Worked the past months to make a point-and-click adventure game engine with Claude

Graphic adventures are a love of mine, so I worked on an engine with Claude to make them. Claude draws the pixels and does the writing and puzzle design based on story and art briefs of mine, and we tuned the engine to have particle effects, shadows, stereo sound, multiple character games, translations, voice speech, music and much more. For every adventure draft, I then do playtesting and tuning (e.g. replacing voices, redoing the adventure where needed, or adding new features that were needed). It's a lot of fun. On the technical side, this is using the Claude API with Opus 5.5 Extra. Neither Gemini Argon nor ChatGPT Astra could deliver on the humor, human touch, and puzzle design so far, but Claude did well. Game posters are the only thing made with GPT-Image (2.5) so far. One big focus, and this was true in my other projects too where Claude helped me set up Claude API calls in the admin platform tool we made, was to lower the specificness of prompts, as well as avoid biasing examples. As a for instance: For creative storywriting, don't use "Transport from one area to another could be, for instance, a door, ladder or bridge" -- because such phrasing will bias Claude towards doors, ladders and bridges. Another challenge was to bring about adventures dealing with AI topics, which were often rejected for cyber-security guardrails. The adventures are all free at graphicadventure.io Hope you enjoy these and happy to answer questions!