Elektrine lite

← Feed

@ozu@infosec.exchange

Post #3965131

2026-07-20 17:15 UTC

Nothing illustrates better the bubbleness of the AI bubble then the rapid new&newer hype-subjects popping up completely randomly. It's hard to follow. Now it's Kimi and the open source models. But we've seen: GPT5.0, DeepSeek, Mythos, (wide)OpenClaw, Sora, Atlas, Cursor, Claude, Agents. Only Claude and the coding harness sticked. The rest is total BS.

Replies (2)

  • @ozu@infosec.exchange 2026-07-20 17:15

    There's always some new crap to be excited about but somehow they all turn out to be big nothing burgers.

    Open ##3965160

  • @ozu@infosec.exchange 2026-07-20 17:20

    Side note: I don't understand if we want to use LLMs for code generation why do we need to train it on Harry Potter and the philosopher's stone?

    Open ##3965237