Everyone told me to use Substack. Or Ghost. Or Medium. Or WordPress with a theme. You are not a frontend developer, they said. Focus on writing, they said. Use a platform and stop reinventing the wheel.
I listened to all of that advice. Then I ignored it and built icanbefitter.com from scratch using Next.js, Supabase, and Tailwind CSS. Deployed on Vercel. Custom everything. My code. My database. My rules.
That decision took 10x longer than signing up for Substack would have. And it was the single best decision I made for this project. Here is why.
The Rented Land Problem
When you build your brand on Substack, you are building on rented land. Substack owns the platform. Substack controls the algorithm. Substack decides what features you get, what your page looks like, and what percentage of your revenue they take.
Right now, Substack is popular and generous. But I have watched this movie before. Blogger was popular. Then Google killed it. Medium was the future of writing. Then they paywalled everything and creators lost their audiences. Tumblr was a thriving ecosystem. Then it was sold for a dollar. Every platform follows the same lifecycle: attract creators, build audience, monetize the audience, squeeze the creators.
I cannot afford that risk. icanbefitter.com is not a side project I will abandon in six months. It is a legacy platform that needs to work in 18 years when Avyaansh is old enough to inherit it. If Substack shuts down in 2030, my 18-year plan is destroyed. If my own website has a problem, I fix it myself.
Ownership is not optional when you are building for the long term. It is the foundation.
Most personal brands are built on rented land. When the landlord changes the rules, the brand disappears. I will not build my son's inheritance on someone else's platform.
The Architecture: Why Each Choice
Every technology in my stack was chosen by asking one question: will this still work in 10 years?
Next.js 14 (App Router) — React is the most widely adopted frontend framework in the world. It is not going anywhere. Next.js adds server-side rendering, which means fast page loads and excellent SEO. The App Router with server components by default means most of my pages load without sending JavaScript to the browser. Fast. Simple. Future-proof.
Supabase (PostgreSQL) — Supabase gives me a full PostgreSQL database with authentication, storage, and real-time capabilities. But the key word is PostgreSQL. If Supabase the company disappears tomorrow, my data is in standard Postgres format. I can migrate to any hosting provider in the world. No vendor lock-in. No proprietary data format. Just SQL.
Tailwind CSS — utility-first CSS that lives in the HTML. No separate stylesheet files to maintain. No CSS naming conventions to argue about. The entire design system is expressed in class names that any developer can read. Five years from now, a developer looking at my code will understand exactly what each element looks like without opening a single CSS file.
Vercel — deploys from GitHub automatically. Push code, site updates. No servers to manage. No DevOps to worry about. Vercel handles scaling, CDN, SSL, and everything else that used to require a dedicated operations team. I push code and the site is live globally in under a minute.
This stack is not trendy. It is boring and reliable. Every piece is backed by large companies or massive open-source communities. Nothing in this stack will be abandoned next year. That is the point.
The 4-Agent AI Pipeline
Here is where the custom build really pays off. Substack does not let you build a 4-agent AI pipeline for content production. Ghost does not let you create a custom blog engine with enriched inputs. WordPress with plugins gets you halfway there, but the plugin ecosystem is a maintenance nightmare.
My blog engine runs a pipeline of four AI agents for every post:
Agent 0 — Research: gathers facts, statistics, and unique angles on the topic. Thinks like a journalist. Outputs a research brief that the next agent uses as source material.
Agent 1 — Strategist: takes the research brief and creates the post strategy — title options, slug, outline, tone, target audience, SEO approach. Thinks like an editor-in-chief.
Agent 2 — Writer: takes the strategy and research and writes the full HTML content. This agent runs on Opus — the most capable writing model — because the quality of the final content is the one thing I refuse to compromise on.
Agent 3 — SEO: takes the finished content and adds focus keywords, secondary keywords, meta descriptions, internal link suggestions, and tag recommendations. Thinks like a search marketing specialist.
This pipeline is only possible because I own the entire stack. I can add agents, change models, adjust prompts, and rewire the pipeline without waiting for any platform to add features. The architecture serves my workflow. My workflow does not bend to fit someone else's architecture.
Supabase as Single Source of Truth
Everything on icanbefitter.com lives in one Supabase project. Posts, settings, members, products, analytics, API keys — one database, one authentication system, one storage bucket.
This simplicity is intentional. When you distribute your data across multiple services — posts on Substack, members on ConvertKit, products on Gumroad, analytics on Google — you lose the ability to see the whole picture. You cannot easily answer questions like "which blog post drives the most product sales?" or "which email sequence converts the most Inner Circle members?" because the data lives in silos that do not talk to each other.
With everything in Supabase, every query is a SQL join away. I can correlate blog post performance with member signups with product purchases in a single query. The insights that fall out of connected data are worth more than any individual tool's features.
Row Level Security on every table means that even if someone found a vulnerability in my application code, the database itself would prevent unauthorized access. Security at the data layer, not just the application layer. This is how serious systems are built.
The Rule That Drives Every Decision
Will this still work in 10 years?
That question has killed more feature ideas than any product review meeting. Fancy animations? Maybe not supported in future browsers — keep it simple. Third-party widget for comments? The company might not exist in 10 years — build it in-house. Trendy CSS framework? Will anyone maintain it in 2034 — stick with Tailwind, which has massive adoption and corporate backing.
This is the luxury of owning your platform. You can optimize for longevity instead of launch speed. Substack optimizes for getting you writing in 5 minutes. I optimize for the platform still working when Avyaansh graduates from college.
Different time horizons produce radically different decisions. When your time horizon is "this quarter," you pick the fastest option. When your time horizon is "my son's lifetime," you pick the most durable option. Every choice on icanbefitter.com is made with the longer horizon.
icanbefitter.com is not a blog. It is a software system designed to outlast its creator. Legacy platform. Treat it that way.
What Substack Does Better — And Why It Still Was Not Enough
I will be honest. Substack does some things better than my custom build.
Substack has a built-in audience discovery mechanism. When someone subscribes to one Substack, they get recommendations for similar ones. That network effect is real and valuable.
Substack has a beautiful, simple reading experience out of the box. No design decisions. No font choices. No layout debates. Just write and publish.
Substack handles email delivery, payment processing, and subscriber management. You do not need to integrate Brevo, Razorpay, or Supabase Auth. It is all done for you.
These are genuine advantages. For someone who just wants to write and build an audience quickly, Substack is excellent. I recommend it to people who ask me what platform to use for blogging.
But I am not just writing. I am building a platform. A platform that includes AI agents, custom blog layouts (the Letter to Avyaansh layout is a completely different emotional world from the knowledge posts), a Telegram bot for administration, a shop, a members area, and an AI fitness assistant. None of this is possible on Substack. And all of it is essential to what icanbefitter.com needs to be.
Avyaansh Inherits the Platform, Not Just the Content
Here is the part that makes the 10x development time worth it.
If I had used Substack, Avyaansh would inherit a collection of essays on a platform he does not control. He could not modify the design. He could not add features. He could not integrate new technology. He would be a tenant on someone else's property, reading his father's words through someone else's interface.
With icanbefitter.com, he inherits a complete software system. The code. The database. The AI pipeline. The domain. The deployment infrastructure. Everything. He can modify it, extend it, rebuild it, or let it run as-is. The choice is his because the ownership is his.
That difference — between inheriting content and inheriting a system — is the difference between receiving a collection of letters and receiving a living, evolving platform. Both are valuable. But only one can grow after the creator stops building.
Every late night spent debugging Next.js routing, every afternoon configuring Supabase RLS policies, every weekend building the AI pipeline — it was all investment in a system that will serve my son long after the development effort is forgotten.
Most people build for today. I build for Avyaansh. That is why I built icanbefitter.com instead of using Substack. And I would make the same choice again without a second of hesitation.
Go Win!