Desktop preview

The desktop where Ari plans, routes, builds, remembers, and hands off.

This is the customer-facing desktop story: Ari receives an ambitious request, classifies the work, picks fit-for-purpose models, controls cost, creates a plan, builds locally, and preserves the Rhumb trail.

Meridian Desktop - Ari
Active workspace YAKKL Ari build

Cost priority: 0

You

Build me a production-grade website with a cinematic hero, consumer and technical Ari product pages, an edge backend, SQL storage, docs, GitHub repo, model routing, and cost guardrails.

Ari

This request spans product surfaces, backend, API docs, model routing, storage, deployment, and cost guardrails. Ari turns it into a plan, routes the phases, and builds from there.

You

Proceed.

Ari

Three phases are in motion. The workflow records the plan, routes the phases, preserves the artifacts, and keeps the audit path available when the stakes justify it.

Representative prompt

The prompt should sound ambitious, messy, and normal.

The point is to show Meridian extracting structure, cost boundaries, deployment assumptions, and missing decisions from a typical ambitious ask.

Build a production-grade website with a strong homepage and two product pages.

Include a REST API spec, example data, edge deployment, SQL storage, docs, and cost guardrails.

Separate the consumer story from the technical buyer story.

Let the system pick the right model for each phase.

Assume fewer than 100k users by month six. Cost control matters immediately.

Structured execution

Small plan when it is enough. Full architecture path when the stakes demand it.

The desktop should make the execution path obvious instead of forcing every request through the same amount of ceremony.

P-01

Product shell

Homepage, consumer page, technical page, brand tone, local preview.

P-02

Backend edge

REST routes, SQL schema, example records, cost guardrails, docs.

P-03

Review and ship

RWP-compatible plan saved, local run shown, optional audit deferred to save cost.

Positioning rule

Sell better work per dollar, not more tokens.

Ari protects cost by controlling model fit, infrastructure choices, artifact reuse, and when heavier audit paths are warranted.

Cost Use small routing and cheap infrastructure first.
Quality Escalate model effort only where it changes the result.
Continuity Save prompts, responses, artifacts, and deployment decisions.
Open protocol Keep the plan RWP-verifiable when the workflow needs proof.