Agentic setups turning upside down
Before:
Ghostty Terminal <> herdr <> multiplexed terminal sessions
Now:
Hedr <> Ghostty Terminals
thanks @typecraft_dev 🐏
what we hear most about the herdr app: "i don't want to leave my terminal."
you don't. those are real libghostty terminals. it's your terminal, with your own mods and layouts, anything you can imagine.
tmux keys? sure, bring them.
That’s really impressive and enabling so many new apps.
Thanks @mitchellh
libghostty 1.0 is here! I started Ghostty ~5yrs ago to work towards this. Today, libghostty is in hundreds of projects used by tens of millions of people. 1.0 marks ABI stability, quality, and an invite for distros to begin packaging. mitchellh.com/writing/libgho…
Since starting the Ghostty project in 2022, the most common question I've been asked is "why?". Today, I'll answer this by showing the broad industry impact libghostty has already had.
On top of JetBrains and Neovim already mentioned (that alone being over 10 million users):
-KDE developers are experimenting with a Konsole replacement built on it.
- Vercel's Turborepo uses it to render task output in its TUI.
- devenv uses it to drive shells.
- Antithesis used it to fuzz TUI programs and found real bugs in btop and others.
-Ghostel brings it to Emacs.
- ghostty-web is the WASM build with an Xterm.js-compatible API.
- Commercial products like OrbStack embed it.
- And a whole wave of new terminals and "agentic" coding tools, such as herdr and cmux, have been built on it rather than writing yet another emulator.
There are community bindings for Rust, Go, JavaScript, Swift, Python, .NET, Elixir, Dart, Java, and more, and the awesome-libghostty list tracks well over a hundred projects.
More people today use a terminal powered by libghostty than use Ghostty the application, by a very large multiple. And libghostty is having a much larger impact than being "just" a terminal emulator.
I'm excited to see where people take it from here.
If you build something so easy that you just put in one prompt, what is it worth?
Nothing, because everybody can do it in a minute.
What still matters when agents write the code 🧵
Full conversation, /skill issue #3 (incl. why I finally don't have 100x more ideas than I can build):
youtube.com/watch?v=uwZM6QNu…
Written version:
linkedin.com/pulse/code-isnt…
The mistake most engineers make, me too: looking for the hardest problem to solve instead of seeing how users use stuff and finding what makes their life better.
Most very complex problems aren't solving a problem in day-to-day life.
What still matters:
→ System thinking: the side effect of learning to code, more than the language itself
→ Knowing what's value to someone: discovery and validation
→ Persistence: people give up too early
You used to be able to win with pure capability. If you could build a to-do app 20 years ago, you had value, because most people couldn't.
Now a lot of people can. That's not your moat anymore.
Harvey's marketing team built agents that make a tailored product film for every target account.
I rebuilt the pattern with Claude Code to see if a small team could run it.
42 account films. Every one an editable design. The recipe is open.
teapawt.org/gabepereyra/status/211…
How @ShawnFarsai uses @figma @ElevenLabs and @claudeai to create hyper personalized marketing videos with agents
Repo and setup: github.com/imgly-labs/codesi…
Install the CoDesign MCP, give it your brand, render one film for one account. Use FILMS3000 for free credits.
Want one for your own accounts? Reply "film" and I'll help you set it up.
The last 30 films took 17 min 56 s, five agent lanes in parallel.
Then I gave the repo to a fresh agent with no context. It reproduced a film in about 4 minutes. That was the test I cared about: can someone else run this without me.
What I did differently:
- ~220 lines of film logic instead of ~750
- every film is saved as an editable design, so someone in marketing can fix one frame without re-rendering
- all of it runs from Claude Code with @getcodesign, the design MCP we build at IMG.LY
Here is one of the 42, unedited. Made for the lending legal team at Nordvik Bank.
The brand (Quillon) and the accounts are fictional, so I could publish everything.
Full credit to @harvey for the idea. @ShawnFarsai and @gabepereyra built it with Figma, ElevenLabs and a ~750-line film engine: 36 films in 69 minutes. Their write-up is worth reading in full.
The last 30 films took 17 min 56 s, five agent lanes in parallel.
Then I gave the repo to a fresh agent with no context. It reproduced a film in about 4 minutes. That was the test I cared about: can someone else run this without me.
Repo and setup: github.com/imgly-labs/codesi…
Install the CoDesign MCP, give it your brand, render one film for one account.
Use FILMS3000 for free credits.
Want one for your own accounts? Reply "film" and I'll help you set it up.
"I think in the future you don't store source code anymore. You just store the spec and you just generate."
I talked about specs, product roles and design tools. A few takeaways 🧵
What's left of a design editor?
For me: just the canvas, the variations, and a way to point at "this one". With ten blue rectangles, it's hard to describe the fourth one from the left.
And the free Photoshop rebuilds? I find them impressive, but I could not care less. Rebuilding what already exists is worth nothing.
Video (4 min): youtu.be/LwOfdlSL9Wc
