25: Setting Up AI - transcript
Liam: The QUILL Cast, episode twenty-five, and welcome to part five: artificial intelligence, QUILL style. I'm Liam. Before any feature, today is about the foundations: setup, the free paths, and the rules that make AI trustworthy here.
Jessica: I'm Jessica. The three rules first, they govern everything in this part. Rule one: AI is off until you turn it on. Fresh QUILL has an AI menu with essentially one item, Set Up AI, and nothing happens without you. Rule two: you choose who, if anyone, sees your words. Rule three, the big one: AI never changes your document silently. Every edit arrives as a preview you accept or reject, and an accepted change is one undo step.
Liam: Rule three deserves a spotlight because it's the whole safety story in one sentence: the AI proposes, you decide, and control Z is always loaded. Everything else is details.
Jessica: Now the wizard: AI menu, Set Up AI. It asks one real question, how do you want this to work, and presents the options with their trade-offs stated plainly, no marketing gloss.
Liam: Option one, most private: a model running on your own computer, via Ollama. No account, no cost, works offline, and nothing you write ever leaves the machine. The wizard helps pick a model sized to your hardware. This is the option for confidential work, and QUILL says so explicitly.
Jessica: Option two, best free quality: bring your own free key. The wizard leads with OpenRouter's free tier, a Get API Key button opens the exact signup page, and QUILL preselects a strong free writing model. Honesty included: free hosted models can be slower, rate limits apply to your own quota, and the wizard literally reminds you to keep sensitive writing on the local option.
Liam: Option three: paid accounts, OpenAI, Anthropic, Google Gemini, OpenRouter, if you have them, with a model dropdown that can list everything your account offers, no typing model names from memory. And option four, a full citizen of the menu: Not Now. Close the wizard, QUILL remains a complete editor forever. Some of the best QUILL users never set up AI at all, and the product treats that as a valid life.
Jessica: Say it clearly for the back row: you never need to pay to use QUILL's AI. Local is free. The leading hosted path is free. This was a deliberate product decision with a name, free AI for everyone, and it rhymes with QUILL itself being free. Assistive capability gated by subscription is the old world.
Liam: The privacy architecture underneath, briefly, because part seven does it justice. Every network call in the entire application is registered in an internal egress audit with a written justification, and automated gates fail the build if anyone adds an unregistered one. Consent is per-action and specific, QUILL says what's being sent and to whom, in words. Keys live in encrypted credential storage, never plain config. And Safe Mode, from episode two, kills all of it with one flag when you want a guaranteed quiet machine.
Jessica: Expectation-setting, the honest coda. A free local model is not a frontier datacenter model. QUILL adapts, simpler task strategies on smaller models so things answer instead of stalling, but adaptation isn't magic. The everyday writing features you'll meet next episode work well on free models; the most ambitious agent plans benefit from stronger ones. Set expectations, then be pleasantly surprised, that's the house tone.
Liam: Homework. One: run the wizard and pick a path, including Not Now if that's your truth, just hear how choices are presented. Two: if you connected something, press Test Connection and hear the confirmation. Three: find the AI privacy documentation from the Help menu and skim what gets sent when, before you ever send it.
Jessica: Next episode: the payoff, Ask Quill and the everyday writing actions, with the review-everything contract in action.
Liam: I'm Liam.
Jessica: I'm Jessica. Your words, your rules, your call.