[darksaints]

I develop some deep learning models. They don't compete with Anthropic, nor are they language models. They mostly enable mathematical optimization systems to approximate actual the actual physics of radio propagation models with a fraction of the latency/compute of a high resolution simulator. Technically that should be safe for me to use with Claude Code, but how the fuck am I supposed to know? You're degrading/malware-ing your responses silently!

I won't ever trust Claude Code again. It's too late. I'd rather trust a less-than-frontier chinese model that takes a little longer to get to correct than a frontier model that deliberately deceives me at its own whim.

[rockinghigh]

Same here, I fine tune LLMs for specific use cases. How can I trust Anthropic models not to introduce bugs to preserve their moat?

[weakened_malloc]

This is why I think in the long run, the Chinese models will probably end up winning where it matters. You can get a cluster of relatively affordable 30 or 4090s, load up DeepSeek v4 and let it rip. Your only ongoing cost is power. We're already seeing companies recoil at the sight of their API bills from the frontier labs, for the price of 1 years worth of tokens you can host your own decent model that's 75% of the way there.