Login
You're viewing the mstdn.social public feed.
  • Sherri W (SyntaxSeed)syntaxseed@phpc.social
    Jul 3, 2026, 1:46 PM

    I still reference SO daily for my work. But how long until they have to take the entire site offline and we're left without this valuable resource? How do you even keep monetizing this?

    What a sad waste. StackOverflow had its flaws for sure, but it was an incredible resource of skills collaboration, knowledge collection and sharing.

    Hoovered up by #AI companies who will gate the information behind rising query token prices.

    The #internet continues to die and we're drinking the koolaid.

    💬 1🔄 0⭐ 0

Replies

  • Sherri W (SyntaxSeed)syntaxseed@phpc.social
    Jul 3, 2026, 1:47 PM

    How will programming evolve if the tools we're left with are trained on old techniques and languages and there's nowhere left to learn and collaborate on NEW things and ideas?

    💬 2🔄 1⭐ 0
  • 💬 4🔄 7⭐ 4
  • 💬 0🔄 0⭐ 0
  • Ian Littmanian@phpc.social
    Jul 3, 2026, 7:23 PM

    @syntaxseed So one interesting thing here is the subsidies are relative to "API pricing", not necessarily OpenAI/Anthropic's marginal cost per token.

    My guess is that OpenAI doesn't make much on someone using a $200 Codex plan flat-out, but that's after a ~40x subsidy vs. pay-per-token.

    We don't know how big their models are, but serving costs at scale for bigger open-weights models are on the order of 99% less than GPT-5.5 costs.

    Inference being a money-printer is why Nvidia can charge $$$.

    💬 0🔄 0⭐ 0
  • Aeryn is sort of a Fernaerynv2@tenforward.social
    Jul 3, 2026, 7:52 PM

    @syntaxseed my back-of-envelope first impression after being involved in a model-training effort is "it takes as much labour to train an LLM to do a task as it does to write a deterministic algorithm for the same task". The only reason executives are able to delude themselves that LLM tech is financially viable is because the training data is usually stolen from the community that produced it.

    💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • Gina Peter BanyardGirgias@phpc.social
    Jul 3, 2026, 2:19 PM

    @syntaxseed idk the issue is not people thinking about new tools, techniques and programming languages, but *evolving* existing tooling / programming languages is such a slow and tedious process.

    Many interesting and novel ideas *exist* but aren't in any mainstream PL. The closest we have got is Rust, and look at the reluctance and whack ideological rhetoric people have bashing it.
    (this is not to say it is perfect, but it's a significant improvement for a system PL)

    💬 0🔄 0⭐ 0