{"slug": "show-hn-uipalooza-do-ai-models-have-a-recognizable-design-taste", "title": "Show HN: UIPalooza – Do AI models have a recognizable design taste?", "summary": "UI Palooza, a Show HN project, publishes a gallery that gives different AI models the same versioned design brief and places their resulting interfaces side by side to test whether models show a recognizable design taste. Each run starts in an isolated workspace, with Pi using Lunaroute and other rows using Claude Code or Codex, under a default generation budget of 30 minutes, 120 model responses and high reasoning. The project separates browser interaction and accessibility checks, which measure functionality, from human design judgments, and states that its two baselines are exploratory evidence rather than a statistically reliable ranking.", "body_md": "Rows are models and columns group selected samples by skill\nvariant. Click a preview to open the working artifact.\n\nNo models match these filters.\nReset filters\n\nEvidence, and your eye.\n\nLatest attempts for the selected brief, samples, and skills.\nFunctional checks and human design judgments remain separate.\n\nDifferent schools of thought.\n\nA curated selection based on adoption, relevance, and distinct\napproaches to interface design. “Top five” is a selection, not a\nproven quality ranking.\n\nA fairer starting point.\n\nHold the task steady. Change the model or the skill. Keep the\nevidence.\n\n01\n\nSame task, fresh workspace\n\nEvery sample starts from its versioned brief, local content and\nassets. The live briefs use browser state; the original backend\nForma experiment is preserved in an offline archive. Baselines are\nindependent; Alternative direction receives recorded descriptions\nof the two baselines. Whacky starts fresh with an\nexperimental-design prompt.\n\n02\n\nIsolated coding agents\n\nEach run starts in an isolated workspace. Pi uses Lunaroute; other rows use Claude Code or Codex. The model, coding agent, skill, and sample condition are shown with each result.\n\n03\n\nVersioned ingredients\n\nSkill commits, supporting references, model catalog, task and\nasset hashes, coding-agent versions, prompts, and transcripts are\nsaved. Reruns are new experiments; model generation is not\ndeterministic.\n\n04\n\nChecks, not invented scores\n\nEach brief has browser interaction and accessibility checks. These measure functionality, not design taste. Use desktop and mobile previews to make your own comparison.\n\nForma compares each skill independently with vanilla. The frontend\nsuite includes Baseline A and Baseline B as independent attempts,\nplus Alternative direction as a deliberately different design\nprompted by descriptions of both baselines. Whacky asks for an\nexperimental interface without showing earlier outputs, initially in\nLunaroute vanilla. These conditions test range, not unprompted\ndefaults. Two baselines are exploratory evidence, not a\nstatistically reliable ranking. Provider aliases and reasoning\nsettings may change upstream; manifests record requested IDs and\nsettings. GLM 5.2 maps explicitly to\nglm-5.2-vision.\n\nNative-agent rows compare both the model and its coding agent.\nSystem prompts, tools, and reasoning budgets differ. Codex has no\nequivalent per-response output-token cap; manifests record this\ndifference and each agent's response-counting method.\n\nThe public gallery contains selected static artifacts and sanitized measurements. Interactive designs run on a separate origin. Original records and private experiments remain in the research workspace.\n\nDefault generation budget:\n30 minutes, 120 model responses, high reasoning; agent-specific\noutput limits are recorded per run. Monetary costs are not compared across providers and subscription\nplans.\n\nA test of taste.\n\nGive different AI models the same brief. What do they choose for the type, the layout, the way an interface works?\n\nUI Palooza puts those choices side by side. Compare models, add a design skill, or look across attempts to see what changes and what keeps showing up.", "url": "https://wpnews.pro/news/show-hn-uipalooza-do-ai-models-have-a-recognizable-design-taste", "canonical_source": "https://uipalooza.ai/", "published_at": "2026-10-26 04:00:04+00:00", "updated_at": "2026-10-26 04:15:48.582890+00:00", "lang": "en", "topics": ["ai-research", "ai-products", "large-language-models", "ai-agents"], "entities": ["UI Palooza", "Claude Code", "Codex", "Lunaroute", "Pi", "GLM 5.2", "Forma"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/show-hn-uipalooza-do-ai-models-have-a-recognizable-design-taste", "markdown": "https://wpnews.pro/news/show-hn-uipalooza-do-ai-models-have-a-recognizable-design-taste.md", "text": "https://wpnews.pro/news/show-hn-uipalooza-do-ai-models-have-a-recognizable-design-taste.txt", "jsonld": "https://wpnews.pro/news/show-hn-uipalooza-do-ai-models-have-a-recognizable-design-taste.jsonld"}}