The Modern Startup Stack

Shreyas Prakash headshot

Shreyas Prakash

Choosing a web framework is like choosing your first pokémon.

I didn’t want to succumb to the ‘new hotness’ problem with the myriad of JS frameworks to choose from (Angular, Vue, React, Solid). I wanted something that i can choose and stick to for atleast a decade. So I resorted to a Rails monolith for building apps (but with a slight twist)

To start with, I chose Ruby on Rails. It’s a Lindy-compatible web framework that’s mature, stable, and well documented. It’s been around for over 20 years, built on a language that’s been around for over 30 years.

I was prompted to take the Ruby on Rails red pill through a recent talk by Irina Nazarova for the RailsConf keynote.

Although Ruby on Rails was touted as a great framework for a solo developer, it still missed the reactive-component jazz. React made complex UI possible. In fact, the #1 request from young Rails startups was the need for proper React support in the Rails community.

React and other reactive frameworks excel at building highly interactive user interfaces (UIs). Not just that, there are some killer libraries which I could only find in the react ecosystem such as framer-motion, svelte-motion etc

Surprisingly, there was a solution to bridge the gap between Rails and React. Inertia.js served as that bridge, seamlessly integrating traditional Rails templates with reactive frameworks like React, Vue, and Svelte.

In fact 90% of the UI can be done on Rails and Hotwire. And the reactive frameworks could be used to selectively sprinkle on top of Rails for more complex UI requirements. What sort of reactive framework should I choose then? React, Vue, Svelte, Angular?

My primary considerations were code simplicity, readability, and community support. Performance mattered, but it wasn’t the sole determinant. Google Trends provided a snapshot of each framework’s popularity and adoption rates, helping me gauge their standing in the developer community.

React continues to capture the lion’s share of attention, whereas Svelte remains obscure, hippie and unnoticed

Svelte tipped the fence for me, owing to their focus on simplicity, more than performance. It’s just like React or Vue, but it’s a compiler.

Svelte generates highly optimized JavaScript code, resulting in faster load times and a smaller bundle size. With Svelte, you only ship the code that your app needs, reducing the initial load time significantly.

Svelte was like Rails for the front end (low barrier of entry, sensible defaults etc) and I wanted to get the two working together. I also wanted to use Inertia which is a sort of ‘glue’ between your front and backend that allows you to use standard controllers without having to write an API.

Svelte’s Tutorial Website is top tier

The way that Svelte has structured their tutorial website and documentation is top tier. When you go to their tutorial page it greets you with some instructions on the left and an IDE on the right. I liked the attention to detail in the tutorials. They even didn’t allow visitors to copy-paste code onto the IDE on the right as they wanted visitors to manually type down the code for better retention. I just fell in love with this tutorial page :)

Their tutorial series consists of 4 parts with each having different lessons all separated by categories.

React also has a tutorial on their website. In no way is that tutorial better than Svelte’s. React has structured their tutorial through documentation rather than an interactive layout like Svelte’s.

React's website for teaching

React’s tutorials are more structured around documentation whereas Svelte is structured around building projects

This may just be personal preference, but I loved how thought-through the learning process for Svelte is. It’s so beginner friendly.

You want to make a component that adds two numbers together?

<script>
let a = 0, b = 0;
</script>

<input type="number" bind:value={a} />
<input type="number" bind:value={b} />
<p>{a + b}</p>

Now switch to React:

import { useState } from 'react';

export default () => {
	const [a, setA] = useState(0);
	const [b, setB] = useState(0);

	return (
		<input type="number" value={a} onChange={(e) => setA(e.target.value)} />
		<input type="number" value={a} onChange={(e) => setB(e.target.value)} />
		<p>{parseInt(a) + parseInt(b)}</p>
	)
}

Goes without saying that the language of Svelte is highly readable.

The reason why I committed to a certain tech in the first place is because I picked the most popular tech at the time I was learning it (~8 years ago).

e.g., I picked MySQL because it’s a very popular database. If I get a bug or a strange error, I can always google it out.

— Tony Dinh 🎯 (@tdinh_me) June 20, 2024

Some hard fought lessons from Tony Dinh here. You just pick a framework and stick to it. No back and forth!

Now that I’ve locked in on my stack for the next decade, I’ll touch a bit upon the setup I use to build Rails+Svelte apps:

Inertia Rails Integration

rails new rails-inertia-svelte

Add the relevant gems:

bundle add inertia_rails
bundle add vite_rails

Install vite:

bundle exec vite install

Setup front end libraries

The previous step will create the package.json file so you can then install the required npm packages:

npm install --save-dev @inertiajs/inertia
npm install --save-dev @inertiajs/svelte
npm install --save-dev @sveltejs/vite-plugin-svelte
npm install --save-dev svelte

To make sure package.json allows for modules,

"type": "module",


{
  "type": "module", 
  "devDependencies": {
    "@inertiajs/inertia": "^0.11.1",
    "@inertiajs/svelte": "^1.2.0",
    "@sveltejs/vite-plugin-svelte": "^3.1.1",
    "vite": "^5.3.1",
    "vite-plugin-ruby": "^5.0.0"
  }
}

Configure Vite Plugin

Add the Vite Svelte plugin to the project by editing the vite.config.ts file (found in the project’s root). After editing it should look like the following:

import { defineConfig } from 'vite'
import RubyPlugin from 'vite-plugin-ruby'
import { svelte } from '@sveltejs/vite-plugin-svelte'

export default defineConfig({
  plugins: [
    RubyPlugin(),
    svelte(),
  ],
})

Next configure the vite.json file (located in the config folder) to point to the correct location for our entrypoint. The sourceCodeDir value needs to be frontend like so:

{
  "all": {
    "sourceCodeDir": "app/frontend",
    "watchAdditionalPaths": []
  },

In the next step we’ll create the entrypoint we refer to above.

Setup Rails Svelte entrypoint

This step was a bit different in the few places I found examples however this seems to be the accepted and most orthodox way of organising the files.

First create the application.js file:

mkdir -p app/frontend/entrypoints
touch app/frontend/entrypoints/application.js

The contents of application.js should be the following:

import { createInertiaApp } from '@inertiajs/svelte'

createInertiaApp({
  resolve: name => {
    const pages = import.meta.glob('../pages/**/*.svelte', { eager: true })
    return pages[`../pages/${name}.svelte`]
  },
  setup({ el, App, props }) {
    new App({ target: el, props })
  },
})

Next create a front end Svelte component as an example.

First create the Home.svelte file:

mkdir -p app/frontend/pages
touch app/frontend/pages/Home.svelte

Now add images under /public/assets/images/ folder. I’ve added a rickroll.gif for demonstration purposes.

Then add some HTML markup to the Home.svelte component:

<script>
    let src = "assets/images/rickroll.gif";
</script>

<img {src} alt="a random image" />

Generate Rails Code

To load our Home.svelte component we’ll need a Rails controller which we’ll create like so:

rails generate controller Home index

Then make sure the index method sends the correct response using the Inertia library:

class HomeController < ApplicationController
  def index
    render inertia: 'Home', props: {
      name: 'Inertia Rails'
    }
  end
end

To get the above to work on the main path edit the routes.rb file like so:

Rails.application.routes.draw do
  root 'home#index'
end

Final Steps

We’re almost done. In fact we have everything we need but this bit is to make it a bit simpler to run. Create a dev file in the bin folder and make it executable:

touch bin/dev
chmod +x bin/dev

Edit the newly created dev file and add the following bash code:

# !/usr/bin/env sh

if ! gem list foreman -i --silent; then
  echo "Installing foreman..."
  gem install foreman
fi

exec foreman start -f Procfile.dev "$@"

Finally create the Procfile.dev file which is referred to above:

touch Procfile.dev

And then add the following lines. The p flag is important and is what caught me out:

vite: bin/vite dev
web: bin/rails s -p 3000

And that’s it! Now you can run bin/dev from your terminal and you should see something like the following:

Inaugurating my first Svelte+Rails app with a rick roll meme

Subscribe to get future posts via email (or grab the RSS feed). 2-3 ideas every month across design and tech

2026

  1. How I started building softwares with AI agents being non technical

2025

  1. Legible and illegible tasks in organisations
  2. L2 Fat marker sketches
  3. Writing as moats for humans
  4. Beauty of second degree probes
  5. Read raw transcripts
  6. Boundary objects as the new prototypes
  7. One way door decisions
  8. Finished softwares should exist
  9. Essay Quality Ranker
  10. Export LLM conversations as snippets
  11. Flipping questions on its head
  12. Vibe writing maxims
  13. How I blog with Obsidian, Cloudflare, AstroJS, Github
  14. How I build greenfield apps with AI-assisted coding
  15. We have been scammed by the Gaussian distribution club
  16. Classify incentive problems into stag hunts, and prisoners dilemmas
  17. I was wrong about optimal stopping
  18. Thinking like a ship
  19. Hyperpersonalised N=1 learning
  20. New mediums for humans to complement superintelligence
  21. Maxims for AI assisted coding
  22. Personal Website Starter Kit
  23. Virtual bookshelves
  24. It's computational everything
  25. Public gardens, secret routes
  26. Git way of learning to code
  27. Kaomoji generator
  28. Style Transfer in AI writing
  29. Copy, Paste and Cite
  30. Understanding codebases without using code
  31. Vibe coding with Cursor
  32. Virtuoso Guide for Personal Memory Systems
  33. Writing in Future Past
  34. Publish Originally, Syndicate Elsewhere
  35. Poetic License of Design
  36. Idea in the shower, testing before breakfast
  37. Technology and regulation have a dance of ice and fire
  38. How I ship "stuff"
  39. Weekly TODO List on CLI
  40. Writing is thinking
  41. Song of Shapes, Words and Paths
  42. How do we absorb ideas better?

2024

  1. Read writers who operate
  2. Brew your ideas lazily
  3. Vibes
  4. Trees, Branches, Twigs and Leaves — Mental Models for Writing
  5. Compound Interest of Private Notes
  6. Conceptual Compression for LLMs
  7. Meta-analysis for contradictory research findings
  8. Beauty of Zettels
  9. Proof of work
  10. Gauging previous work of new joinees to the team
  11. Task management for product managers
  12. Stitching React and Rails together
  13. Exploring "smart connections" for note taking
  14. Deploying Home Cooked Apps with Rails
  15. Self Marketing
  16. Repetitive Copyprompting
  17. Questions to ask every decade
  18. Balancing work, time and focus
  19. Hyperlinks are like cashew nuts
  20. Brand treatments, Design Systems, Vibes
  21. How to spot human writing on the internet?
  22. Can a thought be an algorithm?
  23. Opportunity Harvesting
  24. How does AI affect UI?
  25. Everything is a prioritisation problem
  26. Now
  27. How I do product roasts
  28. The Modern Startup Stack
  29. In-person vision transmission
  30. How might we help children invent for social good?
  31. The meeting before the meeting
  32. Design that's so bad it's actually good
  33. Breaking the fourth wall of an interview
  34. Obsessing over personal websites
  35. Convert v0.dev React to Rails ViewComponents
  36. English is the hot new programming language
  37. Better way to think about conflicts
  38. The role of taste in building products
  39. World's most ancient public health problem
  40. Dear enterprises, we're tired of your subscriptions
  41. Products need not be user centered
  42. Pluginisation of Modern Software
  43. Let's make every work 'strategic'
  44. Making Nielsen's heuristics more digestible
  45. Startups are a fertile ground for risk taking
  46. Insights are not just a salad of facts
  47. Minimum Lovable Product

2023

  1. Methods are lifejackets not straight jackets
  2. How to arrive at on-brand colours?
  3. Minto principle for writing memos
  4. Importance of Why
  5. Quality Ideas Trump Execution
  6. How to hire a personal doctor
  7. Why I prefer indie softwares
  8. Use code only if no code fails
  9. Personal Observation Techniques
  10. Design is a confusing word
  11. A Primer to Service Design Blueprints
  12. Rapid Journey Prototyping
  13. Directory Structure Visualizer
  14. AI git commits
  15. Do's and Don'ts of User Research
  16. Design Manifesto
  17. Complex project management for product

2022

  1. How might we enable patients and caregivers to overcome preventable health conditions?
  2. Pedagogy of the Uncharted — What for, and Where to?

2020

  1. Future of Ageing with Mehdi Yacoubi
  2. Future of Equity with Ludovick Peters
  3. Future of Tacit knowledge with Celeste Volpi
  4. Future of Mental Health with Kavya Rao
  5. Future of Rural Innovation with Thabiso Blak Mashaba
  6. Future of unschooling with Che Vanni
  7. Future of work with Laetitia Vitaud
  8. How might we prevent acquired infections in hospitals?

2019

  1. The soul searching years
  2. Design education amidst social tribulations
  3. How might we assist deafblind runners to navigate?