Calling llms word blenders or random text generators feels similar to dismissing early computers as number blenders. Not a great fit for many situations, but already showing lots of potential

By mimicking an inner thought process they certainly seem better at reasoning. The models with CoT can explore bad ideas and course correct before generating output.

Does pretty well if you use the current model

What's your preferred way of dealing with internal and third party errors in large rust apps?

Yeah that's a great tip. Was thinking I'd make a headless game server and have participants get state and send controls via http so I would provide a basic client (with fully typed API endpoints) as a starting point.

Oh very nice, could be a winner 🏆

Taking the boat up to the sounds is on my bucket list for sure

Safely back to Lyttleton

Lunch aboard

Plenty of dolphins joined us!

Smooth sailing! Stopped in at Okains bay

Very light wind in the morning but enough to sail