Generative History:Google is A/B testing a new model (Gemini 3?) in AI Studio. I tried my hardest 18th century handwritten document. Terrible writing and full of spelling and grammatical errors that predictive LLMs want to correct. The new model was very nearly perfect. No other model is close.
I had the opposite experience: I tried to transcribe & let an LLM correct my handwritten notebook data for gym and daygame approaches. I tried Claude Opus 4.1 and Sonnet 4.5, with basically useless results for both. I’ll probably try this with GPT-5-Thinking soon (I heard it has better image capabilities) and Gemini 3 when it comes out, but I’d be surprised if it did much better. My guess is that it’d take an unskilled human a week to transcribe the data correctly.
I had the opposite experience: I tried to transcribe & let an LLM correct my handwritten notebook data for gym and daygame approaches. I tried Claude Opus 4.1 and Sonnet 4.5, with basically useless results for both. I’ll probably try this with GPT-5-Thinking soon (I heard it has better image capabilities) and Gemini 3 when it comes out, but I’d be surprised if it did much better. My guess is that it’d take an unskilled human a week to transcribe the data correctly.