Fibs and Fibbonacci - is LOGOS faster than Zig? Fibs and Fibbonacci - is LOGOS faster than Zig?<br>Aug 10, 2026<br>#musings<br>#ai<br>#zig<br>#compilers<br>#beef<br>A fib is a trivial or childish lie.

The Fibonacci sequence is an elegant progression of numbers first documented by Indian poet-mathematician Acharya Pingala circa 200 BCE. It was dubbed “Fibonacci” after Leonardo of Pisa, who included it in his “Book of Calculation” (Liber Abaci) in 1202.

The same book helped popularize what we now call Arabic numerals. My top 5 favorite Arabic numerals are 1, 2, 3, 4, and 5. No order. Honorable mention 0.

The Fibonacci sequence begins 1, 1, 2, 3, 5, 8, 13, more on that later.

We are gathered here to evaluate the claim:

LOGOS is the fastest general purpose programming language on earth.

For disambiguation, we’re speaking of the LOGOS at logicaffeine.com, not the Logos at logos-lang.dev or the Logo educational language from 1967 that drew the pictures with the turtle.

Is LOGOS the world’s fastest programming language?

No. We are well within Betteridge’s law of headlines. My assessement:

There is no general concept of a “fastest programming language”, just the fastest for specific jobs.

LOGOS helpfully provides full details of their benchmark runs (archive).

The examples are not realistic workloads, they are algorithm teaching examples.

As a basline, LOGOS performs like Rust because it generates Rust code.

LOGOS claims speedup in a handful of the examples in these ways:

Implementing well-known optimizations that systems languages deliberately leave to the programmer, like Memoization.

Replacing loops with closed-form solutions that directly target these toy benchmark problems.

Implementing C and Zig examples in non-idiomatic ways, e.g. choosing naive recursion over Iteration, signed integers for unsigned data.

In practice, these are non-optimizations .

My conclusion is that calling this the “fastest general purpose programming language” is not a credible claim . To be charitable, it’s best to assume the authors are not intentionally deceptive but just not yet literate about compiler performance trade-offs. They are upfront about staying distant from implementation details, generating millions of lines of code with “3 Claude Code Max subscriptions […] running at all times”.

By the way, a great introductory book on compiler design is Essentials of Compilation - An Incremental Approach in Racket by Jeremy Siek.

You may now be asking the same question I’m asking myself: Why am I wasting my time reviewing someone’s vibe coded compiler? Who even asked me to pay attention to this?

We were protested

Last month I had the honor of speaking at the first Software Should Work conference. It was an amazing two days with a wealth of knowledge on stage and off. Footage is on YouTube, you should check it out!

When we arrived, we were shocked and amused to find a pair of protesters across the street. I have to admit that if their goal was to get attention, I cannot resist taking the bait here. It’s a genuinely interesting thing that happened. Everyone I’ve told this story to has asked the same question:

Who would protest Software Should Work? Do they think software shouldn’t work?!

What an enigma! My first clue on the motivation was a cardboard sign saying “Stop Gatekeeping Programming ”.

Gates and their keepers

When we think of gatekeeping in software nowadays, it’s often part of the AI conversation: who should be programming and how, Vibe Coding and such. As far as that goes, I stand by my remarks from April 2023 in “Will AI Democratize Programming?” We have always sought to democratize programming and should continue to. There are many successes, including Excel, HyperCard, Visual Basic, and Python. We’re not done because making things simple is harder than it looks.

Who else gatekeeps programming? It’s worth remembering that some groups were historically marginalized or excluded from the field due to their identity - regardless of the fact that they learned the skills. We still have work to do in addressing that. Employment inherently has gatekeepers, and so does funding (see: Discrimination In the Venture Capital Industry). I had my doubts that the two white men waving signs were among those affected, but I was ready to hear them out.

“Who exactly is gatekeeping programming?”, I asked.

It turns out they were “protesting” being denied a speaker slot at the conference . They had submitted a lightning talk after the deadline and were upset that the organizer would not alter the schedule to add them (while continuing to deny the other late submitters, one assumes).

Lacking a speaker slot, they still wanted to promote their product and apparently didn’t see the organic conversations as enough opportunity. They then proceeded to publicly bash the event until they were asked not to attend, came anyway, and were ejected. (That sequence of events is not in dispute; I heard the same facts from the organizer and the...