Jump to main content

Nicolas Hoizey

  • articles
  • billets
  • links
  • notes
  • talks
  • archives
  • about

Grant Gross

Devs gaining little (if anything) from AI coding assistants

  • Nicolas Hoizey
  • 7 October 2024
  • AI, LLM, Copilot
  • 86 reactions
Screenshot of “Devs gaining little (if anything) from AI coding assistants”

https://www.cio.com/article/3540579/devs-gaining-little-if-anything-from-ai-coding-assistants.html

“It becomes increasingly more challenging to understand and debug the AI-generated code, and troubleshooting becomes so resource-intensive that it is easier to rewrite the code from scratch than fix it,” says [Ivan Gekht].

86 reactions

35 likes

  1. Rye avatar
  2. Cory Dransfeldt :demi: avatar
  3. eyenoaino avatar
  4. ToastyFlux :mastodon: avatar
  5. Tommi ???? avatar
  6. Darek Kay avatar
  7. Baldur Bjarnason avatar
  8. Michelle Barker avatar
  9. Seen_Canary avatar
  10. llewelly avatar
  11. Konstantin ???? avatar
  12. lj·rk avatar
  13. Alien software, human hardware avatar
  14. Skylar MacDonald avatar
  15. ^,,^ :flanell: avatar
  16. Hen Gymro Heb Wlad avatar
  17. Dave Spector avatar
  18. GP avatar
  19. Mike Anderson avatar
  20. Caroline avatar
  21. angelthorns avatar
  22. marymessall avatar
  23. James Earl Douglas avatar
  24. Steve avatar
  25. Troels Thomsen avatar
  26. Kaosc avatar
  27. ms. liz avatar
  28. Chris Johnson avatar
  29. Paul Wouters avatar
  30. MatthewToad43 avatar
  31. Fancy Sandwiches avatar
  32. surfingreg avatar
  33. Nick avatar
  34. Cory Dransfeldt :demi: avatar
  35. Marc Friederich avatar

40 reposts

  1. Chris Ferdinandi ⚓️ avatar
  2. AI Feed avatar
  3. Cory Dransfeldt :demi: avatar
  4. kay@foodnpolitics avatar
  5. Stéphane Deschamps avatar
  6. BohwaZ avatar
  7. Atx :d20: avatar
  8. Spookybot avatar
  9. Mollie Garza :fuego: avatar
  10. Bartosz avatar
  11. Anais Crosby avatar
  12. Tom Morris avatar
  13. Baldur Bjarnason avatar
  14. Alien software, human hardware avatar
  15. llewelly avatar
  16. Seen_Canary avatar
  17. Debora Weber-Wulff avatar
  18. delCano :d12: :savage: avatar
  19. Peter Krupa avatar
  20. lj·rk avatar
  21. [object Object] avatar
  22. Uncommon Loon avatar
  23. Graham K avatar
  24. Don-kun avatar
  25. Stephen Homewood avatar
  26. JRT avatar
  27. Kaosc avatar
  28. Chris Johnson avatar
  29. Paul Wouters avatar
  30. kat avatar
  31. Jonathan Koomey avatar
  32. samiamsam avatar
  33. MatthewToad43 avatar
  34. Alice Marshall avatar
  35. Cruiser avatar
  36. Fish Id Wardrobe avatar
  37. Shonin avatar
  38. Tofu Musubi avatar
  39. Cory Dransfeldt :demi: avatar
  40. AI Feed avatar

11 replies

  1. Rye avatar Rye
    @nhoizey it's so bad, at best it can help with writing a test but even then I've yet to see it get it right to a degree I can trust it
    • 28 November 2024, 03:13
    • Source
  2. Ike avatar Ike
    @nhoizey This tracks Brandolini's law[1]: The amount of energy needed to refute bullshit is an order of magnitude bigger than that needed to produce it.Note that in this case "effort" refers to the user's and not the compute effort.[1]: https://en.wikipedia.org/wiki/Brandolini's_law

    Brandolini's law - Wikipedia
    • 28 November 2024, 04:34
    • Source
  3. Nicolas Hoizey avatar Nicolas Hoizey
    @rye exactly, I have the same feeling
    • 28 November 2024, 08:57
    • Source
  4. Nicolas Hoizey avatar Nicolas Hoizey
    @ike indeed ????
    • 28 November 2024, 08:58
    • Source
  5. Hen Gymro Heb Wlad avatar Hen Gymro Heb Wlad
    @nhoizey My wife is a professional translator. She has the same experience with most "AI" translation tools. Fine if you just want something that *looks* like program code/a translation in the target language. But for complex tasks where the solution needs to be *good*, you need a human in the loop. And then it's usually easier to let a skilled human do the job properly in the first place.
    • 28 November 2024, 12:03
    • Source
  6. Ondřej Synáček avatar Ondřej Synáček
    @nhoizey It's really hard to tell where the conversation with LLMs is going. You have one side telling you it's inevitable future — feels like Big Tech employers want to take back power from engineers; and than the other one dismissing it. There's some usefulness there but it's probably not sustainable at the level of power/water consumption vs its value.
    • 28 November 2024, 15:29
    • Source
  7. Nicolas Hoizey avatar Nicolas Hoizey
    @hengymrohebwlad I agree. I know multiple professional translators. ????
    • 28 November 2024, 18:31
    • Source
  8. Nicolas Hoizey avatar Nicolas Hoizey
    @ondrej it’s difficult to know indeed, and I agree some of it can be useful.
    • 28 November 2024, 18:33
    • Source
  9. Fancy Sandwiches avatar Fancy Sandwiches
    @nhoizey who could have possibly predicted this
    • 29 November 2024, 02:38
    • Source
  10. StaringAtClouds avatar StaringAtClouds
    @nhoizey LLM generated code generally needs to be checked & corrected by a "dev that knows what they're doing"But all "devs that know what they're doing" were once entry level or "devs that need to learn stuff"Now LLM's have taken the place of the entry level devsSo once all "devs that know what they're doing" move on, leave or even dieWe'll just be left with the LLM's & no one to check & correct their output
    • 29 November 2024, 03:47
    • Source
  11. Emmanuel ???? avatar Emmanuel ????
    @nhoizey Déjà qu'on a parfois du mal à relire son propre code… ????
    • 6 December 2024, 17:49
    • Source
  • Older: Separating content and presentation: Moving past FUD
  • Newer: SPAs: theory versus practice

Related contents with similar topics

  1. screenshot of Writing Code Was Never The Bottleneck

    Pedro Tavares

    Writing Code Was Never The Bottleneck

    • Nicolas Hoizey
    • 8 July 2025
    • AI, LLM, development

    The marginal cost of adding new software is approaching zero, especially with LLMs. But what is the price of understanding, testing, and trusting…

  2. screenshot of Why extracting data from PDFs is still a nightmare for data experts

    Benj Edwards

    Why extracting data from PDFs is still a nightmare for data experts

    • Nicolas Hoizey
    • 11 March 2025
    • AI, OCR, LLM, PDF
    • 11 reactions

    Despite their promise, LLMs introduce several new problems to document processing. Among them, they can introduce confabulations or hallucinations…

  3. screenshot of Why Copilot is Making Programmers Worse at Programming

    Darren Horrocks

    Why Copilot is Making Programmers Worse at Programming

    • Nicolas Hoizey
    • 24 September 2024
    • AI, LLM, Copilot
    • 12 reactions

    If you as the developer do not understand the code, do not understand how you got to where you are, do not understand how to solve the problem yourself.…

  4. screenshot of Models All The Way Down

    Christo Buschek and Jer Thorp

    Models All The Way Down

    • Nicolas Hoizey
    • 9 April 2024
    • AI, LLM
    • 2 reactions

    As artists, academics, practitioners, or as journalists, dataset investigation is one of the few tools we have available to gain insight and understanding…

If you want to share an error or suggest an enhancement of this content, please edit the source on GitHub.

© Nicolas Hoizey

Built with Eleventy