2025-08-09XThread

Source item

We could address this via constrained execution. Constrain output length, like the 140 character limit on Twitter. Or constrain the runtime, like real-time mode in Linux. https://t.co/f9nVR7FS7I You don’t need to use the concepts exactly, but ideas from real-time OS development could be baked in as constraints during AI training and evaluation. Soft real-time may be sufficient. Unless it’s a car crash if it doesn’t work, like hard real-time. https://t.co/OpEnybplIy

2 segments

Segment 1 · characters 0–196 · Original segment ↗

Segment 2 · characters 198–473 · Original segment ↗

Quoted context

These are the quoted author’s words.

I'm noticing that due to (I think?) a lot of benchmarkmaxxing on long horizon tasks, LLMs are becoming a little too agentic by default, a little beyond my average use case. For example in coding, the models now tend to reason for a fairly long time, they have an inclination to start listing and grepping files all across the entire repo, they do repeated web searchers, they over-analyze and over-think little rare edge cases even in code that is knowingly incomplete and under active development, and often come back ~minutes later even for simple queries. This might make sense for long-running tasks but it's less of a good fit for more "in the loop" iterated development that I still do a lot of, or if I'm just looking for a quick spot check before running a script, just in case I got some indexing wrong or made some dumb error. So I find myself quite often stopping the LLMs with variations of "Stop, you're way overthinking this. Look at only this single file. Do not use any tools. Do not over-engineer", etc. Basically as the default starts to slowly creep into the "ultrathink" super agentic mode, I feel a need for the reverse, and more generally good ways to indicate or communicate intent / stakes, from "just have a quick look" all the way to "go off for 30 minutes, come back when absolutely certain".
karpathy ↗