1. A 200k context is bittersweet with that 70k->195k error rate jump. Kudos on that midsection error reduction, though!
2. I wish Claude had fewer refusals (as erroneously claimed in the title). Until Anthropic stops heavily censoring Claude, the model is borderline useless. I just don't have time, energy, or inclination to fight my tools. I decide how to use my tools, not the other way 'round. Until Anthropic stops injecting bias into their models to create some byzantine, manic LLM omertà, I'll stick to more effective models, thanks. I'm too swamped to add "tech company decided what's best for me this time" app bugs to my backlog.
2. I wish Claude had fewer refusals (as erroneously claimed in the title). Until Anthropic stops heavily censoring Claude, the model is borderline useless. I just don't have time, energy, or inclination to fight my tools. I decide how to use my tools, not the other way 'round. Until Anthropic stops injecting bias into their models to create some byzantine, manic LLM omertà, I'll stick to more effective models, thanks. I'm too swamped to add "tech company decided what's best for me this time" app bugs to my backlog.
[EDIT] To avoid replies to further "the only people who want privacy must have something to hide" style arguments, my reply: https://news.ycombinator.com/item?id=38368352