Quiddy was built on a simple conviction: the most powerful AI interaction isn't the one from the most capable model — it's the one where the question was asked with the right context.
LLMs are stateless by design — every conversation starts from zero. You type a question; it answers generically, without knowing your role, your tools, your current project, or how you like to think. The quality of the answer is bounded entirely by the quality of the question. And most questions are shallow because they carry no context.
"The bottleneck isn't the model. It's the question. Give the model enough context and even a modest model gives you exceptional answers."
Quiddy's answer is the SEEK model: a structured way to define not just what you want the AI to do, but how it thinks, how it communicates, what domain it operates in, and how intensely it applies those traits. Combined with persistent memory, a knowledge vault, and fine-grained behavioral controls, Quiddy becomes an extension of the way you work — not a generic chatbot you tolerate.
Your conversations contain your best thinking — ideas you're still developing, code you haven't committed, business strategies you haven't announced. That data shouldn't live on someone else's server, logged, analyzed, and used to improve a model you didn't consent to train.
Quiddy runs a native Swift server on your Mac. Your data lives in SQLite on your disk. Your API keys are in the macOS Keychain. The only external communication is the LLM API call you deliberately make. We can't read your conversations because we never receive them.
Not a checkbox. Local-first isn't a limitation — it's the point. Your data is yours, always.
A focused, well-tuned persona outperforms a generic model with 10× the parameters. Context is everything.
Every parameter exposed in Quiddy has a clear, objective effect. No vague sliders. No mystery settings.
A tool you have to configure for an hour before it's useful isn't a tool — it's a project. Quiddy works in minutes.
Memory extraction, persona refinement, and knowledge accumulation make Quiddy better the longer you use it.
Claude, GPT-4o, Gemini, or your own Ollama instance. Quiddy doesn't lock you into any single provider.
SEEK is Quiddy's four-dimensional system for defining how an AI persona thinks and communicates. Each dimension is independently adjustable with three intensity levels, and supports a free-form custom option.
Controls the affect and interpersonal style of responses. Objective gives you neutral, fact-first communication. Empathetic adds warmth and acknowledgment. Witty brings sharp, clever observations. Energetic is motivational and upbeat.
Combine with intensity levels: a Slightly Witty engineer persona adds personality without sacrificing precision; a Very Witty copywriter persona is distinctly characterful.
Defines how the AI approaches problems. Analytical breaks queries into components with structured comparisons. Practical goes straight to step-by-step directives. Creative uses lateral thinking and unexpected connections. Explanatory simplifies with analogies. Narrative thinks in stories and context.
Controls how responses are structured and how long they run. Concise targets under 250 words, cutting to the point immediately. Standard aims for 250–600 words. Structured enforces Markdown headers and lists. Flowing produces smooth narrative paragraphs. Persuasive uses rhetorical devices and calls-to-action.
Anchors the persona in a specific domain. Engineer produces technically rigorous, code-centric responses. Educator guides with questions and encourages critical thinking. Strategist focuses on planning and trade-offs. Scientist is evidence-based and methodical. The intensity levels here — no prefix, Professional, Expert — adjust how deeply the role is embodied.
Each built from a specific SEEK configuration with matching conversation starters.