Webflow's Builder Day

1. Learning by Building Works

• Builders valued having dedicated time to explore without delivery pressure — it encouraged risk-taking and creativity.
• Many reflected on the joy of hands-on learning, often describing it as "fun," "energizing," and "a reminder of why I love to build."
• Confidence grew as prototypes and automations started to work; participants described wanting more frequent Builder Days or mini build sessions.

Sample reflections: "I could do this all week." • "Having space to play around helped ideas form organically." • "The protected time made all the difference."

2. Working With AI Tools — Strengths & Frictions

Cursor

• Excellent for rapid prototyping, syntax help, and scaffolding ideas quickly.
• Less effective for deep architecture, debugging, or UI polish.
• Works best when builders give it structured plans, small prompts, and guardrails ("Plan mode" cited frequently).
• Builders learned to treat it as a thinking partner, not an autopilot.

Figma Make

• Great for low-fidelity interaction prototypes and visual exploration.
• Some found it "too eager" — jumping to implementation before confirming details.
• Especially useful when paired with Cursor (using Figma for visuals and Cursor for logic).

Marvin, Hex, and Insights Tools

• Marvin and Hex sped up data analysis, research synthesis, and SQL generation.
• Insights participants highlighted the need for better data structure and export formats.
• AI-assisted data work revealed friction around trust and transparency: understanding why the AI produced a given result.

3. Prompts, Process, and Planning

A universal learning theme: "Think first, prompt second."

• Builders who spent time planning prompts or writing pseudo-PRDs achieved faster, more accurate results.
• Iterative, incremental prompting consistently outperformed single large requests.
• Combining multiple tools (e.g., Figma → Cursor → ChatGPT for JSON) created powerful workflows.
• Participants emphasized balancing fidelity vs. learning — getting rough ideas working before refining visuals.

4. Craft, Fidelity, and Focus

• Many learned that perfectionism slows learning — low fidelity leads to faster insights.
• Builders reframed success as "getting something working" rather than matching pixel-perfect mocks.
• A recurring theme: "Focus on flow and storytelling, not polish."
• Participants cited the value of quick user feedback: even 15–20 minute interviews clarified priorities and scope.

5. Collaboration & Support

• Builders frequently co-debugged or shared solutions with others — spontaneous cross-functional pairing emerged.
• Mentions of #builder-day threads, Open Jam Room, and vendor office hours show high engagement with support channels.
• Many thanked team members for help debugging or prompting guidance.

Recommendation: Keep emphasizing real-time pairing and shared learning in future events.

6. Insights Track Highlights

Builders using Marvin and Hex showed that AI can accelerate research synthesis and data exploration, but data prep and structure remain critical.

Insights participants noted:

• Need for centralized, searchable data sources (Glean, Drive, Confluence).
• Importance of human validation: "AI can compile, but we still need to read and fact-check."
• Excitement about agentic AI tools like Braintrust and Hex Agent for research automation.

"AI can speed up synthesis, but it still needs human judgment to interpret and trust the results."

7. Cultural Impact & Takeaways

• Hands-on time = confidence. Repetition across submissions showed that direct experimentation built lasting comfort with AI tools.
• Cross-functional learning. Designers learned code structure; PMs prototyped; analysts automated.
• Shift in mindset. Participants recognized that AI fluency isn't about output speed — it's about understanding how to think with the tools.
• Desire for more. Multiple reflections suggested doing Builder Day quarterly or extending it to multi-day hackathons.

💪 Strengths

• "Cursor reduced the friction of opening a new codebase."
• "It's so great at getting ideas out quickly."
• "Using AI (via Cursor) to create AI-powered products … is incredibly fun and powerful."
• "Really easy to spin up interactions in Cursor and create callouts to OpenAI."
• "I found it super helpful … as I don't work with Zustand and node.js regularly I would have had to google the syntax."
• "Cursor rules take time to set up and fine tune, but very fun to use once they're ready."

⚠️ Pain Points & Limitations

• "Ultimately my learning was that I'd use Cursor for prototypes or demos and assistance with syntax, but not for the structural, architectural, or logic bits of dev."
• "It's not a shortcut to avoid thinking. Either you think first or you think later."
• "Very hard to dependency manage or problem solve in Cloud and have AI fix things."
• "Doing any fine tuning on the UI can be difficult and quickly doesn't feel worth the time investment."
• "Cursor had weird sample data … ask for the JSON format it needs then generate sample data separately."

💡 Tips from Builders

• "Small, more frequent prompts yield the best results."
• "Spending the time up front to work through a plan before going into agent mode was helpful."
• "Plan mode was helpful to brainstorm different ideas."
• "Used Plan mode to break complex features up as a PRD with phases."
• "For net-new UIs, Cursor did better once I started wireframing and mapped out UI flows very precisely in text or flows."
• "Include screenshots of designs to kick things off. Cursor works better when it can see context."

💪 Strengths

• "I think Figma Make is so good for low fi prototypes. It's great at visual design and interactions."
• "Really helpful in quickly prototyping a new experience."
• "It helped fill in logic gaps without needing me to add in every step in between."
• "Had a great experience trying out Figma Make. It added an 'apply animation' button when I told it the action I wanted it to do."
• "Used screenshots of the existing live Webflow experience to prompt new experiences."

⚠️ Pain Points & Limitations

• "It was often too eager, just implements code without confirming if that's what I want."
• "Sometimes it was harder than expected to get the prototype to look like the Figma mocks."
• "Hard to not focus too much on small design details … forced myself to move on when it wasn't working."

💡 Tips from Builders

• "Start with screenshots and rough sketches. High fidelity isn't necessary to get good results."
• "Figma Make and Cursor work well together. Use Figma for visuals and Cursor for logic."
• "Small, frequent prompts and iterating in Make led to better outputs than one big ask."

💪 Strengths

• "Hex agent is amazing. We barely fed it 2 sentences, and it produced exactly what we needed, including the SQL that we can now refine and reuse."
• "It probably would have taken me a week without Hex's AI assistant."

⚠️ Pain Points & Limitations

• "Most of my time was spent debugging, but especially for math heavy functions AI was so much faster."
• "Configuration of the cloud agents' virtual machines is challenging … difficult to troubleshoot."

💪 Strengths

• "This POC showed how much potential AI has to speed up research synthesis and amplify how we communicate insights."

⚠️ Pain Points & Limitations

• "My biggest obstacle was structuring Marvin's data in a way that AI tools could easily interpret and trust."
• "I'm not sure yet whether that friction came from my own workflow choices or from limitations in Marvin's current export capabilities."

Enablement & Setup

• "Better documentation for how AI tools connect with Cloud would make setup smoother."
• "Keep early tool access and office hours, those made it easier to get unblocked."

Prompting & Process

• "Short refreshers on how to plan and write good prompts would go a long way."
• "Start small and build step by step, Plan mode and simple checklists really help."

Collaboration & Support

• "Pairing up or asking questions in Slack threads made a huge difference, keep encouraging that kind of real-time help."

Building Mindset

• "Focus on functionality and flow, not polish. The best insights came from rough builds."
• "Having protected time to play and learn is what made this work, keep that spirit."

After the Event

• "Would love more chances to share what we built, maybe a gallery or mini-showcase."