You can now interrupt ChatGPT as it learns to take feedback on the fly (www.techradar.com)

OpenAI added a real‑time “interrupt” capability to ChatGPT that lets users halt a model mid‑response and inject clarifications or corrections via an “Update” button in the sidebar. The feature is aimed especially at higher‑latency, higher‑cost use cases — GPT‑5 Pro and ChatGPT’s Deep Research mode — where answers take minutes and users may have limited prompt quotas. Instead of stopping and restarting a long query, you can add new context (for example “use 2024 data, not 2022”) while the model is still composing and watch the answer adjust immediately. For researchers and analysts this reduces wasted time, tokens and reruns on multi‑step tasks: long explorations can be steered on the fly, wrong premises corrected early, and iterative reasoning made more fluid. Technically it shifts the UX from strict turn‑taking to a continuous, overlapping feedback loop where the model ingests mid‑generation edits and revises its output — a practical move toward more interactive steering and better alignment, though not a change in model consciousness. The update signals broader UI/UX trends for LLM tooling (real‑time editing, tighter human-in-the-loop control) and should improve productivity and fidelity in deep, exploratory AI workflows.