4 ms·

This is a great anecdote. SOTA models will not provide “engineering” per se, but they will easily double productivity of a product manager that is exploring ne

This is a great anecdote. SOTA models will not provide “engineering” per se, but they will easily double productivity of a product manager that is exploring new product ideas or technologies. They are much more than intelligent auto-complete. I have done more with side projects in the last year than I did in the preceding decade.

llm_trw 2y ago
One of my friends put it best: I just did a months worth of experimentation in two hours.
- Sateeshm 2y ago
  I find this hard to believe. Can someone give me an example of something that takes months that AI can correctly do in hours?
  - jvanveen 2y ago
    Not hours; but days instead of months: porting around 30k lines of legacy livescript project to typescript. Most of the work is in tweaking a prompt for Claude (using Aider) so the porting process is done correctly.
    - cdchn 2y ago
      Thankfully it seems like AI is best at automating the most tedious and arguably most useless endeavor in software engineering- rewriting perfectly good code in whatever the language du jour is.
      - disgruntledphd2 2y ago
        Again, what AI is good at shows the revealed preferences of the training data, so it does make sense that it would excel at pointless rewrites.
      - protomolecule 2y ago
        Legacy code in a dynamically typed language is never good.
  - llm_trw 2y ago
    Use Undermind to gather a literature review of a field adjacent to the one you’re working in but with a wealth of information that you don’t yet know. Use OpenAI to convert a few thousand lines of code from a language you're familiar with to one you’re not, as all the state-of-the-art tools in the field above use that language. Debug all the issues that arise from the impedance mismatch between the languages. Recreate the results from the seminal paper in the field to verify that the code works, and run it on your own problem. Write a stream-of-consciousness post without spell-checking, then throw it into GPT and ask it to fix it.
- hnisoss 2y ago
  sounds to me like you're tooting your own horn.

llm_trw 2y ago

One of my friends put it best: I just did a months worth of experimentation in two hours.

Sateeshm 2y ago
I find this hard to believe. Can someone give me an example of something that takes months that AI can correctly do in hours?
- jvanveen 2y ago
  Not hours; but days instead of months: porting around 30k lines of legacy livescript project to typescript. Most of the work is in tweaking a prompt for Claude (using Aider) so the porting process is done correctly.
  - cdchn 2y ago
    Thankfully it seems like AI is best at automating the most tedious and arguably most useless endeavor in software engineering- rewriting perfectly good code in whatever the language du jour is.
    - disgruntledphd2 2y ago
      Again, what AI is good at shows the revealed preferences of the training data, so it does make sense that it would excel at pointless rewrites.
    - protomolecule 2y ago
      Legacy code in a dynamically typed language is never good.
- llm_trw 2y ago
  Use Undermind to gather a literature review of a field adjacent to the one you’re working in but with a wealth of information that you don’t yet know. Use OpenAI to convert a few thousand lines of code from a language you're familiar with to one you’re not, as all the state-of-the-art tools in the field above use that language. Debug all the issues that arise from the impedance mismatch between the languages. Recreate the results from the seminal paper in the field to verify that the code works, and run it on your own problem. Write a stream-of-consciousness post without spell-checking, then throw it into GPT and ask it to fix it.
hnisoss 2y ago
sounds to me like you're tooting your own horn.