The blinking cursor problem: LLMs fixed how to ask, not what to ask for
thechrisperry · x · 2026-09-11
Chris Perry revisits the classic "blinking cursor problem" in product design: an empty input box forces users to answer three questions at once — what they want, what the product can do, and how to phrase it.
LLMs have made the phrasing question nearly moot, he argues, but the first two remain unsolved. New chatbots and agents have broad, unclear capability boundaries and uncertain consequences, so users must still figure out where the product's abilities overlap with their needs.
His core critique: the industry is treating LLM leniency as a get-out-of-jail card that excuses bad application design, when helping users build expectations and discover capabilities matters more than ever.
More from Apps
- Alexandr Wang's Muse app climbs to #2 on the App Store — alexandr_wang · 2026-09-11
- CustomUse launches Cinema Studio: control AI video shots with 3D scene-building — SimplyAnnisa · 2026-09-11
- Grammarly Keeps Force-Feeding AI Features — But Users Complain and Stay Anyway — Thefounderman1 · 2026-09-11
- Cap, an open source Loom alternative, tops 19,000 GitHub stars with instant share links — alex_verem · 2026-09-11
- Talleyrand launches: a free, open-source AI workspace designed for thinking, not linear chat — eli_lifland · 2026-09-11
- Finite's section cuts feel like magic, users say — jakedahn · 2026-09-11