or Tailwind CSS as starting points for component libraries
- Try this hybrid approach: use AI to generate components, then organize them into a consistent system
- For client work, start with system definition before touching visuals
- For larger teams with existing products, tools like Onlook, Builder and TempoLabs are working on integrating with your code bases
4. Time Remains the Ultimate Constraint
The most honest moment came after the competition when both admitted 45 minutes was nowhere near enough time.
Brett said "45 minutes is not easy" and Henrik admitted "it went incredibly horrible for me."
In fact, Dan Mall wrote about his process for competing against me in the Relume Design League a few years ago (yes - it's hard).
For all the advances in tools, quality design still requires time for refinement, iteration, and polish. Neither AI nor traditional tools have solved this fundamental reality.
It's why I created a course about making fast, informed UI decisions when time is limited.
What you can do today:
- Budget realistic timeframes regardless of tools - 2-3x what you think AI will save you
- Split projects into clear phases: rapid generation, systematic refinement, and final polish
- Use timeboxing: 1 hour for AI exploration, then commit to a direction and refine
- Soren Iverson does an improve timebox of bad ideas with his teams
5. AI Tools Lower the Floor, Not the Ceiling
Henrik, a 19 year-old with zero design experience, created a functional landing page in 45 minutes.
Five years ago, that would have been impossible. The barrier to entry has dramatically lowered.
But Brett's output was clearly superior in cohesion, hierarchy, and storytelling. AI has made "decent" more accessible but hasn't yet cracked "exceptional."
What you can do today:
- If you're not a designer: use AI to get to "decent" faster, then iterate based on feedback (watch Aakash Gupta prototype 5 features in 80 minutes)
- If you are a designer: focus on the skills that elevate work from "decent" to "exceptional"
- Create a checklist of quality factors AI tends to miss: consistent spacing, typography rhythm, intentional emphasis
- Use AI outputs as first drafts, not final products
6. The Future is Hybrid, Not Either/Or
The most successful approach is likely a combination: AI for rapid scaffolding and ideation, human refinement for quality and nuance. The tools reflect this already - Lovable adding visual editing, visual tools adding AI assistance.
This isn't about AI versus traditional tools; it's about finding the right blend for each specific challenge.
What you can do today:
- Try this workflow: use v0/Lovable for initial structure, export to Figma/Webflow for refinement
- Experiment with AI for parts of your process where you typically get stuck
- Use Cursor instead of VS Code to add AI assistance to your development workflow
- Try Windsurf for hybrid design/code workflows where AI can suggest implementation
7. Design Quality is Becoming More Important
This is a luke-warm take, but it was certainly reinforced.
If AI can help non-designers produce decent work, the bar for "good enough" rises for everyone (a good thing).
Paradoxically, as basic design execution becomes more accessible, the value of exceptional design thinking increases.
The gap between "acceptable" and "outstanding" remains, but the baseline is rising rapidly.
What you can do today:
- Audit a recent project: what elements truly differentiate it from what AI could produce?
- Focus learning efforts on strategic skills: information architecture, user journey mapping, visual storytelling
- Create a personal quality checklist to review all work before finalizing
- Look for opportunities to elevate AI-generated work with your unique perspective