When Modernization Breaks Working Code: A Study of LLMs on Legacy PHP(arxiv.org)
22 pointsby tambourine_man2 days ago9 comments
  • plorkyerana day ago
    > We study a common intermediate setting: a developer pastes one complete file into a chat assistant [5], which sees the file’s internal structure but not its callers or dependencies. We evaluate the first answer, not an agent’s result after iterating on errors.

    Very unsurprising results given this methodology. The agentic approach where models are allowed to iterate on changes is a very big part of what has made LLM coding tools actually useful.

    • yodona day ago
      > We study a common intermediate setting: a developer pastes one complete file into a chat assistant [5], which sees the file’s internal structure but not its callers or dependencies.

      Ack! Do people actually try to develop software this way?

      • plorkyerana day ago
        Not any more, but it was common before Claude Code came along (~18 months ago). Based on the model versions used it may have still been the most common thing when the paper was started.
      • refactor_mastera day ago
        I do it from time to time when I’ve run out of tokens at the end of the month. But even then I’ve created a script to pack multiple code files and their dependencies into a single txt, with a tree overview.

        And of course you never change the test suite in the process… That would be bad developer practice, LLM or not.

        Gemini 3.1 Pro can easily work with this effectively, if not a bit slower than an agent.

        • 1718627440a day ago
          > But even then I’ve created a script to pack multiple code files and their dependencies into a single txt, with a tree overview.

          Like:

              $ cat $(find project/ -name '*.py' -type f)

?
      • 2000UltraDeluxea day ago
        Non-technical people often have no clue that there are options
      • customguya day ago
        No, they test what LLM are capable of this way. They then examine the results by reading the code and reasoning about the changes the LLM made, rather than just throwing up their arms and going "Ack!" as if that's anything.
      • antisthenesa day ago
        When I need modification to a small independent Python script (<1000 loc), sure.

        Not for anything serious though.

      • TZubiria day ago
        It's pretty much the standard way to develop in OOP.

        You are not supposed to hold all of the code in context, but to have a single object and reason about with what it see.s

  • kgeista day ago
    The average age of the explored models is ~1.5 years. They discuss PHP 8.3, which was released in 2023 and has not been actively supported since 2025. Agentic coding isn't used.

    I'm not sure what the point is of publishing results on the mid-2025 state of the art in October 2026.

  • StevenWatermana day ago
    Models:

    - Claude Sonnet

    - Gemini 2.5 Flash

    - Gemini 2.5 Pro

    - GPT-5 Codex

    - Llama 3.3 70B

    It's inconvenient how quickly things are improving, makes potentially-interesting results like this dubiously relevant

  • AdrianB1a day ago
    One year ago I was very skeptical about such attempts. Today, it would be my first choice as LLMs evolved a lot and the result of carefully using a LLM for this task can be very, very good.
  • TZubiria day ago
    When code output becomes cheap, you write 15 million lines of code per weekend, and of course everything gets deprecated pretty fast version 1? that was like 4 geological ages ago (1 month) that's legacy.
  • apishare_apia day ago
    [flagged]
  • ahmedk292 days ago
    [flagged]
  • madebyezra6 hours ago
    [flagged]
  • emscba day ago
    [flagged]