The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
Most AI features that wow people in a demo quietly fall apart in week two of production.
The demo works because it ran on three clean examples. Production breaks because:
– no cost controls, so token spend balloons
– no guardrails, so it confidently answers things it shouldn't
– RAG that retrieves the wrong chunks and sounds certain anyway
Shipping AI isn't a prompt. It's evals, fallbacks, retrieval quality, and a token budget — the same discipline as any other system.
17 years of building software taught me that "it works on my machine" applies to AI too.
What's the most painful AI-in-production lesson you've learned the hard way?
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Challenges
View all
Related posts
Custom liquid interaction built in Framer.
Turning a static hero into a responsive visual surface through cursor-driven distortion.
How do you adapt the cursor-driven effect for touch screens?
Ledger — AI CFO Platform
Ledger is an AI platform that acts as an always-on Chief Financial Officer for founders — connecting to accounts, learning cash rhythm, and surfacing risks before they become real problems. I designed the full landing page.
My role: UI/UX Design
Scope: Landing page
Financial tools for founders usually feel either cold (dashboards) or intimidating (accounting software).
For Ledger, we borrowed from old-world institutional finance instead — serif type, oil-painting textures, classical architecture — so an AI CFO feels like real authority, not another app.
This color pallet works so well with data and makes everything look premium. love it!
Built this over a weekend as an experiment with AI-assisted design and development.
Duo Case is an absurd concept for turning two iPhones into a very unofficial foldable iPhone. The goal was to see how far AI could accelerate the process while keeping creative direction, design decisions and final execution in human hands.
A weekend build that looks this polished is wild. Where did AI actually save you the most time here, the concepting or the production grind?