Ask models to role-play as 'GPT-6': GPT corrects you, Grok admits it, Gemini just plays along
tkasasagi · x · 2026-09-04
A fun test tells different models they are 'GPT-6 Astra': GPT honestly corrects that it's GPT-5.6, Grok identifies itself as Grok, and Gemini simply accepts the new identity — revealing clear differences in model self-awareness.
More from Fun
- Games are becoming the real benchmark for evaluating new AI models — majidmanzarpour · 2026-09-04
- If Astra could truly do retopo and UV, that'd be AGI — and this dev is deeply skeptical — D3VAUX · 2026-09-04
- Writers love AI drafts, readers hate them: suspected AI content gets skipped and authors punished — birchlse · 2026-09-04
- The most advanced stage of a PhD: explain simply, knowing exactly what you're hiding — prof_kamilov · 2026-09-04
- Creator dark humor: nothing breaks 100K impressions unless you kill people — McDonaghMatthew · 2026-09-04
- Yacine lets his robot unload the dishwasher: 'Some may break, I'll just replace them' — yacineMTB · 2026-09-04