Forget AGI—Sam Altman celebrates ChatGPT finally following em-dash formatting rules (arstechnica.com)

OpenAI CEO Sam Altman posted on X that ChatGPT finally respects a user’s custom instruction to avoid using em-dashes, a small but notable change that arrived just after the release of GPT-5.1. Em-dashes have become a common giveaway of AI-generated prose — frequently appearing in chatbot outputs — so the update addresses a long-standing formatting gripe from users who have struggled to get ChatGPT to consistently follow stylistic preferences. The moment is telling for the AI community: it shows incremental progress in instruction‑following and user-level controllability (likely tied to updates in the new model), but also underlines persistent limits. If a state-of-the-art model struggled for years to obey a basic punctuation rule, that raises questions about how well we understand model behavior and alignment more broadly. For practitioners, the episode highlights the practical importance of robust preference handling, better instruction parsing, and fewer brittle heuristics in production models — and tempers rhetoric about near-term AGI by pointing to remaining engineering gaps. Users and developers should expect continued, pragmatic improvements rather than sudden leaps to “general” intelligence.