Fetching from the wire…
Public story · 2026-07-30 · high
JuliaHub's July 30 test found a 0.366-point harness gap, more than double the 0.162 gap between the best and worst model.
Why now: JuliaHub's evaluation and OpenAI's harness settings post both landed July 30, alongside a July 29 study tying agent security complaints to configuration, not model weights.
JuliaHub ran four frontier models through five sealed modeling and simulation problems on July 30. The harness gap beat the model gap by more than double: swapping models produced a 0.162-point spread, swapping the harness produced 0.366. For teams building or shipping agents, that means the scaffolding wrapped around a model matters more than which model you pick.
The hardest problem was a full NASA HL-20 flight vehicle with six-degree-of-freedom dynamics. Claude Fable 5 took the top score at 0.889 and swept all twelve trials on the four core problems. GPT-5.6-Sol followed at 0.814, then GPT-5.6-Terra at 0.786, then GPT-5.6-Luna at 0.727.
OpenAI backed this up the same day. Turning on retained reasoning and compaction in the Responses API took GPT-5.6 Sol from 13.3% to 38.3% on the ARC-AGI-3 public set. Output tokens dropped sixfold, same weights.
Under the official evaluation harness, the same model scored as low as 7.8%. Its private chain of thought got discarded after every move, forcing it to rebuild the game state from scratch each turn. Two settings, nearly a 3x swing.
A July 29 arXiv study adds a third data point. Researchers mined 1.1 million Reddit posts across 29 subreddits and isolated 446 threads on security and privacy problems in Cursor, Copilot, and Codex. After reading over 6,000 comments, most reported issues traced to system-level choices, data access scope, unchecked autonomous actions, not the model underneath.
In my own work, every time output quality drops, my first instinct is to blame the model. It's almost never the model. Usually it's a tool grant revoked in a refactor, or a context budget that quietly truncated.
Before you migrate models, check what your setup can actually reach versus what your prompts assume it can reach. Next time you read a leaderboard, ask whose configuration produced the number.
Each link below shares sources, entities, or timing with this story.
Responses API built by OpenAI / Shared entities / Same source domain / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Responses API built by OpenAI); both cover Codex, Cursor, GPT, July; reported by the same outlet (arxiv.org).
Responses API built by OpenAI / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (Responses API built by OpenAI); both cover Claude Fable, Codex, GPT, July; reported by the same outlet (openai.com).
Responses API built by OpenAI / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Responses API built by OpenAI); both cover Copilot, GPT, July, Luna; overlapping topics (agent, claude, same).
OpenAI acquired Roi / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (OpenAI acquired Roi); both cover Cursor, GPT, July, Same; reported by the same outlet (arxiv.org).
Responses API built by OpenAI / Shared entities / Same source domain / Shared topic
Linked by a graph relationship (Responses API built by OpenAI); both cover Claude Fable, Codex, GPT, July; reported by the same outlet (openai.com).
OpenAI released Frontier / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (OpenAI released Frontier); both cover AGI, ARC, GPT, Same; reported by the same outlet (arxiv.org).
Responses API built by OpenAI / Shared entities / Shared topic / Earlier coverage / Downstream implication
Linked by a graph relationship (Responses API built by OpenAI); both cover Codex, GPT, Luna, OpenAI; overlapping topics (gpt 5, model).
OpenAI released Luna / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (OpenAI released Luna); both cover Cursor, GPT, July, Luna; overlapping topics (agent, model, output).