Sometime in late January 2026, I decided to give Claude Code another try. I had tried a few times, spread out since its initial release, and had always quickly decided that AI for coding was useless (useless to me, from my 30 years of experience/POV). As you all know, this time felt very different - something had changed - an immediate “wow, this is much better than that last time”.
A couple of weeks later, our usual foursome for bridge found that we could not gather in person for our weekly Sunday game, so we signed up for a one-month, non-renewing subscription on an online bridge website. I will not name the site given what comes next…
Each of us found the signup and onboarding to be a horrible experience, both technically and from a customer support POV. We got that all sorted out, and then we started to play. What a miserable UX. A constant need to keep hitting an “I am still here” button to not get kicked out of the table, nothing more really needs to be said, as the rest of the UX was just as bad or worse.
I casually said to one of the foursome, “I bet I could create a much better online bridge game in short order”, using this as a prototype for my recent new view of Claude Code. Later that night, I had a basic, fully functional, deployed bridge website with only local account creation/management (including self-service registration, email verification, forgot my password, etc.). Was it perfect? Absolutely not, but it was surprising how “easy” it was to produce a working prototype - less than one day.
Next came all the feature creep, rubber vs duplicate scoring, lobby and table chat, how about an AI player (and then also a Mentor)??? By the end of the week, it was all there (all the feature creep), including AI players (to fill out a less than 4-person table, or to play as a single player) that frankly could beat members of our foursome.
So what’s the problem? I would say that this was, at best, the 80% mark; now the typically long 20% to hit release 1.0 and maybe start charging for play. Next, the rounds of “cleaning up the slop” have now begun. This is where, at least for now, years of human experience can see the issues and then redirect the AI to clean up at least some of the mess.
This first project with CC clarified how I needed to change the nature of my interactions with it to avoid some of these pitfalls up front. This was also the time when I got into that semi-manic “wow, what about this, or that, or that” (all projects related to my own interests in programming, photography, options trading, and my husband’s circle of architects struggling with their design automation tools) - but that is for a future writing including how launching down that path yet again refined how I interact with AI for coding…