Hands-On Report: Opus Made Multiple Errors Coding a MUD Bot, Needed Manual Fixes
jessi_cata · x · 2026-11-15
The developer noted her recent project was fairly easy eval-harness work, not hard AI programming. She attributes some of the issues to Codex rather than Opus, but also recalls using Opus to code a MUD game bot — where the model made multiple errors she had to manually point out and get fixed.
It echoes a common experience: even on moderate coding tasks, frontier models still can't be fully trusted without human review and correction in the loop.
More from Models
- Why do LLMs always end replies with a question? Reddit users suspect engagement tuning — TaurusGold · 2026-11-15
- Chart shows Gemini 2.5 Pro sliding from LMArena #1 to #42 in about a year — Sauers_ · 2026-11-15
- Dev says Google Antigravity's Hydrogen model is "VERY good and VERY cost effective" — jasonkneen · 2026-11-15
- Grok 4.8 reportedly rolling out to Grok Bot users, unconfirmed yet — mark_k · 2026-11-15
- Llama 405B Base Model Natively Mimics o1's Chain-of-Thought from 128 Pages of Context — Sauers_ · 2026-11-15
- Gemini 4 'Hydrogen' Rumored Faster Than GPT and Claude, a Third Model Class — mertdumenci · 2026-11-15