Intermediate developer edition | Updated: October 2026
How to use this guide
This guide covers all 14 tools mentioned in the original script, their roles, official resources, free-versus-paid considerations, a recommended learning order, hands-on exercises, and a 30-day plan. The goal is not to master every product at once. Build one working SaaS MVP while learning a repeatable AI-assisted development workflow.
The 14 tools at a glance
OpenCode
Open-source AI coding agent for terminal/desktop/IDE workflows; connects to supported model providers.
Official resource: https://opencode.ai/docs/
Learning / cost note: Use for model flexibility and agent workflows. Software can be free while model/API usage may cost money.
Claude Code
Agentic coding tool that can inspect repositories, edit files, run commands, and help test software.
Official resource: https://code.claude.com/docs
Learning / cost note: Best for repository-level engineering tasks. Access depends on eligible plan or API billing.
Claude Opus
Anthropic model family for complex reasoning, coding, architecture, and analysis; a model, not a separate IDE.
Official resource: https://docs.anthropic.com/en/docs/about-claude/models/overview
Learning / cost note: Use for difficult tasks; choose faster/cheaper models for routine edits when appropriate.
Google Antigravity
Agent-first development environment that can work across editor, terminal, and browser workflows.
Official resource: https://antigravity.google/
Learning / cost note: Experiment with browser-assisted coding on a separate branch. Check current plan and usage limits.
Cursor
AI-powered code editor with codebase context, autocomplete, agent edits, and terminal workflows.
Official resource: https://cursor.com/docs
Learning / cost note: Recommended primary editor. Free tier is limited; paid plans provide additional usage.
Grok
General-purpose AI assistant from xAI with supported research, writing, coding, and media capabilities.
Official resource: https://grok.com/
Learning / cost note: Useful for brainstorming and exploring X discussions. Verify claims against primary sources.
Perplexity
AI-powered search and research assistant that synthesizes answers with source links.
Official resource: https://www.perplexity.ai/
Learning / cost note: Use for competitor research and technical comparisons. Free access is available with limits; advanced features may be paid.
Lovable
Prompt-driven app builder for creating and iterating on web application prototypes.
Official resource: https://lovable.dev/
Learning / cost note: Good for rapid UI and product prototypes. Credits and features vary by plan.
Replit
Browser-based development environment with AI app-building and deployment workflows.
Official resource: https://replit.com/
Learning / cost note: Useful for prototypes and cloud-based development. Compute, AI, and deployment limits vary.
Hermes Agent
Agent project from Nous Research for tool use, reusable skills, and repeatable assistant workflows.
Official resource: https://github.com/NousResearch/hermes-agent
Learning / cost note: Explore after the core SaaS works; model providers and integrations may add cost.
Higgsfield AI
Creative platform for AI-generated images/videos, cinematic effects, and supported model workflows.
Official resource: https://higgsfield.ai/
Learning / cost note: Useful for product ads and social creatives; generation credits/features vary.
Google Veo
Google's generative video model family for text/image-to-video and supported audio/video workflows.
Official resource: https://deepmind.google/technologies/veo/
Learning / cost note: Access depends on product, region, account, model version, and plan.
Kling AI
Generative media platform for text/image-to-video and motion-oriented creative workflows.
Official resource: https://klingai.com/
Learning / cost note: Useful for product animation and short clips; plans, quotas, and model capabilities change.
Hugging Face
AI ecosystem for discovering models, datasets, Spaces, libraries, and inference options.
Official resource: https://huggingface.co/docs
Learning / cost note: Great for model exploration. Licenses, compute, and commercial-use permissions vary by model.
Recommended learning order for SaaS builders
1. Cursor (Days 1–7)
Make it your primary editor. Learn project instructions, codebase context, small agent tasks, diff review, terminal use, and debugging.
2. Claude Code (Days 5–12)
Use for repository-level changes, tests, refactoring, and multi-step tasks. Always inspect changes.
3. Claude Opus or another strong reasoning model (Days 2–15)
Use for architecture reviews, difficult bugs, security analysis, and trade-off comparisons.
4. Antigravity (Days 13–16)
Try browser-assisted agent workflows on a separate branch; compare its test findings with your own.
5. OpenCode (Days 17–18)
Experiment with compatible model providers and an alternative coding-agent workflow.
6. Lovable and Replit (Days 19–20)
Build small parallel prototypes and compare iteration speed, code access, and deployment constraints.
7. Perplexity (Throughout)
Research competitors, current documentation, product positioning, and technical options. Verify important claims in original sources.
8. Hugging Face (Days 27–28, optional)
Explore model cards, inference, hardware needs, licenses, and commercial-use terms only if an AI feature is justified.
9. Hermes Agent (Days 29–30, optional)
Explore repeatable agent tasks and skills in a sandbox, not with production credentials.
10. Grok, Higgsfield, Veo, Kling (Later)
Learn these when you need social research, promotional videos, product demonstrations, or creative campaigns.
Hands-on capstone project
Build a Mini Website Change Monitor
Build a small SaaS inspired by a website-monitoring product. Users register, add URLs, see the latest check result, and browse a change-history timeline. The project is deliberately limited so that you can finish a useful MVP in 30 days.
Suggested stack:
Next.js App Router + TypeScript
Tailwind CSS
Supabase Auth + Postgres + Row Level Security
Background worker or scheduled job for URL checks
Playwright for browser tests
GitHub for version control and pull requests
MVP acceptance criteria:
Landing page with clear value proposition and call to action.
Sign-up/sign-in and a protected dashboard.
Add, list, validate, and delete monitored URLs.
Background checking job with timeouts and useful failure handling.
Change history visible to the authenticated owner.
Server-side authorization and RLS policies verified with tests.
Staging deployment, smoke tests, basic SEO metadata, and error monitoring.
Documented limitations and estimated operating costs.
30-day action plan
Week 1 — Set up AI-assisted development
Day 1: Configure Cursor, Git branch workflow, and project instructions. Ask the agent to explain the repository before editing.
Day 2: Define users, core workflows, MVP scope, database entities, and acceptance criteria. Research competitors with Perplexity.
Day 3: Review the architecture with a strong reasoning model. List assumptions, security risks, and implementation phases.
Day 4: Build landing page, navigation, dashboard shell, empty states, and URL form.
Day 5: Implement a small feature with Claude Code and compare the changes with Cursor's approach.
Day 6: Connect Supabase authentication and database operations; verify ownership boundaries.
Day 7: Run linting, type checks, production build, and basic UI tests. Fix failures before adding features.
Week 2 — Build working SaaS features
Day 8: Create the monitored-URL table and indexes.
Day 9: Implement add/list/delete flows, validation, loading states, and error states.
Day 10: Add server-side authorization and Supabase RLS policies.
Day 11: Implement a background checking job with timeouts and failure handling.
Day 12: Store change history and display it in the dashboard.
Day 13: Use Claude Code to review the implementation and identify missing tests.
Day 14: Test the dashboard in a browser using Antigravity on a separate branch; compare findings with your own tests.
Week 3 — Compare tools and improve quality
Day 15: Review architecture, data isolation, SSRF risks, and job reliability with a strong reasoning model.
Day 16: Add Playwright tests for critical user flows.
Day 17: Try OpenCode with a compatible model on a bounded bug-fix task.
Day 18: Write reusable project instructions, code-review checklists, and feature prompts.
Day 19: Re-create one dashboard screen in Lovable and compare UI and code.
Day 20: Build the same small prototype in Replit; assess code access, deployment, and cost.
Day 21: Choose your main workflow; do not migrate the product solely because a prototype was faster.
Week 4 — Deploy and measure
Day 22: Improve onboarding, empty states, and calls to action.
Day 23: Add usage limits and subscription-ready design; payment integration is optional.
Day 24: Separate environments, protect secrets, and configure deployment.
Day 25: Deploy to staging and run smoke tests.
Day 26: Test security, accessibility, mobile responsiveness, and failure handling.
Day 27: Check SEO metadata, performance, analytics, and error monitoring. Explore Hugging Face only if an AI feature is useful.
Day 28: Test realistic URLs and edge cases; document known limitations.
Day 29: Explore Hermes Agent for a sandboxed test report or pull-request review.
Day 30: Deploy the MVP, document the workflow, review costs, and choose the next customer-driven feature.
Reusable prompts for coding agents
Architecture planning
Act as a senior SaaS architect. Review my Next.js App Router, TypeScript, Tailwind, and Supabase project.
Before making changes:
1. Explain the current architecture.
2. Identify missing requirements and security risks.
3. Propose a minimal production-ready architecture.
4. Break implementation into small, testable phases.
5. Identify assumptions and trade-offs.
Do not modify files. Wait for my approval.
Feature implementation
Implement the “Add monitored URL” feature.
Requirements:
- Follow existing project conventions.
- Validate URLs on the server.
- Verify the authenticated user's identity.
- Enforce ownership and database access policies.
- Include loading, success, and error states.
- Add tests for invalid input and unauthorized access.
- Avoid unnecessary dependencies.
First explain your plan. Then implement, run relevant checks, and summarize files changed and remaining risks.
Production readiness review
Act as an independent production-readiness reviewer. Inspect changes for:
- Authentication and authorization bypasses.
- Supabase RLS policy mistakes.
- SSRF risks in URL fetching and redirects.
- Secrets exposed to the browser.
- Missing validation and error handling.
- Database performance issues.
- Missing tests and deployment risks.
Do not claim tests passed unless you actually ran them. Rank findings by severity and provide reproducible verification steps. Do not change files until I approve the findings.
Security note for website monitoring
A server fetching user-supplied URLs must defend against Server-Side Request Forgery (SSRF). Block localhost, private network ranges, cloud metadata endpoints, and unsafe redirect destinations. Revalidate each redirect and resolve DNS carefully to reduce DNS-rebinding risks. Use network-level egress controls where possible. URL syntax validation alone is not sufficient.
Free vs. paid strategy
Start with free tiers and pay for one primary coding workflow only when usage limits slow delivery. Open-source software may still require paid model/API usage. Check official pricing pages immediately before purchase because plans, quotas, and regional prices change.
Cursor pricing: https://cursor.com/pricing
Antigravity: https://antigravity.google/
OpenCode providers: https://opencode.ai/docs/providers/
Claude Code setup: https://code.claude.com/docs
Lovable: https://lovable.dev/
Replit pricing: https://replit.com/pricing
Supabase pricing: https://supabase.com/pricing
Core technical resources
Next.js App Router: https://nextjs.org/docs/app/getting-started
Supabase documentation: https://supabase.com/docs
Playwright: https://playwright.dev/docs/intro
GitHub documentation: https://docs.github.com/
Hugging Face documentation: https://huggingface.co/docs
Day-30 success checklist
☐ Configured one primary coding-agent workflow.
☐ Built and deployed a working SaaS MVP.
☐ Verified authentication, ownership, and database access policies.
☐ Added automated tests for important user journeys.
☐ Completed a production-readiness and SSRF review.
☐ Compared at least one alternative coding agent and one app builder.
☐ Documented model/tool costs, workflow instructions, and next product priorities.
Final advice
Do not measure progress by the number of AI tools you have tried. Measure it by whether you can turn a clear requirement into secure, tested, maintainable, deployed software. Start with Cursor, Claude Code, a strong reasoning model, Next.js, Supabase, and Playwright. Treat the other tools as focused experiments.
Want the members-only resources too?
A free account unlocks them all, including the ones shared in DMs.