Claude

Spex Glance: I Wanted a Better Kalshi Window, So I Made One

I wanted a Kalshi window that looks like Apple Stocks. No coding background. Claude ran the builds, I made the calls, and I never opened Xcode. What I learned shipping Spex Glance from idea to v0.2.2.

David Marcantonio

David Marcantonio

— 3 min read
Share
Spex Glance: I Wanted a Better Kalshi Window, So I Made One
One window, every open position, no browser tab.

tl;dr - Casual Kalshi user, bland UI. Wanted something like Apple Stocks. No coding background. Claude drove the build. Shipped v0.1.0 in a day, v0.2.2 now.

I was looking to solve a problem I had with Kalshi. I'm a casual prediction market person, but the Kalshi UI on your phone or the web is bland. I wanted a compressed UI with just the data. Detailed but simple. Glance at every open order on my sporting matches and move on. Something like the Apple Stocks app, which gives me what I need in a clean, compressed view. I had no coding experience, but I figured I could try using Claude to close that gap. Partly for what I wanted out of Kalshi, partly for my own learning.

Some background. If I want to do something, I'll read the manual front to back, more than once. Trial and error works for me. But coding an app from scratch is something I've tried to tackle for decades. Back in the early days of the internet, learning HTML was easy enough, but the standards were moving fast and it was a lot to keep up with. Where I worked at the time we had Adobe PageMill, an early WYSIWYG HTML editor. You dropped elements onto the page however you wanted it to look. That took the coding-from-scratch part out of the equation and got some early web pages built for my job. Ever since, I've looked for ways to skip the learning curve, lean on my design instincts, and build with off-the-shelf tools, paid or free.

So where I landed next took me from an idea to a shipped release. After a few conversations with Claude, it dawned on me that Claude could drive Xcode's build tools on my Mac directly. I connected it, and what happened next made my jaw drop. Claude edited the code, ran the builds, and took screenshots of the running app so I could react. I never opened Xcode myself. With a lot of questions, requests, and ideas from me, I wound up with a polished app that would've taken me years of dedicated learning to reach.

The idea from the start was a Mac app that pulls live data from Kalshi and shows it in one window. Not all the data, just what's relevant to me. Claude went out, read Kalshi's API, scoped it down to the pieces I cared about, and gave me a list of options for how to display them. Once it had everything it needed for the project, I set it to work.

I've had an eye for design and layout over decades of using open source, paid software, and the web. I've used apps that were polished and apps that were horrible, so I knew what I wanted the UI to be. Claude and I had long conversations about colors, where data and icons sit, logos, settings, you name it. I thought through every setting and idea I wanted in the app. Claude took all of it and constructed the app piece by piece. Big chunks of churning, then quick little fixes. At the end of all that directing, I had a 0.1.0 release. Since then there've been follow-up bugs and features, and as of this writing we're at v0.2.2 of my new app, Spex Glance. I even had Claude build a site to showcase my app (and future apps) using back and forth decisions.

So what's my takeaway? What was I hoping to get out of this?

For one, I fixed the way I keep up with my open Kalshi orders. After the app shipped I had a real sense of accomplishment, and that part was great. It would've taken me years to make what Claude made in 24 hours. But the fundamental takeaways are these:

  • With the right person in control, an app that fills a gap can be made for just about anything. My decades of UI and software experience are what let me design and ship something other people might find useful.
  • If you've got a hobby app or a small business need, Claude can stand in for hiring a third-party developer.
  • I now understand the whole development process, conception to deployed. From an IT standpoint, I can support development teams at any company that needs to fold AI into its toolbox.
  • Humans are the final decision makers. AI shouldn't be in total control of the process. It should probe you with questions, give recommendations, accept that it's wrong every now and then, and double check its own work against other sources.

I learned something new and I can now say I've shipped software. I've gotten comfortable enough with the tools, through Claude, to make the app I wanted.

If you'd like to check out Spex Glance (more apps are planned and already in progress with Claude), head over to spex.bet.

Links

Read more