The Shopify CEO’s Pull Request Will Never Be Merged Nor Closed

The CEO of Shopify, Tobi Lütke, recently
made headlines
by using the AI tool
autoresearch to
increase Liquid parsing
speed by 53%.

Turns out this was a case of
“CEO Said A Thing!” journalism
where no one checked the primary source and all nuance was lost.

Some articles imply the code actually shipped
and, well, works. Both false.

I’ve dabbled in Liquid parsing myself,
so I was compelled to read the code.
Here’s Tobi’s actual pull request.
It makes the code significantly less readable, partially because
optimizations tend to be confusing, but mostly because the code
quality is just bad. See
lib/liquid/variable.rb
if you enjoy the feeling of your eyes glazing over.

On top of that, tests are failing.

To be fair, not that many are failing: only
3 out of 4,192
specs. Tricky ones, though. The remaining failures are code
quality related, like too many layers of nesting.

If I maintained Liquid, I wouldn’t bother cleaning this up
because it seems unsalvageable. But I also wouldn’t have the
guts to close a PR opened by the CEO. Therefore, I predict it
will stay open forever.

Disclaimer

This post is not intended as a criticism of Tobi. I think
autoresearch is a cool idea and I like the fact that AI coding
tools are giving more people the opportunity to code. I am
critiquing the hype-based news coverage of this event and the
public’s lack of interest in primary sources.

TL;DR

If most of your knowledge comes from headlines, you are less
informed than you think.