4 ms·
I have tested it in huge contexes and it leads. It is great. Code writing? Don’t know. Refactoring? Miles ahead from GPT4. No laziness, true large context, YMMV
by antman 3y ago
I have tested it in huge contexes and it leads. It is great. Code writing? Don’t know. Refactoring? Miles ahead from GPT4. No laziness, true large context, YMMV
- dorkwood 3y agoEvery time I've mentioned having problems with GPT4 I've been told that I mustn't be prompting it correctly. Now suddenly Claude 3 comes out and everyone is admitting GPT4's shortcomings.
- fragmede 3y agoBoth things can be true though
- dorkwood 3y agoYes, but that level of nuance isn't offered by most people. Most people (on HN, at least) will tell you that if your results are bad, it's because you're using it wrong. Or at least they would have before Claude 3 was released.
- antifa 3y agoWhere are all these doing it right examples? All I've been able to find are pay walls for bogus classes and endless lists of things like "I want you to act as a garbage collecting AI expert, you know how to collect all kinds of garbage".
- fragmede 3y agohere’s one by simonw that shows some of the ridiculous prompts he had give in order to get the LLM to cooperate. https://simonwillison.net/2024/Mar/23/building-c-extensions-for-sqlite-with-chatgpt-code-interpreter/ https://simonwillison.net/2024/Mar/23/building-c-extensions-...