LevelUp
Catalog
Section titled “Catalog”Every material is a sequenced roadmap, not a reference dump. Each step states why it comes where it does, what you should be able to explain afterwards, and how to verify you actually got it.
Spring Boot — Event Ticketing Marketplace
Go from "I know Java, new to Spring" to a production-grade backend and fullstack app, built around one real domain: an event ticketing marketplace.
Start learningLaravel — Multi-Tenant Booking SaaS
Go from "I know PHP, new to Laravel" to a production-grade multi-tenant SaaS, built around one real domain: the appointment-booking platform a barbershop, a dental clinic and a yoga studio all run on.
Start learningAdonisJS — Last-Mile Delivery Dispatch
Go from "I know TypeScript, new to AdonisJS" to a production-grade delivery platform, built around one real domain: the last-mile dispatch system a food stall, a pharmacy and a hardware store all run their deliveries on.
Start learningGo — Webhook Delivery Platform
Go from "I can program, I have never written Go" to a production-grade Go service, built around one real system: a webhook delivery platform with retries, signing, ordering and a replay dashboard.
Start learningLaravel — Business Operations Integration Hub
Go from "I can build a Laravel app" to the integration work businesses pay a retainer for: webhooks that arrive twice, supplier files that arrive half-written, and money that has to reconcile.
Start learningFastAPI — Grounded Support Assistant
Go from "I know Python, new to FastAPI" to an AI feature you can prove works — grounded answers, real citations, refusal when the documents do not say, and an evaluation report to show for it.
Start learningLaravel — Off the Spreadsheet Back Office
Go from "I can build a Laravel app" to the job small businesses actually hire for: moving a living business off its spreadsheets — discovery, messy import, entity resolution, reconciliation and cutover — without stopping it for a day.
Start learningFastAPI — Document Intake Pipeline
Go from "I know Python" to the most boring and most sellable AI product there is: invoices and delivery notes turned into validated rows, with confidence scores, a human review queue, and an accuracy number anyone can check.
Start learningMCP — Action-Taking Operations Agent
Go from "I have built a chatbot" to an agent that takes real actions in real systems — authority limits, approval gates, idempotent writes, compensation for what cannot be undone, and a suite that proves what it refuses.
Start learningn8n — AI Operations Automation
Go from "I can write code" to the work businesses pay a retainer for: AI steps inside a real daily process, with replay, human sign-off, a confidentiality answer a solicitor accepts, and a cost dashboard the client reads.
Start learningPlaywright — Legacy System Automation
Go from "I know TypeScript" to automating the systems that have no API and never will — a resilience ladder, evidence for every run, a vision model only where selectors fail, and a vendor update that breaks your bot on purpose.
Start learningLiveKit — Voice Intake Agent
Go from "I have built a chatbot" to a phone number that answers — a latency budget, a confirmation strategy for names and numbers, a warm handoff to a person, and ten recorded hostile callers that prove how it behaves.
Start learningPlaywright — Quality Engineering for a Release You Can Defend
Go from "I write automated tests" to owning the quality of a release: a risk-based strategy, a suite proven by the defects it catches, flake treated as a defect, contract tests for the service with no UI, and a gate that blocks a build for a stated reason.
Start learning
How these materials are built
Section titled “How these materials are built”Each material follows the same shape, so what you learn about how to work through one transfers to the next.
- One real domain per material. The hard problems are load-bearing, not bolted on as demos.
- Sequenced, not alphabetised. Every step exists because the previous one created the pain it solves.
LEARNandBUILDsteps. Some things only survive being struggled with. Those steps tell your AI agent to stay a tutor and refuse to write the code.- Verifiable. Acceptance criteria and anti-pattern lists per step, so “done” is checkable rather than felt.
- Agent-ready. Each material ships the harness files —
AGENTS.md, a reviewer prompt, rubrics — so a coding agent is useful rather than merely plausible.