Mitesh ShahGet to know me
As a kid, I played with Legos a lot. At first, I built the picture. Then, I built what I could picture. Eventually I wanted to build things that I didn't have the pieces for. And that's where it began.
Solve the problem by building your own pieces.
I've been a builder ever since. I couldn't afford the Lego motors, so I soldered cheap model motors into the housings. I wanted a PlayStation steering wheel, so I broke apart a controller and built one out of styrofoam and cardboard with the original circuitry intact. I fixed a VCR. Wrote songs in Logo beep. Made games. Made apps. Now I bring bigger ideas to life through startups. And I feel like I'm just getting started.
Snap together the right pieces, and we can build anything we want — the right teams, the right software, the right products, and even the right robots.
How I work
Small teams. Simple stacks. Surface area you can hold in your head.
The work I'm proudest of has come from groups tight enough that one person can hold the whole product in their head, and stacks simple enough that the whole architecture fits there too. People mistake this for a constraint. It's the opposite. Simplicity is what makes ambitious work possible.
Every tool, every framework, every additional person has to earn its place. Not by what it'd look like at Netflix scale. Not by what's in fashion this quarter. The bar is: does this help us make the thing?
The hardest one to defend in big organizations is the one I'm most right about: trust over process. Process is what you reach for when you don't trust anyone in the room. With the right people, you don't need much of it. With the wrong people, no amount of it will save you. And it's even easier to trust when the team is shipping actual product instead of closing tickets.
Why it works
Big teams have a quiet failure mode. Coordination scales geometrically. Meetings and ceremony eat the runway. Each person looks busy and productive. The organization atrophies.
A small group with the right stack will outpace a committee of forty. Almost always.
In the AI era, this is even more true. The bottleneck isn't building speed. It's decision quality. And decision quality lives in small groups thinking clearly together.
That's the whole game.
An idea I keep coming back to
Product IDEA Loop
Insight, Design, Engineering, and Analysis in the same working session. Build something concrete, learn from it, and decide what happens next. The goal is better decisions with every loop.
Explore the IDEA LoopThe Me Stack
Technology, product decisions, and team coordination belong in the same stack. Together they determine what a small group can actually ship. The name is a nod to G Stack and P Stack. Mine keeps the whole product understandable, leaves room for judgment, and checks the thing that actually runs. These are defaults, not doctrine. Each product earns its own choices.
- 00
AI-first from here on out.
Think with AI. Do with AI. Build with AI.
- 01
Sketch it out.
Build a prototype with AI in v0 or Replit instead of a doc or slide deck. It answers questions up front. Kills bad ideas earlier. Strengthens great ones until your teammates are thrilled to build them.
- 02
TypeScript / JavaScript everywhere.
Frontend, backend, even native. Seamless prototyping. Code reuse across boundaries.
- 03
Intentional frugality.
Not chasing token leaderboards. Not opening 10 agent windows at once. Not hiring 5 engineers at once.
- 04
Assume rewrites are cheap when needed.
Need a native Swift app? Cutting-edge Apple SDKs? Platform-grade transitions? First we answer why. If the product needs it, a small team can go native with AI instead of treating it as a new hiring plan.
- 05
Keep the team small enough to decide.
5 people × 10 hours ≠ 10 people × 5 hours. The math misses the handoffs, lost context, and time spent getting everyone aligned. Fewer people can mean more room to make the thing.
- 06
Coordinate around decisions.
Trust the people doing the work. Talk when it changes a decision, write down what matters, then get back to making. Keep recurring meetings only when they earn the time.
- 07
Prove it in the product.
Read the code, then run the flow. Check the actual response, the saved data, and what the person sees. A green test is a clue; working software is the evidence.
Technology choices
One language across the product where it helps. Services that take work off a small team. Tools to measure what happens, coordinate and communicate when there's a team, and think with AI. Here are some current favorites and why they earn a place.
| Choice | Why it earns a place |
|---|---|
| shadcn/ui | Beautiful, accessible components whose source lives in the app. The defaults get us moving; owning the code lets the interface become ours. |
| Next.js | Front-end and back-end in one mental model. Three boxes on the architecture diagram collapse into one. |
| React | The component model. Universal, predictable, and every tool — human or AI — already knows it. |
| Tailwind | The styling layer underneath. Fast to shape the details without maintaining a separate CSS naming system. |
| Supabase | Postgres, auth, storage, realtime. One service, four problems solved. Still just Postgres underneath. |
| Expo | React Native without the native pain. iOS and Android from one codebase, one toolchain. |
| TanStack Start | Type-safe full-stack with sharper routing. The right call when Next isn't. |
| Vercel | A branch becomes a working preview. Feedback can happen on the product before anything goes live. |
| PostHog | Product analytics and experiments in one place. Measure what people actually do, test changes, and feed the next product decision. |
| Linear | When there’s a team, keep projects, backlogs, and ownership visible so people can coordinate without turning every decision into a meeting. |
| Slack | The team’s communication backbone. Channels and threads keep questions, decisions, and updates moving while Linear keeps the work visible. |
| Frontier AI | Claude, OpenAI, and Cursor together: one extremely useful workhorse for thinking, writing, coding, and review. Different models bring different strengths; Cursor keeps that help close to the work. |
Have something to build?
A few lines about the product, the team, or the hard part are plenty.