This:

 Due note that this is a set of examples of the types of shortcuts in the sets of skills one can develop.

Most of it was figuring out what needed to go where.
Now the key in the example has been deleted. So. It is only good for example of where to put said key.

Beyond that, use any combination of these sets of information, one can build a career out of them. These are things valuable for building compatible parts for the thing you're building. The least number of hoops to jump through, in this case, the better. Trial and error is bound to happen, and no system is perfect. But, if you needed to, this is where I'd consider starting. Because if all one has to do is get the machine to do such with a small set of vocabulary, tools to get and return the data it needs via python functions, give it access to the CLI with all the file acorn, docker, and your kubernetes stack, etcetera, it can likely get itself to perform as good as a human engineer.

Give it modified bits of conversation correcting the actual error where it got 95%+ correct objectively, the last 1% to 5% error can be human corrected, and most of what you were trying to do should work. While the python library for openai in python doesn't have a way to store the usage at present of the conversation in the python shell, manual implementation should work fine. One can even store it in a replit database and with an additional function given, have it figure out where it left off last in the conversation based on the dictionary/json layout of the messages & calls that interface with the API.

Functionally this gives one an option to even give it bits that emulate brain function that have been trained separately that get rpc interfaced at a local point pretrained. Useful for if you want to make use of many MRI/megan test readouts or EEG in addition. This would potentially give one something closer to AI than what we have as of late.

And yes, this is a legitimate way to get GPT to talk to itself and complement itself even in a replit CircleCI container.

>

>