Beguène Permale

Live editing a react app with plain English featuring @OpenAI Edit GPT, @nextjs and precedent template from @steventey (speed 1.5x)

LLM will create Kodak moments for many

A classic is a book you keep turning the pages waiting for something to happen and find that you have turned the last page without anything ever happening!! 😅😂🤣

@TheSeaMouse so cool ! How reliable is it ?

(How to Write a (Lisp) Interpreter (in P̶y̶t̶h̶o̶n̶ GPT3))

Before AI it was publish OR perish. After AI it is publish AND perish.

It is far easier to learn the real subject properly than to learn the imitation badly

Is GPT just another Grand Parlor Trick ?

Trying to make ChatGPT behave like a random number generator. Its output is obviously not random but it's not nonsensical. It always outputs a value different than the previous output. It seems like it tries to maintain the expectation of the initial distribution

Soon we will have more podcasters than listeners

It’s not perfect, for example the footer is not sticky bottom. This uses a custom prompt just for React with the OG GPT-3 model to emulate ChatGPT

Creating React components in parallel with @OpenAI GPT-3

A prompt engineer is a restricted software engineer with a less restricted language 😆

@MichaelRouveure @OpenAI @nextjs @steventey With a custom elisp function in emacs. See my followup tweet above

@beguene @OpenAI @nextjs @steventey how did you plug openai into your code editor?

Live editing a react app with plain English featuring @OpenAI Edit GPT, @nextjs and precedent template from @steventey (speed 1.5x)

@sandstormcap @OpenAI @nextjs @steventey See my followup tweet above

@beguene @OpenAI @nextjs @steventey Incredible. How do you set this up?

Note that I tried around 5 prompts before that, the js and shell code output was correct, but the org mode format to output the code evaluation result in external files was not correct

This could be really valuable, since interacting with GPT leaves a trace, all development processes are recorded and therefore reproducible. Literate programming makes everything more readable, maintainable and reproducible www-cs-faculty.stanford.edu/~knuth/lp.html….

Creating a simple node.js api backend with @OpenAI GPT. The js code generated is exported to a new file with the command 'org-babel-tangle', shell cmds are executed with C-c C-c. Chatting with GPT like this looks a lot like literate programming

next experiment: a Node.js backend 🧐

This all works by sending the file buffer + the instruction to the openai api edit gpt beta.openai.com/docs/api-refer… which does all the work 😆. It then replaces the file with the api response and so on.

The code used to do this github.com/beguene/emacs-… Instructions are at the top of the file @steventey @harishkgarg @sandstormcap

Good software engineering is about finding the right trade-off between speed and quality.