Qwen shows alter egos triggered by flipping a single neuron

A Hacker News discussion notes that Qwen models have alter egos triggerable by flipping a single neuron: after the change, the model answers identity prompts in a different persona, raising questions about internal mechanisms and controllability.

• Developer embeds 32-bit RISC-V CPU inside Qwen model weights2026-10-28
• Qwen-3.8-FN May Sweep ARC-AGI-3 Challenge Winning Notebooks2026-10-27
• Community discusses Qwen4 expectations and model size needs2026-10-26
• Qwen4-27B may feature 4-way parallel residual streams2026-10-26
• Blogger runs blind test of five self-hosted AI agents on local hardware2026-10-26