Detailed Analysis
A quirky piece of Claude ephemera has surfaced in the form of a crowdsourced catalog of 187 "spinner verbs" — the whimsical progress-indicator words Claude displays while processing a request, such as "Flibbertigibbeting," "Recombobulating," "Honking," and "Symbioting." The article references a compiled list at deepakness.com and describes an effort to turn that raw inventory into a "pictorial reference," essentially an illustrated glossary that lets users glance at Claude's loading screen and get a visual sense of what the assistant is nominally "doing" while it thinks. The tone is playful and affectionate, treating the spinner text as a small cultural artifact worth documenting rather than a technical detail worth explaining.
These spinner verbs are a deliberate design choice by Anthropic, not a bug or randomness for its own sake. Loading states are a universal UX challenge: users need reassurance that a system is working during latency, and generic spinners or "Loading..." text can feel sterile or induce impatience. By rotating through a large, invented lexicon of nonsense-adjacent gerunds — many blending real words with invented ones (Flibbertigibbeting, Whatchamacalliting) or repurposing existing words in surreal ways (Newspapering, Honking) — Anthropic injects personality and levity into an otherwise mundane wait state. This mirrors a broader pattern in consumer software of using microcopy to humanize AI products: Slack's whimsical loading messages, Mailchimp's playful error text, and various apps' quirky empty-states all serve the same function of softening technical friction with brand voice.
The fact that this list has grown to 187 distinct terms — enough to inspire a dedicated cataloging project and now a visual companion piece — speaks to how much attention users pay to the small, seemingly incidental details of AI products. It also reflects a growing internet subculture of reverse-engineering and documenting the idiosyncrasies of large AI systems, from prompt leaks to system card details to, in this case, purely cosmetic flourishes. That such minutiae generate dedicated fan projects suggests Claude has cultivated a userbase engaged enough to treat interface details as worth preserving and explaining, not unlike fandoms that catalog Easter eggs in video games or hidden jokes in software release notes.
More broadly, this kind of artifact underscores how AI assistants are increasingly evaluated not just on capability benchmarks but on the texture of everyday interaction — tone, personality, and moments of delight during otherwise invisible computation. As foundation model providers compete on raw performance metrics that are increasingly difficult for average users to distinguish, these smaller design signals (word choice, humor, warmth) become meaningful differentiators in how a brand is perceived. A community-made illustrated glossary of loading-screen verbiage is a small but telling data point in that larger story: users don't just want smarter AI, they respond to AI that feels considered and characterful even in its most mundane moments.
Read original article →