Weekly reading: 2026-06-01
Intro
This was a thinner reading week, but the saved links still fit together surprisingly well. The strongest thread was about operating surfaces around AI: what makes coding agents actually useful inside an organization, and what happens when trust in an AI platform starts to fray.
There was also one good lighter link worth keeping in the mix, a reminder that not every useful reading roundup item has to be breaking news. Sometimes an old classic is still the right thing to resurface.
Agent infrastructure and shared operating context
1) Under the River
https://shopify.engineering/under-the-river
Shopify’s engineering write-up on River and the Aquifer platform underneath it is the clearest item of the week. The core claim is that useful coding agents depend on substrate quality first: monorepo access, reproducible environments, written-down skills, and a clean separation between durable session state, harnesses, and disposable sandboxes. The especially interesting organizational choice is making agent work happen in public Slack threads so the transcript becomes shared memory instead of private local context.
Why it matters: a strong example of “agent-friendly infrastructure” really meaning “human-friendly infrastructure with better memory, visibility, and reproducibility.”
Platform backlash and AI ecosystem trust
2) Everything That Went Wrong With Claude
This is a satirical anti-Claude timeline, surfaced via Peter Steinberger, but it is useful as more than just a joke. It compiles a broad set of backlash narratives around Anthropic and Claude, pricing changes, outages, bans, legal disputes, policy reversals, quality complaints, and anti-competitive optics, into one artifact.
Why it matters: even when exaggerated for effect, sentiment indexes like this are a useful signal. They show where trust is breaking, what users remember, and how quickly platform risk becomes part of the product story.
Old internet classics worth revisiting
3) A Brief, Incomplete, and Mostly Wrong History of Programming Languages
https://james-iry.blogspot.com/2009/05/brief-incomplete-and-mostly-wrong.html
Shriram Krishnamurthi resurfaced James Iry’s classic programming-languages satire, and it still holds up. It speed-runs decades of PL history through jokes about FORTRAN, Lisp, C, Smalltalk, Haskell, Python, Java, Scala, and more.
Why it matters: a good reminder that engineer culture has its own enduring canon, and that older essays can still be more memorable than most new discourse.
Closing note
My short version of the week: the AI story keeps moving away from raw generation and toward trust, memory, substrate quality, and institutional shape. And when the week is light, it is still worth carrying one timeless piece that gives the whole roundup some texture.