About 70% of the commands I need on a daily basis, I've already ran it someday. So I record every cmdline in my bash/zsh sessions with some prompt magic (history 1|cut -c7-) and use an alias ("hag": history-silver-searcher) to search the .log files, copy-paste them and done.
For the other 30% of commands, bringing chatGPT slippery tongue right into my session feels suicidal. Actually, a simple, well-crafted command builder that can query real-life recipes would do. Then I can copy-paste without shame and edit accordingly, the same way I do with "hag" or maybe with bash tab-completion.
This cookbook searcher would be built from a good corpus of command histories like mine and from others (ie. extracted from Stackoverflow and Github resources or even chatGPT), trained into a much, much simpler ML model that fits the bill and landlocked to my personal realms.
Here's an outdated, yet illustrative, basic example:
This is a good idea that you could likely build out using just embeddings and performing semantic search over them. ChatGPT could be relegated to a role of translating intent into semantic searches and refining the output. You could even imagine fine-tuning an existing LLM to do this as well given a large enough corpus of examples.
The problem with this implementation is that it just blindly executes whatever ChatGPT says - that's quite scary.
Yup. But the thing I would like to alert to *GPT startup hopefuls is that first-to-market won't cut it here. Raising will require wow tech that cannot be replicated by the VC's 11-year old kid with an OpenAI API key.
For the other 30% of commands, bringing chatGPT slippery tongue right into my session feels suicidal. Actually, a simple, well-crafted command builder that can query real-life recipes would do. Then I can copy-paste without shame and edit accordingly, the same way I do with "hag" or maybe with bash tab-completion.
This cookbook searcher would be built from a good corpus of command histories like mine and from others (ie. extracted from Stackoverflow and Github resources or even chatGPT), trained into a much, much simpler ML model that fits the bill and landlocked to my personal realms.
Here's an outdated, yet illustrative, basic example:
https://medium.com/unkempt-thoughts/jeeves-predicting-bash-c...