Elektrine lite

← Feed

@borgbackup@fosstodon.org

2026-07-22 20:48 UTC

@argv_minus_one@mastodon.sdf.org "this bad at programming" seems to be something you experienced. My experience is quite different, so I disagree. It's hard to see whether something was reviewed, but that is not an AI specific problem. You can also not see how much a human developer was thinking while they developed some piece of code. A small bit of evidence (but no proof) sometimes can be seen if there is a big change (AI assisted) followed by some small changes (human corrections).

Replies (2)

  • @borgbackup@fosstodon.org 2026-07-22 20:57

    @argv_minus_one@mastodon.sdf.org To get a bit more practical / specific: Did you notice the increased amount of bugs being fixed recently? Tons of CVEs and also lots of non-CVE fixes. Commercial SW as well as FOSS. Quite a lot of these were discovered by AI. A lot of these were in the code since many years and were NOT discovered by humans - neither in code review at commit time, nor in the years after that.

    Open ##4137139

  • @borgbackup@fosstodon.org “"this bad at programming" seems to be something you experienced.” Indeed. Repeatedly. With several different LLMs, some of which are named Claude. In light of which, you'll have to forgive me when I start to consider it a red flag that AI is used to write safety-critical code. (Losing my backups could ruin my life, so Borg is safety-critical.) “It's hard to see whether something was reviewed, but that is not an AI specific problem.” Indeed, but as I said, AI use is a red flag.

    Open ##5175201