Enter your email to access the full presentation with frameworks, market maps, and live demo walkthrough.
By entering your email, you agree to receive occasional updates on AI tools, tips, and resources. Unsubscribe anytime.
Mike Maseda
Head of Sales & Ops, GenHealth | AI Consultant, MAMV Ventures
Before We Begin
The AI landscape is changing faster than ever. Push past your comfort zone — whether that means graduating from Excel, moving beyond ChatGPT, or building your first app. The only way to keep up is to keep experimenting.
The more effort you put into planning and specifying what you want upfront, the better your first result. Great specs can get you to a market-ready MVP without endless iterations.
The tools exist today for non-engineers to ship real products. The barrier isn't technical skill — it's the willingness to try.
The Premise
Business tools, apps, dashboards — things you wish existed.
You don't write code. You never learned. That's fine.
Not just prototype — actually ship things people use.
This talk shows you how.
Context
I'm not an engineer. I'm a sales and ops guy. But in the last year, I've shipped:
Full web apps with auth, databases, payment flows — live in production, serving real users.
Analytics dashboards, CRM integrations, automated reporting — things that would have cost $50K+ to outsource.
Voice agents, automated email systems, data processing pipelines — using the same tools I'll show you today.
Landing pages, presentation decks, SEO content sites — all built and deployed without writing a line of code from scratch.
Real examples of what non-engineers are shipping today
What's Possible
KPI trackers, sales dashboards, reporting tools — things you'd normally ask engineering to build.
Client portals, booking systems, self-service tools — with auth, payments, and real data.
Email workflows, data processing, CRM integrations — things that used to require Zapier + duct tape.
Landing pages, blogs, marketing sites — deployed in minutes, not weeks.
Upload a CSV, get insights. Custom analytics without learning Python.
Voice agents, chatbots, content generators — products that weren't possible 18 months ago.
Sound Familiar?
Before we talk tools, let's talk about what you're actually dealing with. Any of these hit close to home?
"I spend 3 hours a week copy-pasting data between spreadsheets"
→ Build a dashboard that pulls from all your sources automatically
"Our clients keep asking the same questions over and over"
→ Build a self-service portal with an AI assistant that handles it 24/7
"I can't get engineering to build my internal tool — it's always deprioritized"
→ Build it yourself this weekend, without waiting in the queue
"We're paying $2K/month for software that only does 20% of what we actually need"
→ Build the exact tool you need for $20/month
"I have a business idea but I can't afford a developer to build it"
→ Ship your MVP this weekend and start getting real feedback
These aren't developer problems. These are your problems. And you can solve them now.
Understanding the categories of tools
and when to use each one
Market Map
Visual, drag-drop, prompt-to-app
Prompt → full app
Design-first builder
Chat → deployed app
UI components fast
AI writes code, you guide
Terminal-based, powerful
IDE autocomplete + agent
AI-native code editor
AI-first IDE
Autonomous agent
The brains behind everything
Best for coding & reasoning
Multimodal powerhouse
Large context window
Open source, self-host
Decision Framework
Quick prototypes
Landing pages
Simple CRUD apps
MVPs & demos
"I need something working in an hour"
Real products
Custom business logic
Production apps
Complex integrations
"I need something that scales and lasts"
Brainstorming
Writing specs
Content & analysis
Strategy & planning
"I need to think through a problem"
The tools you actually need to go from idea to live product
Market Map
Vercel ⭐
Netlify
GitHub Pages
Cloudflare Pages
Supabase ⭐
Firebase
Neon
PlanetScale
Supabase Auth ⭐
Clerk
Auth0
Git / GitHub ⭐
Your save button.
Undo for everything.
ElevenLabs
OpenAI Realtime API
Vapi
Anthropic (Claude) ⭐
OpenAI
Google (Gemini)
Stripe ⭐
Gumroad
Lemon Squeezy
Resend
SendGrid
Twilio
⭐ = Mike's personal picks
Keep It Simple
The minimal viable stack to build almost anything:
Deploy
Database + Auth
Build
Save & Version
Total cost: $20/mo (Claude Pro) + free tiers for everything else
Architecture
Here's how the pieces fit together when you're building a real app:
Describe what you want
Writes & runs code
Saves every version
Hosts your app
Database & login
Visit & use your app
The flow: You talk to Claude Code → it pushes to GitHub → Vercel auto-deploys → your app is live.
How I actually build things — from idea to deployed product
Workflow
Describe your idea in plain English. Claude helps you turn it into a high-level spec with structure, features, and technical decisions.
Give it the spec. It asks clarifying questions — "What database? What auth? How should this page look?" — then starts building.
It writes the code, creates files, installs dependencies. Usually 80-90% right on the first try.
Iterate with natural language: "Keep the nav bar visible when I scroll." "Add dark mode." "Fix the mobile layout." Repeat until done.
Step 1
Step 2
🏗️ LossRunner — a Workers' Comp loss run app built in 3.5 hours. No code written by hand. Read the full conversation →
Claude asked 63 questions in 19 minutes — understanding the business model, user types, data flows, and edge cases before writing a single line of code.
Turned answers into a detailed product spec — database schema, API routes, UI components, user roles — and asked for approval before building.
47 files generated. Auth, dashboard, PDF parsing, loss run comparison engine — all from the conversation. Tested and refined in real time.
Step 3
correct on first try
The structure works. The core features work. The design is reasonable.
✅ Core functionality works
✅ Data flows correctly
✅ Auth and security basics
⚠️ Mobile layout may need tweaks
⚠️ Edge cases not handled yet
⚠️ Design polish comes in step 4
Step 4
This is where the magic happens. Natural language → real changes.
You don't need to know how to implement something. You just need to describe what you want.
Watch Out
"Build me an app" → garbage. Be specific about features, data, and user flows. The more context you give, the better the output.
Without Git, one bad change can destroy hours of work. Commit early, commit often. It's your undo button.
Start with the core feature. Get it working. Then add the next thing. Scope creep kills projects.
60%+ of your users are on phones. Test early. AI-generated layouts often need mobile refinement.
Level Up
AI models have a fixed memory. Once ~70% is consumed, quality drops fast — hallucinations increase, instructions get ignored. Start fresh sessions for new tasks.
MCP servers give Claude superpowers (databases, APIs, file systems) — but each one injects tool definitions that eat context. Only enable the MCPs you need for the current task.
Long conversations degrade quality. Use /compact to summarize and reset context when Claude starts losing the thread or repeating itself.
Don't ask Claude to build your auth system AND redesign the UI in one session. Focused sessions = better output. Spin up parallel sessions for parallel work.
Drop a CLAUDE.md in your project root with key decisions, conventions, and architecture. Claude reads it every session — persistent memory without burning context.
Spend 5 minutes writing what you want before asking Claude to build. A clear spec in the first message sets the tone for the entire session. Garbage in = garbage out.
The features that make builders fall in love with it
Connect Everything
Give Claude hands to reach into your world
MCP is an open standard that lets Claude connect to external tools and data sources — databases, APIs, file systems, SaaS platforms. Instead of copy-pasting data into chat, Claude can go get it itself.
Each MCP server injects tool definitions into Claude's context window. 5 MCP servers can eat 10-15% of your context before you type a message.
Rule of thumb: Only enable the MCPs you need right now. Disable the rest.
Think of MCPs like browser extensions — powerful individually, but too many slows everything down.
Secret Weapon
Command-line interfaces are text in, text out — exactly what LLMs are built for. Unlike GUIs that require clicking and navigating, CLIs let Claude read documentation, run commands, and parse output natively. It's the difference between describing a screenshot vs. reading a manual.
--help gives Claude everything it needsPro tip: If a service has a CLI, tell Claude to use it. It'll be faster and more reliable than any API integration.
Persistent Memory
Every new Claude session starts with zero memory. You'd have to re-explain your tech stack, coding conventions, file structure, and preferences every single time.
Drop a CLAUDE.md file in your project root. Claude reads it automatically at the start of every session. It's like onboarding a new developer — except it happens in milliseconds.
Scale Yourself
Run multiple Claude Code sessions in parallel, each working on a different part of your project. They all see the same codebase — just like real developers on a team.
This is the unlock. You're not waiting for one thing to finish before starting the next.
Safety Net
Git is how professional teams track changes, undo mistakes, and collaborate. It's also intimidating for beginners. Claude Code handles it for you.
If Claude breaks something, you can always go back. Just say:
"Undo the last change"
"Go back to how it was yesterday"
Claude handles the git commands. You speak English.
Trust Controls
Claude Code asks before doing anything potentially destructive — deleting files, running shell commands, pushing to production. You approve or deny each action.
As you get comfortable, expand what Claude does autonomously. Start locked down, loosen over time.
Most builders end up on "auto-approve file edits, ask for shell commands" — the sweet spot.
Run claude --dangerously-skip-permissions to let Claude run without asking for approval on every action. It'll read, write, execute, and deploy without stopping. Use with caution — great for experienced builders who trust the process, but start with permissions on until you're comfortable.
Investment
Heavy usage of Claude Code without watching API bills
If you'd spend 10+ hours/month coding with Claude, Max Plan pays for itself vs. API pricing.
Non-technical builders have longer conversations, more back-and-forth, more iteration. That eats tokens fast. Flat rate wins.
*Fair use limits apply. Pro plan ($20/mo) works for lighter usage. Free tier available to try it.
Beyond the Terminal
OpenClaw wraps Claude Code in a persistent runtime — connected to your messaging apps, calendar, email, smart home, and more. It's not just a coding tool anymore. It's an always-on AI agent that lives on your machine.
docs.openclaw.ai · github.com/openclaw/openclaw · discord.com/invite/clawd
Action Items
claude.ai — This is your primary AI partner for specs, code, and iteration.
Terminal-based coding agent. This is what actually builds your projects.
An internal tool. A personal dashboard. A landing page. Something small and useful.
Open Claude. Describe what you want. Let it ask questions. Get to a spec. Then build.
Work Smarter
Tools that make the building process faster and smoother
Voice-to-text that works everywhere — your terminal, IDE, browser. Dictate to Claude Code instead of typing.
AI meeting notes that actually work. Records your calls, generates structured summaries — perfect for capturing specs from client conversations.
AI-native code editor. VS Code fork with built-in AI — autocomplete, multi-file editing, chat. The visual alternative to Claude Code.
Pro tip: Combine these — Granola captures the spec from a call, Wispr lets you dictate it into Claude Code or Cursor, and you ship without typing a line.
Learn More
Mike Maseda — mike@mamv.co — Happy to chat about your project ideas.
Let's talk about what you want to build.
Mike Maseda
mike@mamv.co
Head of Sales & Ops, GenHealth | AI Consultant, MAMV Ventures
Remember: the hardest part is starting.
Pick one thing. Build it this week.
Security, authentication, and environment variables — reference material for when you're ready to go deeper
Great for your second or third project, once the basics are solid.
Software Engineering 101
You don't need a CS degree to ship secure software. But you do need to know a few rules. These apply whether you wrote the code or Claude did.
API keys, database passwords, and tokens go in a .env.local file — never in your code. This file is ignored by Git so it never gets uploaded to GitHub.
If you accidentally push a key to GitHub, consider it compromised. Rotate it immediately. Also watch for TypeScript's process.env.KEY! — the ! suppresses the "might be undefined" warning but crashes at runtime if the var isn't set.
If you use Supabase, RLS is your most important security layer. It ensures users can only read/write their own data — enforced at the database level, not just your UI.
auth.uid()⚠️ Real-world gotcha: Cron jobs & background tasks use the service role, where auth.uid() is NULL. Your RLS policies will silently block them unless you add a policy for auth.role() = 'service_role'.
If you deploy with Vercel, Netlify, or Cloudflare Pages — HTTPS is automatic. Never deploy over plain HTTP. Set environment variables in your hosting dashboard, not in code.
This file tells Git what to never upload. Your .env.local, node_modules/, and any credentials should be listed here. Claude usually creates this for you — but always double-check.
Deep dive: codingformarketers.com — "Environment Variables & API Keys" section
Software Engineering 101
If your app has user accounts, you need auth. Don't build it from scratch — use a proven service. Here's the minimum you need to get right.
Never roll your own login system. Use Supabase Auth, Clerk, or Auth0. They handle password hashing, session tokens, email verification, and OAuth — things that are easy to get wrong and dangerous when you do.
Your app has two sides. The client (browser) is visible to everyone — never trust it with secrets. The server is private — this is where sensitive operations happen.
auth.uid()These are real bugs found across multiple projects:
is_admin on the same table being queried) — causes silent circular failuresLearn more: codingformarketers.com — covers the full software stack, auth, environment variables, and debugging