FAQ
Frequently asked questions
Clear answers about wallet credit, usage, subscriptions, and how Tycoon charges for work.
Can an AI frontend engineer actually match a senior human?
On well-scoped tickets against an established codebase, yes. What AI does exceptionally well: converting mockups to components, writing tests, fixing lint and type errors, responding to clear review comments, refactoring for consistency. What it does less well: novel architecture decisions, UX judgment calls (when the mockup is wrong), and debugging race conditions that require deep mental models. The practical outcome for most founders: the AI handles 70-80% of frontend work without supervision, your human time goes into the 20% that's novel or ambiguous.
Which frameworks does it know?
First class: React, Next.js (App Router and Pages Router), Vue 3, Nuxt 3, Svelte, SvelteKit, Solid. Component libraries: shadcn/ui, Radix, Headless UI, Chakra, MUI, Ant Design. Styling: Tailwind, CSS Modules, Emotion, Styled Components, vanilla CSS with design tokens. State management: Zustand, Redux Toolkit, Jotai, Pinia, TanStack Query. Testing: Vitest, Jest, Playwright, Cypress. If you work in a framework not on this list, the AI can work in it but with somewhat slower iteration while it learns your codebase conventions.
How does it handle code review feedback?
Every review comment becomes a discrete task. The AI Frontend Engineer reads the comment, makes the change in a fresh commit (so history is clean), and re-requests review. It explicitly avoids batching unrelated comments into one commit. For comments that require a product decision ('should this be a dropdown or a dialog?'), it surfaces the question to you with a recommendation and waits. Most reviewers — AI or human — find the iteration loop faster than working with a mid-level human engineer, because there's no ego in the response.
What about TypeScript strictness and type safety?
TypeScript strict mode is the default. The AI Frontend Engineer writes narrow types, avoids `any`, and uses branded types for IDs where appropriate. Generic components get full type parameters. API responses use zod or valibot for runtime validation when the backend doesn't already provide types. PRs that introduce `any` require an explicit comment explaining why, and the CTO reviews them. This level of discipline is often easier to enforce with an AI than a human team because the AI doesn't push back under deadline pressure.
How does it work with my existing team?
As a peer, not a replacement. Most founders running one or two human engineers find the AI Frontend Engineer slots in as the 'volume' member — takes the tickets nobody wants, handles refactors, writes tests, does accessibility passes, migrates old code. The humans focus on novel work, architecture, and the moments where judgment matters more than speed. Team velocity typically doubles in the first month because the ratio of 'interesting work' to 'maintenance work' per human engineer improves dramatically.