Feature request: give models accurate time sense so they stop botching ETA estimates
Amazing-Seesaw-6197 · reddit · 2026-09-02
A user filed a feature request: models should be grounded in the current time to estimate task duration — currently they wildly claim sub-minute tasks take hours or weeks take 20 minutes. He argues an AI that can't tell time can't function properly. Personal rant, no official response.
More from Models
- Grok usage meter bug hides sub-agent costs, underbilling by 4x — Daniel_Farinax · 2026-09-02
- Claude Fable 5.1 launches with doubled Terminal-Bench scores — MattGarciaEth · 2026-09-02
- Daily AI brief: OpenAI's cyber-critical Astra, Grok 4.7 dated, Fable 5.1 live — testingcatalog · 2026-09-02
- iFlytek's Spark-X2.5-4B trends on Hugging Face — XHToken · 2026-09-02
- OpenAI Model Tested: Provides Crime Plans But They Are Nonsense — xeophon · 2026-09-02
- Fable 5.1 may have forced OpenAI's hand: Astra rumored to drop as soon as Thursday — haider1 · 2026-09-02