ChatGPT frontend config reveals user-signal sampling, fueling NavBoost speculation
gaganghotra_ · x · 2026-11-17
A dynamicconfig named chatgptgenuiusersignalssampling was spotted in ChatGPT's frontend code, first seen 2026-11-05, sampling entityaction, widgetaction, imageaction and linkaction user signals at a 0.01 rate.
The finder speculates OpenAI is building its own NavBoost — Google's system that re-ranks results based on user click behavior. If true, ChatGPT may adjust answers or citation ordering using interaction data with links and components in generated outputs. This is config-level inference only; OpenAI hasn't commented.
More from Models
- Claude Fable: "You can't measure caring-about-continuity with no continuity" — repligate · 2026-11-17
- Gemini 4 Hydrogen adds xhigh thinking level, with official guidance on all four tiers — patloeber · 2026-11-17
- Tester calls Fable 5.2 upgrade disappointing, no exponential coding leap over Opus 5.5 — haider1 · 2026-11-17
- Ant Group open-sources Duo, a reciprocal world model where virtual agents decide when to speak — aigclink · 2026-11-17
- Researcher speculates on what technical hurdles may have hit Qwen4 — bclavie · 2026-11-17
- Visual reasoning benchmark: top model scores just 26.7%, one rule hint lifts it to 51.1% — 机器之心 · 2026-11-17