No-Code, Low-Code, and Low-Effort-Code

Show and Tell
1

UPDATE: While this post is relatively dated (more than a year ago) and from the original Airtable Community, it is now suddenly mainstream in the wake of GPT attracting ~100 million users in a few months. Being one of the fortunate developers who leapt at the chance to test CoPilot during the alpha phase, I have used it in code production since mid-2021. I thought it was worth resurrecting this observation here and in a new light as GPT-4’s release is imminent.

Two of these terms are increasingly familiar to you; the third one – low-effort-code – is new, but you’ll be hearing about it more and the engineers at Airtable are likely very familiar with it. I just wish they’d bake the underlying AI into the Airtable code editors. Maybe the folks at SeaTable will listen because they support javascript and Python and GPT-3 supports both. :winking_face:

When I’m writing code, I have another programmer with me every day all the time. It’s an AI model that is constantly scanning my source code and making really fast lookups into about a trillion lines of code written by other programmers. It’s powered by GPT-3 and delivered as a plugin to VS Studio Code.

I typically use the VS Studio code editor to create Airtable scripts and as a result, my paired programmer has already learned how to write Airtable SDK scripts. As such, my low-code activities with Airtable have become more optimized - truly low-effort-code.

Here’s a little video demonstrating that this AI tool even knows how to comment my code.

https://cdn.loom.com/sessions/transcoded/5bcd4bed44a54912954b05debf692f79.mp4?Expires=1958508156&Policy=eyJTdGF0ZW1lbnQiOlt7IlJlc291cmNlIjoiaHR0cHM6Ly9jZG4ubG9vbS5jb20vc2Vzc2lvbnMvdHJhbnNjb2RlZC81YmNkNGJlZDQ0YTU0OTEyOTU0YjA1ZGViZjY5MmY3OS5tcDQiLCJDb25kaXRpb24iOnsiRGF0ZUxlc3NUaGFuIjp7IkFXUzpFcG9jaFRpbWUiOjE5NTg1MDgxNTZ9fX1dfQ__&Signature=mz26MyE~fuCidAv0vD0dzrVUi7-npWrnCz8Cd~2RAXF9gj9hD0QvmofioJrAy5Z63-uKnl4CYWSaeL5py4romtpml7HudsyrVXkVnlVOKr2CO-T8jJx89fRCY15DUiiQaEB3Ag3dFH5Z21gynkHv0~T7ahW1Xin7mqr-mXrzhle2rT-PSVpynfftH-5PhmPzFSxPNBBaFqmbUPGuEL0tntbWT8ZmEur3V~RRsNp~QiWt39wNsCTtHXgmNH9-jgsPBf1sw3i-rjXwk~uL~Y-kS7FvSejrlFEhg1ArsRXAItQ3DwoD7I0bkQnwxuUpcxZpDLP7W6fSmXnFW97o27QNYQ__&Key-Pair-Id=APKAJQIC5BGSW7XXK7FQ

3 Likes
2

I am using GitHub copilot in my VS Code and I absolutely swear by it. It saves me a lot of time typing by providing really good code completions. I usually use it for completions of 1-2 lines and that goes very smooth.

I am hesitant to take bigger chunks as suggestions, as sometimes even if the suggested logic is fine, I might miss in a big chunk that e.g. variables are named slightly differently. That results in painful debugging, but the real question is who named variable in a wrong way me or Copilot :joy:

The future is clearly in progress.

1 Like