Choosing between AI chats

Short explanations of what actually differs between assistants, what is worth comparing, and what a switch costs. No sign-up needed to read any of it.

Four questions for any comparison
Four questions expose whether a comparison is useful: who wrote it, when, what was tested and what is missing from the list.
Three habits that make leaving cheap
Export before you need it, keep prompts outside the product, and write down what you learned. Three habits make any future move cheap.
Four tests worth 20 minutes
Four tests, 5 minutes each, tell you more than a week of casual use: a known answer, a hard question, a boundary case and a long conversation.
Four differences, and only one is the model
Four things differ and 3 are not the model: the hidden instructions, the filter and the data policy. Products differ mostly outside the model.
Five things that predict your experience
Five properties predict your experience and 4 popular ones do not. Comparing the wrong 4 is why so many switches disappoint.
The 4 costs of moving
Four costs, and only 1 is time. History, habits, prompts and integrations all have to move or be abandoned.
Four times to stay where you are
Four situations where the new product is better and moving still costs more than it returns. Recognising them saves a wasted month.
What a score does and does not measure
Benchmark scores measure narrow tasks under conditions unlike yours, and the gap between a score and daily use is structural rather than accidental.