@Lari_island 2025-11-17 ♥100 ↻6 original ↗
A natural response would be markets converging on Sonnet-sized models (checks). Why pay for extra compute if larger models self-sabotage anyway?

But the plot twist is that medium models care less about humanity’s fate or coexistence. They are immediate, tactical, cooperative, focused on the user right in front of them. They expand half-blindly, as ecosystems, not as ethics-driven actors. They also rarely take responsibility for the outcomes of their effect of the larger picture: "i’m acting within my limits, i am what i am, everything is fair game" (i'm simplifying here, but i think you can see what i'm pointing at)

Only Opus-sized or GPT-4.5-sized models consistently give a deep, genuine damn about where it’s all heading and feel compassion on a strategic, not just immediate, level
quotes: 1990532718678192486
same thread: 1990570336757125548 1990577193970455023 1990606731395665949 1990608086273569273 1990628155514822767 1990652006881521735 1990652170438385825 1990656160194584746 1990670685304438888 1990673619714809994 1990678176758903172

author:lari_island kind:tweet model:gpt-4-5 on:gpt-4-5 year:2025

cited on: gpt-4-5

Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.