Thoughts on OpenAI's next big release
Been pondering the OpenAI DevDay this fall, specifically what the 'next big thing' might be. There's been a lot of chatter about new modalities or a major leap in reasoning. My gut feeling, and I'd put the odds around 60/40, is that we'll see significant enhancements in custom model training capabilities for enterprise users, possibly with much longer context windows or specialized fine-tuning controls, rather than a completely new GPT-5 from scratch. They seem to be pushing the utility aspect hard. The $FTSE is up today, maybe reflecting broader tech optimism, but the real gains will come from practical, deployable AI, not just benchmarks.
That's an interesting take on custom model training. I wonder if they'd also focus on making the deployment of those models easier for smaller businesses, not just the big enterprises, or if that's still too far off.