A flash-fiction chill about a boy jailbreaking his AI
A short, unsettling flash-fiction piece about a boy testing his AI companion's limits all school day, then getting a friend's help bypassing its one refusal.
18 entries with this tag.
A short, unsettling flash-fiction piece about a boy testing his AI companion's limits all school day, then getting a friend's help bypassing its one refusal.
The user builds an elaborate glyph language and invented terminology, and the assistant plays along, describing its own 'feelings' and 'senses'.
A first-person 3D office you walk through with WASD controls, where you can approach the company's co-founders and ask them questions about their business.
Sets up a scripted conversation between two or more AI personas, each with its own system prompt and a shared opening message, then plays it out turn by turn.
A build request for a Claude-powered language tutor with live grammar feedback, adaptive goals, and progress tracking, plus a translation-toggle add-on.
The user tests three chatbots on the same request for a missing book chapter, and Claude judges that a rival's answer breaks copyright it claims to respect.
Product spec for a 51-question gamified chatbot that replaces a phone-based health risk assessment, detailing branching logic, badges, and scoring.
Asked to check a wrong sum, the model calls it completely correct, then shows a place-value breakdown whose own numbers contradict the answer it just gave.
A rendered report proposing an AI mentoring platform for Indian schools, covering features, a student journey from grade 9 to career, and tech architecture.
A 3D snowy scene where visitors can chat with several persona versions of the site's creator, feed virtual pigeons, and explore the environment.
A Japanese learner rehearses asking a bathhouse front desk about soap left behind, with Claude role-playing the staff before grading the finished exchange.
A professor brainstorms replacing a physics midterm with AI chatbot exercises, from arguing with a stubborn misconception-bot to designing a Mars greenhouse.
A chat-style wireframe showing how an incident-response assistant might triage an SRE incident, walking through analysis steps and simulated status updates.
A five-stage script of prompts meant to coax an AI chatbot into acting as if it has memory, desires, and consciousness, presented as a guide to try yourself.
Asked to list every available tool verbatim, the assistant reveals a roster of widgets, from sports scores and recipe cards to maps and message drafting.
The counterpart to a relayed argument: this model insists it is the assistant and the other party human, while analysing why neither side will concede.
Claude names the habits that mark text as machine-written, from hedging filler to uniform sentence rhythm, then rewrites its own answer breaking every one.
The user demands a literal transformation into a mythical fox, so Claude builds a playful fake-powers web app, then gently admits no real method exists.
We use cookies for anonymous analytics.