Roupen Odabashian, Hematology/Oncology Fellow at Karmanos Cancer Institute, and Podcast Host at OncoDaily, shared a post on X:
“I don’t know if this phenomenon exists, but I’m experiencing ‘model fatigue’.
Over the last three years, every single day I got excited. I looked forward to each new model release.
But now? I really don’t care.
Here’s what changed:
The marginal benefits are declining. GPT-4 to GPT-4.5 to 5.1 to 5.2 each jump feels smaller. The wow factor is gone.
I already have tasks where models work. My workflows are set. I’ve found what works for documentation, for research summaries, for drafting content.
It’s exhausting to keep up. New model every week. New features. New pricing. New context windows. New benchmarks that don’t reflect real use.
I feel we’ve hit the plateau of practical utility.
For 90% of what I do, GPT-4 from 18 months ago would still work fine. The incremental gains don’t justify the constant context-switching.
Here’s what nobody talks about:
The industry moves faster than adoption. We’re launching models faster than people can integrate the last one.
We’re optimizing benchmarks, not workflows. A model that scores 2% higher on USMLE or AIME doesn’t make my Tuesday easier.
We’ve gone from ‘AI can do this?!’ to ‘Oh, another model.’ The excitement has turned into noise.
Maybe this is maturity, not fatigue.
When a technology stops feeling magical and starts feeling like a reliable tool, that’s actually progress.
I don’t get excited about new versions of Excel. I just use Excel.
Maybe that’s where we are with AI models. They work. They’re useful. And that’s enough.
I think we are so far behind in understanding integration and how this technology integrate to our workflow.
Anyone else feeling this? Or am I the only one who stopped refreshing the model release calendar?”
More posts featuring Roupen Odabashian.