Elektrine lite

← Feed

Emil Sit

emilsit@discuss.systems

<p>I lead and develop leaders in technical organizations. This is a personal account.</p>

Posts

  • View post

    RE: https://mastodon.social/@scalzi/117328348252321409 I did say nice things about the security write-up that Meta posted, but that was just the theory. Making something secure in practice is much harder, especially with LLMs in the mix confidently making bad decisions.

  • View post

    RE: https://mastodon.social/@mcc/117328124035599647 Guilty.

  • View post

    RE: https://toot.cafe/@baldur/117328308464019225 Imma ironically boost this without watching the video 👍

  • View post

    RE: https://hachyderm.io/@hazelweakly/117328020199133154 “I’m in a FinOps review every Thursday. The AI team has never ever been to FinOps” 😭

  • View post

    If you train something (an LLM) to perform well in a specific, forgiving environment (harness)… turns out it might not perform well in other ones. Claude Code being slop-y makes their models worse https://lucumr.pocoo.org/2026/7/4/better-models-worse-tools/