Superlogical costs four times tmux to start, and wins from the second full terminal

Mitchell Hashimoto published the memory numbers for his new terminal multiplexer and left out the one that decides the argument.

The benchmarks went up on 2 September, measured on an M4 Max under macOS, using physical footprint. His framing is scale, and the reason he gives for caring about it is agents. "They like to spawn more sessions than we've ever seen, and they are all their own clients."

The charts are honest, including about where Superlogical loses. An empty server takes 10.6 MiB against tmux's 2.50 MiB. An empty terminal costs it 68 KiB against tmux's 15 KiB. On both of those, tmux wins by roughly a factor of four.

Hashimoto says there is "an early breakpoint with busy terminals and clients" where Superlogical pulls ahead. He does not say where it is. So where is it?

So we did the subtraction

The numbers are all in his own charts. They just sit across four separate images (which is probably why nobody had added them up).

Fill a terminal with ten thousand lines of prose and tmux spends 4.89 MiB holding it. Superlogical spends 407 KiB, about a twelfth as much. Call it a saving of four and a half megabytes per filled terminal, against a starting deficit of roughly eight.

Which puts Superlogical behind at one filled terminal and ahead at two.

Two.

That is a narrower claim than "scale"

Superlogical is not better at having many terminals. Per empty terminal it is four and a half times worse than tmux, and always will be.

What it is better at is holding whatever is inside them. Which happens to be roughly what agents produce: long, noisy scrollback nobody is reading. So the AI framing probably lands, but through storage rather than session count, and those are different products with different competitors.

Per client connection, with fifty filled terminals attached, Superlogical adds 85 KiB and tmux adds 157 KiB. Both fine. Zellij out of the box, with no configuration at all, adds 1.56 GiB.

Sit with that last one for a moment

Fifty attached clients of default Zellij would come to something like 78 GiB of server memory before anybody runs a command.

Hashimoto went and found out why, and the answer is the sort of thing you only learn by reading somebody else's plugin system. Zellij ships a tab bar and a status bar as wasm plugins. Each tab instantiates them. Each instantiation is its own wasm runtime (yes, really). He also reports the memory was never reclaimed after he closed the tabs, though he says he did not dig into that further.

He is generous about tmux, and specific about why it wins. "Credit where credit is due tmux has excellent memory usage for empty states."

This is a recruiting document aimed at one person

That person is the tmux user who has never had a reason to move, and the numbers are good enough to work on them. With one caveat worth saying out loud. They are the vendor's numbers, on one machine, on one operating system, and nobody has reproduced them.

Superlogical is not shipping yet either. "Hope we can bring you something you can use soon" is where the demo ended in late August.

The architecture underneath is what we would want stress-tested first. Hashimoto describes "a custom binary protocol where the server is maintaining N replica distributed terminal state machines," which is how you get a session that survives closing your laptop and reopening it on a phone. It is also a class of system with a long history of interesting failure modes, and that tiny per-terminal footprint is likely only true while the replication keeps doing something clever with it.

So what is this company actually selling?

Not a terminal, on its own telling. The plan on superlogical.com runs to three steps: build an incredible multiplexer, make everything in it composable, make it safe and operable in production. The team is Hashimoto plus HashiCorp's first employee, Vercel's former VP of Design, and a designer from Poolside (a multiplexer does not usually need a Head of Experience).

That is not a roster you assemble to compete with tmux.

Hashimoto says the platforms for the first public release are undecided. We would bet on macOS and the web first and Linux after, and we would be glad to be wrong, because the people most likely to be running fifty agent terminals at once are not on a Mac.