Tintello: Judgment Is Harder to Delegate Than Code | by Paul Gedeon | May, 2026 | MediumSitemapOpen in appSign up<br>Sign in

Medium Logo

Get app<br>Write

Search

Sign up<br>Sign in

Tintello: Judgment Is Harder to Delegate Than Code

Paul Gedeon

11 min read·<br>Just now

Listen

Share

A year ago, I shipped Piplo in 50 hours. I spent most of them in an IDE. This time I never opened a source file.<br>Piplo is an iOS relationship-reminder app, in about 50 hours using Cursor and ChatGPT. That project proved to me that AI could compress the timeline enough for a solo build. In the past months, I learned from scaling my own responsibilities with Claude Code that the models had taken a step-function leap in Q1 this year. I wanted to build a new app to test how much had changed in a year. I just needed an idea.<br>Then, while reading a book, I came across “zomp,” a color I had never heard of before. I looked it up, then fell into a rabbit hole of bizarre color names (e.g., Caput Mortuum, Isabelline, Smaragdine, Feldgrau). A quiz app to learn more about both common and quirky colors felt like a good candidate.<br>During the Memorial Day long weekend, I built Tintello, a calm color-learning app with 700+ named colors, 200+ themed palettes (i.e., colors grouped together in an evoking scene such as Moroccan Riad Pool), 2,000+ sourced lore facts, and a daily practice loop. I used Claude Code, OpenAI Codex, and ChatGPT. The codebase is larger, the content is deeper, and the product is more polished than Piplo.<br>Press enter or click to view image in full size

The Tintello logoHere is the sharpest version of the difference: with Piplo, I spent most of the 50 hours inside Cursor catching bugs the AI introduced while fixing other bugs, reinforcing coding standards, editing the business logic in the code when needed, and manually testing the app in the simulator. I was in the code. With Tintello, I never read a single line of Swift. Not once across the entire build. I operated entirely through product feedback: play the app, react to what I saw, direct the next change. The code was not my problem. The product was.<br>That freed every hour I would have spent debugging to go toward UX iteration and content quality instead, which turned out to be exactly where this app needed the most attention.

The First Night<br>At 6:45 PM, I spent 35 minutes on a voice conversation with ChatGPT to produce the BRD and logo direction, similar to what I did for Piplo. Then I handed the BRD to Claude Code and told it to scaffold the app. I stepped out to do laundry while it was working.<br>By 7:30 PM, Claude Code hit its session limit, which forced an early tool switch. I moved to Codex, gave it the right guidance to pick up with Claude Code left off and started clearing my personal inbox while it was working. By 8:10 PM, the app was running on the local simulator, functional and playable. The same evening, with laundry and email in between.<br>With Piplo, a fast session still meant sitting in the IDE, addressing build-failing bugs, approving changes, testing, and staying close to the implementation. This first night of building Tintello looked nothing like that. By the time I sat down to play the app, I already had a backlog of UX feedback. Not a single item was about code.

The Product Pivot<br>Tintello started as a color quiz. Show a color name, present four tiles, and ask the user to choose the correct one. I was planning a game similar to Wordle, where everyone worldwide would get a single daily palette, score points based on correctness and speed. The main hook was competitive with a global leaderboard and also a discussion-starter.<br>After the initial build, I started focusing on minor UX improvements (e.g., copywriting, answer handling, no screen-wide feedback effects). The color tiles were the hero and nothing should upstage them. I had something functional to show others at that point. The bigger pivot came from a friend. I was at dinner on Saturday and she tried the app on my phone. Based on her feedback, it was clear that Tintello should feel like a calm, daily ritual. Something relaxing and educational, not competitive. The idea clicked immediately. The palettes should paint sceneries, evoke the imagination, and help the user mentally travel through the colorful scenes.<br>While I hit the restroom midway through the dinner, I typed a quick prompt into Codex on my phone through the remote control functionality, describing the new direction. When I got home, there was already a bare-bones version of the ritual to experiment with. That guidance restructured everything, removing the leaderboard and ranking emphasis. It redesigned the Explore tab to show four evocative palette choices instead of a firehose.<br>Piplo’s core job was stable from the start: help me not miss reaching out to people. Tintello’s core job had to be discovered through interaction. Codex implemented fast enough that I could feel the product, reject directions, and pivot the identity repeatedly. The tool...