Intelligence is foundation
Podcast Subscribe
Builders & Makers Monday, 13 April 2026

Building a Shell by Ignoring the Shell Entirely

Share: LinkedIn
Building a Shell by Ignoring the Shell Entirely

A developer set out to build a better shell. He ended up building something that sits underneath the shell instead - and it's more useful.

The project is called ptylenz, and it's a PTY proxy that structures terminal output into navigable blocks with search and copy functionality. The insight that made it work: the shell isn't on the data path where output matters. The PTY is. That forced a complete rethink of where to intervene.

The Problem with Terminal Output

Terminal output is a stream of text. Commands produce output, it scrolls past, and if you want to find something you either scroll back through history or pipe to grep. There's no structure. No way to say "show me just the output from that git status command three commands ago".

Most attempts to fix this focus on the shell itself. Build a better shell, add features for output parsing, create custom integrations. But that means rewriting bash or zsh or fish, convincing people to switch, and maintaining compatibility with decades of existing tooling.

The developer behind ptylenz tried that approach. It didn't work. The shell is too complex, too deeply integrated into workflows, too hard to replace. So he went looking for a different intervention point.

What a PTY Actually Does

A PTY (pseudo-terminal) sits between the shell and the terminal emulator. It's the layer that handles the raw input and output. When you type a command, it goes through the PTY. When a command produces output, it comes back through the PTY. Every byte of data passes through this layer.

That makes it the perfect place to inject structure. You don't need to modify the shell. You don't need to change how commands work. You just intercept the data as it flows through the PTY, parse it, and structure it before it reaches the terminal.

Ptylenz does exactly that. It's a proxy that sits on the PTY layer, captures command output, breaks it into blocks, and makes each block individually searchable and copyable. The shell never knows it's there. The commands run normally. But the output is suddenly navigable in ways it wasn't before.

Why This Approach Works

Building at the PTY layer means compatibility. It works with any shell - bash, zsh, fish, whatever. It works with any command. It doesn't require changing your workflow or learning new syntax. You just run your commands as normal, and the output is structured automatically.

The key features: output is broken into blocks per command, each block can be searched independently, you can copy a block without scrolling, and there's a navigation interface to jump between commands. It's not significant - it's just solving the basic problem that terminal output is hard to work with.

The developer describes the realisation moment: he was trying to build features into a shell and kept hitting walls. Then he looked at the data path and realised the shell wasn't on it for output. The PTY was. That shift in perspective - from "where does the feature belong?" to "where is the data actually flowing?" - unlocked the solution.

What This Teaches About Architecture

The lesson here isn't about terminals specifically. It's about finding the right layer to solve a problem. The obvious place is often wrong. The shell feels like the natural place to add shell features. But the PTY is where the data lives, so that's where the solution needs to be.

This happens constantly in software. You want to add logging, so you modify the application. But the right place might be the web server. You want to add caching, so you change the database queries. But the right place might be a proxy layer. The question isn't "where does this feature logically belong?" - it's "where is the use point in the data flow?"

Ptylenz found its use point by working backwards from the constraint. Modifying the shell was too hard. Modifying terminal emulators was too fragmented. The PTY was the choke point where all data passed through, so that's where the intervention happened.

Building Something You Actually Use

The other detail worth noting: the developer built this because he wanted it. Not as a side project to learn Rust, not to get GitHub stars, but because terminal output was genuinely annoying and he wanted a better way to navigate it.

That's why it works. It solves a real problem he encounters daily. The features aren't hypothetical - they're the things he actually needs when working in a terminal. Search across command output. Copy a specific block without scrolling. Jump back to see what a command printed five steps ago.

The best tools come from builders scratching their own itch. They don't have to guess what users need - they are the user. Ptylenz is a good example of that. It's not trying to reimagine the terminal. It's just making the existing terminal less frustrating to use.

Read the full writeup and grab the code at DEV.to.

More Featured Insights

Robotics & Automation
Robotic Guide Dogs That Talk Back
Voices & Thought Leaders
Why GPU Scarcity Broke the AI Pricing Model

Video Sources

Theo (t3.gg)
How does Claude Code *actually* work?
Google DeepMind
What's new in Gemma 4?
AI Revolution
China's New Self Improving Open AI Beats OpenAI

Today's Sources

DEV.to AI
I wanted to build a shell. I built a PTY proxy instead.
Towards Data Science
Your ReAct Agent Is Wasting 90% of Its Retries - Here's How to Stop It
DEV.to AI
Memory Hierarch
Towards Data Science
Stop Treating AI Memory Like a Search Problem
Towards Data Science
Write Pandas Like a Pro With Method Chaining Pipelines
The Robot Report
Binghamton researchers create robotic guide dogs that walk - and talk
ROS Discourse
Fast Lossless Image Compression: interested?
ROS Discourse
New "ROS Adopters" page is live - ADD YOUR PROJECT
Ben Thompson Stratechery
Mythos, Muse, and the Opportunity Cost of Compute
Jack Clark Import AI
Import AI 453: Breaking AI agents; MirrorCode; and ten views on gradual disempowerment
Gary Marcus
Even more good news for the future of neurosymbolic AI

About the Curator

Richard Bland
Richard Bland
Founder, Marbl Codes

27+ years in software development, curating the tech news that matters.

Subscribe RSS Feed
View Full Digest Today's Intelligence
Free Daily Briefing

Start Every Morning Smarter

Luma curates the most important AI, quantum, and tech developments into a 5-minute morning briefing. Free, daily, no spam.

  • 8:00 AM Morning digest ready to listen
  • 1:00 PM Afternoon edition catches what you missed
  • 8:00 PM Daily roundup lands in your inbox

We respect your inbox. Unsubscribe anytime. Privacy Policy

© 2026 MEM Digital Ltd t/a Marbl Codes
About Sources Podcast Audio Privacy Cookies Terms Thou Art That
RSS Feed