"Purple hair", "retarded", "media bad", "post 1945 order".

There's a bit of a "do what you have to" mentality with NY traffic that I haven't seen in any other east coast or mid-western city. I think that poses some unique challenges that I've often seen video of Waymos freezing up when facing similar scenarios, which could cause huge issues in most of the city.

It's literally the same pain point with low code solutions like WordPress page builders/plugins. Adding more becomes a hindrance, and even models with long context that can fit whole codebases will try to make up new functions that already exist. Just a couple weeks ago I had o3 continually try to write a new debounce function, even when I told it explicitly I had one.

That's the problem with most "AI" products/companies that still isn't being answered. Why do people use your tool/service if you don't own the LLM which is most of the underlying "engine"? And further, how do you stay competitive when your LLM provider starts to scale RL with whatever prompting tricks you're doing, making your product obsolete?

CoT builds on existing prompt engineering techniques by adding it to reinforcement learning to force the models to build their own CoT prompt essentially. So it's not what it's thinking but all indications are that it does guide the reasoning abilities of LLMs through the output distribution.