Step-by-step tutorials to get the most out of ALIN.
This guide walks through creating and executing a complete TBWO from start to finish.
Step 1: Open the Chat Station and type a complex request, like "Build me a portfolio website with an about page, projects gallery, contact form, and dark theme." ALIN's intent detector will recognize this as a sprint-worthy task.
Step 2: ALIN generates an execution plan showing phases (Setup → Design → Content → Polish → QA), estimated time per phase, pod assignments, and quality targets. Review the plan. You can adjust time budgets, add specific requirements, or remove phases you don't need.
Step 3: Click "Approve Plan" to start execution. The TBWO Command dashboard shows real-time progress — active pods, current phase, completed tasks, and a progress timeline.
Step 4: At each checkpoint, ALIN pauses and shows what's been completed. You can preview files, add instructions ("make the hero section taller"), or approve to continue.
Step 5: When complete, files appear in the chat as attachments. A receipt documents everything. The self-model records the outcome for future improvement.
Step 1: Open Settings → Memory. Enable all 8 layers (Pro plan required for layers 3-7).
Step 2: In the Chat Station, tell ALIN about your project: "We're building a SaaS dashboard using Next.js 14, TypeScript, Prisma, and PostgreSQL. The design system uses Tailwind with a custom theme." ALIN stores this in semantic memory (Layer 4).
Step 3: Set preferences: "I prefer functional components, named exports, and comprehensive error handling." These go to relational memory (Layer 5).
Step 4: Pin important notes: "The staging server is at staging.example.com. The API base URL is /api/v2." These persist in explicit notes (Layer 6).
Step 5: As you work, ALIN automatically consolidates patterns. After a few sessions, check the Memory Dashboard to see what ALIN has learned — you can edit, delete, or promote any memory entry.
Direct Mode is ALIN's fast path — no TBWO, no plan, just tool-assisted execution. It's how 80% of interactions should work.
When to use Direct Mode: Bug fixes, code changes, file operations, quick research, explanations, refactoring, running commands, and any task that doesn't need multi-phase orchestration.
How it works: You type a request. ALIN reads files, makes edits, runs tests, and verifies — all in a tight loop without asking permission for each step. At the end, it summarizes what changed.
Tips: Be specific. "Fix the authentication bug in server.js where the JWT token isn't being validated on the /api/settings endpoint" is better than "fix the auth bug." Direct Mode chains tool calls rapidly — the more context you give upfront, the fewer wasted steps.
Override: If you want to force Sprint Mode for a task the intent detector classified as direct, add "use sprint mode" to your message. Conversely, "use direct mode" forces direct execution even for complex-sounding requests.
ALIN supports multiple AI models simultaneously. Here's how to use them effectively:
Claude Opus 4: Most capable. Use for complex reasoning, nuanced writing, and critical decisions. Slowest and most expensive.
Claude Sonnet 4.5: Excellent balance. Good default for most tasks. Best for coding, structured output, and tool use.
Claude Haiku 4.5: Fastest. Use for simple tasks, quick lookups, and high-volume operations where speed matters more than depth.
GPT-4o: Strong alternative. Good for creative writing and tasks where a different perspective is valuable.
In Settings → Models, you can set a default model and enable auto-routing. With auto-routing, ALIN picks the best model per task: fast model for clarification questions, strong model for planning, coding model for code generation.