replit chased the ai hype and left a simple coding gap
I miss the old Replit.
Not because it was perfect. It was not. But it had one beautiful promise: open the browser, pick a language, write code, run it, share the link.
That was the magic.
No local setup. No Dockerfile. No SDK version fight. No “give me twenty minutes to install Python again.” If you were teaching, prototyping, interviewing, or just trying to show someone a tiny idea, Replit was the easy answer.
It was not trying to be your engineering organization. It was a place to run code.
then the ai wave arrived
Replit saw the AI wave and jumped on it. That part is understandable. Everyone did.
The problem is what happened to the simple runtime experience underneath.
As of May 2026, Replit’s own pricing page describes the free Starter plan around daily Agent credits, built-in database support, media/artifact creation, and publishing up to one project. Core is $25 per month, or $20 per month billed annually, and adds monthly credits, collaborators, parallel agents, unlimited workspaces, publishing in any region, and removing the “Made with Replit” badge. Pro is $100 per month, or $95 per month billed annually, with more credits, more collaborators, more parallel agents, stronger model access, database rollbacks, and premium support.
The Starter documentation is even clearer about the shape of the product now: free users get daily Agent credits, monthly cloud credits, Lite Build, and one free published app. That published app link automatically goes down after 30 days, although it can be republished. Full Build, Plan Mode, third-party connectors, additional published apps, and badge removal require Core or Pro.
This is not evil. It is a business model.
But it is a different product.
The old “I just want to run a small program in the browser” use case now lives inside a platform optimized for AI app generation, credits, publishing, deployments, and upgrade paths. The simple runtime did not disappear entirely, but it is no longer the center of gravity.
That matters because a lot of people never needed an AI pair programmer.
They needed a place to:
- run a quick snippet to test an idea
- share runnable code with a student or colleague
- learn a new language without installing anything
- prototype something tiny without creating a full project
- paste a few lines and see what happens
That use case did not go away.
The tool that served it drifted.
the gap is boring, which is why it matters
There is a pattern in developer tools that keeps repeating.
A tool wins because it is simple. Then growth arrives. Then enterprise features arrive. Then AI arrives. Then the homepage starts explaining a bigger vision. Then the thing people originally loved becomes one feature among many.
The boring use case is usually the one that gets squeezed first.
“Run this code.”
“Show this example.”
“Try this language.”
“Share this tiny experiment.”
These are not glamorous workflows. They do not look great in investor decks. They do not justify a giant pricing table. But they are real, daily developer workflows.
And they are especially important for learning.
Students do not need a full-stack AI builder just to understand a loop. A colleague reviewing a small algorithm does not need a hosted deployment product. Someone experimenting with Clojure or Python in five spare minutes does not need an upsell journey.
They need a text box, a runtime, and a run button.
so I built the thing I wanted
That is why I started coding.pvgomes.com.
It is intentionally simple.
Pick a language. Write code. Run it.
No AI upsell. No paywalled runtime. No account flow before you can think. Just a browser-based coding playground for the use case Replit used to serve so well.
Today it supports:
- Python, powered by Pyodide
- JavaScript, executed in an isolated Web Worker
- Clojure, powered by Scittle
- Markdown preview
- HTML preview
- local file trees per language
- localStorage persistence so your code survives refreshes
- bundled editor/runtime assets where practical, instead of depending on a pile of remote CDNs
The project is deliberately static. The repo is plain HTML, CSS, and JavaScript. It can be served from GitHub Pages or any static host. There is no server to operate and no account system to maintain.
That is the point.
It is not trying to replace a professional IDE. It is not trying to become an AI software factory. It is a small, fast place to run code in the browser.
That category still deserves to exist.
the useful contribution list
The code is open source here:
github.com/pvgomes/coding-playground
There are no open issues at the time I am writing this, which means the contribution path is simple: try it, break it, and tell me what is missing.
Concrete contributions that would help:
- add more language runtimes
- improve the mobile layout
- add better examples for each supported language
- improve accessibility and keyboard navigation
- make the file tree smoother
- add tests for runner behavior and persistence edge cases
- open issues for small papercuts you hit while using it
Even a tiny issue is useful. “The Python runner behaves weirdly with this snippet” is useful. “The Clojure output is confusing here” is useful. “This button is awkward on mobile” is useful.
Open source works best when people contribute the boring friction they actually feel.
the bigger point
I do not think Replit is terrible now. I think Replit became something else.
It is chasing the AI builder market, and maybe that is the right business decision for them. But when a platform moves upmarket or sideways into the hype cycle, the old users do not magically stop needing the old workflow.
They just need a new home for it.
That is what open source is good at.
When a platform drifts away from the people who loved it first, the community can build the missing piece again. Smaller. Sharper. Less ambitious in the good way.
Try the playground:
Contribute here:
github.com/pvgomes/coding-playground
Sometimes the best answer to a bloated tool is not a better pitch.
It is a simpler tool.