The weekly Vivarium for Week 8

What Sports Ball Can Teach Us About LLMs

What fresh horrors await each new day in this era of LLMs? Recently, it was an article by folks at StrongDM who claim that you don't even need to look at the code LLMs write, you can just check the "acceptance tests" (the article uses some words that sound like "we checked it did a thing" so I'm paraphrasing).

We'd never suggest that even the smartest human engineer's software should just be trusted without ever looking at the code, right? I mean, is there a respectable software house out there that will stand on "fuck code reviews, we shipped it and the users liked it"? If so, please do get in touch.

Alternatively, do you think those folks have ever heard of lineage-driven fault injection? The number of ways something can go wrong is... let's just say that number is very very very big.

"Works on my machine..."

Morgan Stanley published some thoughts on AI investment.

There've been a lot of ideas that have been rolling around in my head like rocks in an empty drum for a few weeks, so I've jotted them down for your amusement or whatever:

Is superintelligence possible (beyond the obvious cases like Musk and Altman): In Which of 4 Universes Do We Live?
Why do we evoporte small lakes and try to un-moth-ball nuclear reactors, again? LLMs are Cancer, Big Data Centers are Tumors
At the risk of being rude: LLMs Are Not Even Stupid, But Humans Sure Are
The modern world is run by "intermediaries" that would make even the brightest and biggest Mafia families blush a little: The Mass Disintermediation of Modern Life
We all know humans just want to believe what they already believe, but this could get awkward: Did an LLM just Predict its own Demise?
A little tale of "what if..." The LLM Grift Playbook

Since it was the Superbowl on Sunday, I hope your sports ball team won. It got me thinking, why don't we have leaderboards and contests for LLMs like we do for all manner of sports ball sports and sports without balls?

So I cooked something up: LLM One Shots Club: Show your best one(-or-more) shot LLM creations to the world and bask in fame and fortune.

News Around Your Towns

Shane and I have been running the Better Futures Club meetups since Jan, 15, 2026 every week, alternating Tuesdays and Thursdays at PDX Hackerspace and it's been great to hear from the small number of people who have shown up what they're thinking and what they're interested in.

Come hang out or consider getting together in your neck of the woods and letting us know how that goes.

Reader's Corner

If you read one thing this week... The Singularity will Occur on a Tuesday.

We’ll send occasional updates about new docs and platform changes.

How's it Tracking?

I've spent most of January getting up to speed on the Tenstorrent architecture and software with the aim of using it with projects like this deterministic computing system and red-candle.

If you've been following along (or not), there's quite a big world of artificial intelligence work out there that has nothing to do with LLMs.

London Calling...

We need painters, writers, philosophers, ethicists, contrarians, sculptors, designers, engineers, chefs, sewers, knitters, crafters, drawers, illustrators, and their friends and more to work on the problems in the world. There is nothing other than humans that can solve these problems right now, and that's going to continue to be true for a while at least, so tell your friends.

Looking Forward to Next Week

If no one even needs to look at the code anymore, do programming languages matter? This might be something we want to mull over a bit.

If you haven't noticed, the world of WebAssembly has some interesting new facets that could be pretty useful in helping define an ecosystem of interoperability, just in case programming languages do still matter:

References

Feb 8, 2026