<- all tokdocs
2026.09.10@starterstory

The $1M ARR Stack: What Fastlane Runs On at $69K/Month

Business
Watch on TikTok

View on TikTok ->

The tech stack behind a near-million-dollar SaaS is almost boring, and that is the point. In this 33-second clip, Fastlane founder Gaurav walks through the entire stack running his AI video tool. Fastlane launched around March 2026 and by late August was doing $69K/month with over 1,000 paying customers and $1M ARR across the business. The stack is a short list of managed services, most of them cheap. What is not on the list matters as much as what is: no self-hosted databases, no Kubernetes, no custom auth.

The Full Stack, Line by Line

Gaurav names each tool and what it does. The on-screen graphics add rough monthly costs:

  • Convex (backend + database) -- shown as free
  • Vercel (frontend hosting) -- ~$40/mo
  • Railway (hosts the microservice that builds the content) -- ~$150/mo
  • Clerk (authentication) -- ~$25/mo
  • Resend (transactional + mailing list emails) -- ~$100/mo
  • Axiom (logs and observability) -- ~$25/mo
  • OpenAI + Claude (generate copy and captions) -- free tier shown
  • Fal AI (image and video generation) -- ~$5k
  • Claude Code, 20x Max plan (shipping features) -- ~$200/mo

Note the transcript renders "Fal AI" as "FileAI," which is a speech-to-text slip. The on-screen card reads fal.ai, the generative media platform that gives developers API access to image and video models like FLUX and Kling.

Managed Services Over Infrastructure

Every choice here trades control for speed. Convex handles backend and database as one reactive service instead of a separate Postgres instance plus an API layer. Clerk replaces the auth code most teams rewrite badly. Resend handles email deliverability. Vercel and Railway remove server management. The founder is not running infrastructure, he is renting outcomes.

The one heavy line item is Fal AI at roughly $5k, and that makes sense. Fastlane generates short-form marketing videos, so image and video inference is the actual product cost, not overhead. Everything else is under a few hundred dollars a month. A stack that produces $69K in monthly revenue costs a small fraction of that to run.

Claude Code as a Shipping Multiplier

Gaurav singles out one tool. Claude Code on the 20x Max plan, roughly $200/mo, is "an absolute must for the speed that we ship at." He is not describing it as a code assistant. He frames it as the thing that lets a small team ship fast enough to keep up with what customers ask for. At two months from launch to this revenue, iteration speed is the constraint, and he is paying to remove it.

The Real Edge Is Not in the Stack

The video is a stack rundown, but the Starter Story writeup around it makes the sharper point. Gaurav did roughly 2,000 customer calls. The stack is replaceable, and most of these tools have three viable competitors each. What is not replaceable is knowing exactly what to build because you talked to two thousand people who would pay for it. The tools let him ship. The calls told him what to ship.

Key Takeaways

  • A $1M ARR SaaS can run on a short list of managed services, most costing under $200/mo each.
  • The largest cost, Fal AI at ~$5k, is the product itself (video generation), not infrastructure overhead.
  • Convex, Vercel, Railway, Clerk, Resend, and Axiom cover backend, hosting, auth, email, and observability without a single self-managed server.
  • Claude Code on the 20x Max plan is called out as the shipping-speed multiplier, not a nice-to-have.
  • The stack is commodity. The moat was 2,000 customer conversations that told him what to build.

Resources

Published September 10, 2026. Writeup generated from a favorited TikTok.