Artificial Intelligence

Out of context: Reply #2081

  • Started
  • Last post
  • 2,575 Responses
  • ********
    -7

    GPT-5 launch looks like a fiasco:

    "For a year or two I have been speculating that OpenAI might take a serious hit if GPT-5 was disappointing. We may finally soon find out.

    Certainly, in a rational world, their valuation would take a hit.

    They no longer have anything like a clear technical lead.

    GPT-5 is unlikely to be ahead of the pack for more than a couple months. (And Grok 4 Heavy is already better on the ARC-AGI-2 measure)

    Many of their best people have left.

    Many of those people left to start competitors.

    Elon is moving faster. Anthropic and Google and many others are nipping at their heels. Their relationship with Microsoft has frayed.

    OpenAI still isn’t making profit.

    Instead they are being forced to cut prices.

    People are wising up that LLMs are not in fact AGI-adjacent.

    People are becoming more skeptical about the company and its CEO.

    OpenAI has the name brand recognition, and good UX. Will that be enough to sustain a $300-500B valuation? Hard to know."

    https://garymarcus.substack.com/…

    • gpt5 is great. 4o was the disaster, it just did well with plebs because it was sycophantic. gemini is similarly gash and opus is too expensive.kingsteven
    • I think it depends a lot what you're using it for. Right out of the gate I'm seeing slightly worse performance with GPT-5 vs o3 (never used 4o much).yuekit
    • I'm getting way less hallucinations and a shitload more speed with 5YakuZoku
    • Which GPT-5 is great? And for what?
      ********
    • gpt-5 is not only a massive improvement for coding and automation over 4o, it dosen't have the sycophantic tone, constant bullshitting and hallucinations -kingsteven
    • the reason why i have been using claude. and stayed away from open ai and google. yet to see if the long context support in claude will put it back on top.kingsteven
    • I'm not dropping entire codebases and eating up tokens. I paid nearly $200 to Anthropic last month without using Opus once.kingsteven
    • Claude code does have advantages but sonnet does really fuck up sometimes. GPT5 may still have the edge for general coding tasks and price.kingsteven
    • What are you coding?
      ********
    • i work in ed tech and have a side job building twitch extensions and tools for streamers - so its a good variety. this month a game engine for developing twitchkingsteven
    • chat games (like a very niche browser based flash) and an app for delivering and evaluating situational judgement tests.kingsteven
    • Oh you have your plate full!

      Good luck!
      ********
    • i've been a developer for 20+ years so it's not much of a change, i spend as much time manually coding as i do engaging with AI agents. the more i dig in to itkingsteven
    • the more i find people creating products 'using AI' know exactly what they want and are are just bridging a small skills gap.kingsteven
    • i think the sycophantic nature of the 'assistant' actually fools a lot of people in to believing that they couldn't have done it themselveskingsteven
    • meanwhile plenty of 'developers' in the design industry have never had an original thought, built a career on implementing libraries and copy/pasting code.kingsteven
    • It removes the learning curve for new coders, which then reinforces the lack of true coding skills in the industry. How long until old school coders are gone?monospaced
    • just grifters for now. i've had two people contact me this week asking how to get their AI generated apps to work! the jobs im not seeing any more are the oneskingsteven
    • that are the ones so small it's awkward to charge for - fixing up a bit of css for someone who got 99% of the way themselves.kingsteven

View thread