Claude 5.5, ChatGPT, and the fun of going back and forth.
Two models, two different ways of getting to a good idea, and The Elkaholic as the latest test.
The latest round
Claude 5.5 has been my latest test, and the efficiency has been genuinely impressive. Over the last couple of days, most of that testing happened on The Elkaholic: giving it real problems, seeing what it understood, and watching what it could make.
Two new ways into the wild
The latest additions are Elk Pinball and Bugle Off. Elk Pinball is a playable arcade table with flippers, ramps, a wolf pack, multiball, and real elk audio. Bugle Off turns the microphone into a challenge: hear a real Yellowstone bull, take your turn, and get scored on the shape of your call. Both are silly in the best way, and both make learning feel like something you want to keep exploring.
The back-and-forth is the point
I keep moving between Claude and ChatGPT with OpenAI. One gets an idea moving; the other catches an edge case, suggests a different structure, or helps me see the thing again. I’m less interested in declaring a permanent winner than in noticing which model is most useful for the problem in front of me.
Why the efficiency matters
When the cost of trying an idea drops, you get more attempts. More attempts mean more chances to find the version that has a little spark in it. That has been the most fun part: not just getting a website built, but having enough room to play, revise, and follow a weird idea further than I expected.
What comes next
I’m not sure what I’ll create next. That is part of the appeal. The models keep changing, the tools keep getting better, and the best use of them still seems to be staying curious long enough to make something real.