From Pin Codes to Hex Codes
What happens when you (and AI tools) follow weird ideas
What happens when you accidentally discover that Indian pin codes look suspiciously like hex color codes? Last Friday morning, I found out. One hour and with few prompts later, I had painted the entire country using nothing but postal codes as colors. The LinkedIn post about it exploded—4,000 likes, hundreds of comments, and over 400 new connection requests.
The concept was absurd: every Indian pin code is a 6-digit number, which can be a hex color (678006 becomes #678006), creating a map where each postal area gets its own unique shade. Here is the Original Post and Github Repo.
Act 1 - The Process
The first question wasn't for AI - it was the basic: Does map data for Indian pin codes even exist? I fired up Perplexity to search while reaching out to friends from OpenStreetMap Kerala, a community I'd worked with on design projects the year before. They pointed me straight to data.gov.in. Knowing who to ask beats knowing what to ask.
Then I opened Claude - but not for code. I positioned it as my technology brainstorming partner and laid out the concept: Indian pin codes as hex colours, mapping possibilities, visual potential. The conversation felt like whiteboarding with a colleague.
Claude, I had this random thought today and wanted to bounce it off you as a tech expert. You know how Indian pin codes are all 6-digit numbers? I just realized they could work as hex color codes too - like 678006 could become #678006. I'm thinking if we can create some kind of visual map where each postal area gets colored by its own pin code.
I know there's map data out there somewhere for Indian postal boundaries, and I'm pretty comfortable with web stuff, but I'm curious about your thoughts on this. What would be some interesting ways to approach this visualisation? Any mapping libraries or tools that would make sense? Also, what kinds of problems do you think I might run into with something like this? I'd love to hear you think through this with me. Don’t write any code.
Claude threw out options - D3.js, Leaflet, various libraries. Since I had already decided on Leaflet and had peeked inside the GeoJSON structure, I cut straight to the directive. Instead of asking for code, I asked Claude to break everything down into a task list - every feature, every edge case, every possible complication - all numbered and explicit.
Great. I've decided to go with Leaflet for this. Could you help me think through this systematically? Think like a product and tech person, break this entire concept down into a simple, numbered task list in markdown.
I want every feature I should consider, every edge case that might bite me, and every potential features and complication laid out clearly. Make it comprehensive so I can make smart decisions about what to actually build versus what to skip. Don’t code. Just markdown numbered list.
Back came around 15 tasks covering file parsing, colour mapping, interactive features, legends, hover effects, zoom controls. On the next prompt, I asked to focus on specific tasks 1, 2, 5, and 7, and to skip everything else.
From that task list, I want to focus only on tasks 1, 2, 5, and 7. Skip everything else for now. Use plain HTML, No react or tailwind. Build a basic page with Leaflet embed.
The code worked. I tweaked zoom levels and added some headers, and suddenly I could see 19,000+ coloured regions across India. But there was a problem - urban areas vanished into white pixels, too small to render at the country scale. Since I was planning for a full-view print rather than web interactivity, I described the rendering issue to Claude and asked for a way to generate an A3 PDF. Claude recommended QGIS.
Hello, I've got my pin code color map working in Leaflet, but there's a rendering issue - all the urban areas with tiny pin code boundaries just disappear into white pixels when I zoom out to see the full country view.
I'm planning to print this as a high-quality A3 poster rather than keep it as an interactive web map, I'm open to switching tools or using other websites/software entirely. Give me multiple suggestions for how to tackle this, and let me know which approach you think would give me the best results for a clean, printable map - as step by step documentation. I have no expertise in GIS.
For the QGIS workflow, I switched to ChatGPT - not because Claude couldn't handle it, but because ChatGPT excels at structured step-by-step processes. I fed it the full context in one detailed prompt, along with the code Claude generated.
The Ripples
What happened next surprised me - My network grew by 500+ connections, I had more Topmate requests than ever. Within days, people shared it on Reddit and someone had created a German map using the same hex-colour logic. Curious Python developer Harshee Pitroda, built a #dd-mm-yy calendar. Arun Ganesh build the same with HSL colors and introduced me to Amche. People were taking the idea and running with it in directions I never imagined.
What I Learned
The biggest mistake I kept making with AI was asking for solutions when I should be asking for problems first. These days I spend more time having AI list what could go wrong than actually building anything, and that made all the difference."
The Task List Approach
Instead of diving straight into build me this, I've started asking AI to dissect concepts into numbered task lists first. For this, Claude gave me around 15 tasks covering everything from file parsing to hover effects to zoom controls. Then I could say "focus on tasks 1, 2, 5, and 7" instead of explaining later why I didn't need the other features.
This felt like having a conversation with someone who thinks through problems the same way I do, just faster.
Making Decisions, Not Documents
Clear communication beats sophisticated vocabulary every time. The real breakthrough came from aggressively shrinking scope instead of getting lost in 16-page documents and endless possibilities. When AI gives you every possibility, your superpower is deciding what deserves to exist. It's infinitely easier to tell AI what to skip upfront than to explain why you don't need something after it's already built. Turn AI into your strategic thinking partner, not just a code monkey.
None of this was about the technical execution or clever prompts. It was about following curiosity down an unexpected path and sharing the messy, imperfect result. Sometimes the things we almost don't post are exactly what spark something bigger. If you've got a weird little experiment gathering dust, this might be your sign to put it out there!
This newsletter accidentally became a novella. Regular-sized This&That will resume once I remember how to write short sentences.
Wow..!
This is incredibly nice..!
How refreshing! And what a great reminder to get back to the basics - curiosity!