I skipped a couple of months, and now I'm covering a whole summer in one go. Three months is too long for the usual link pile, so with this edition, I'll go long pieces — reads and watches for your weekends 🧘♂️
Compression Is the Whole Game
The most quietly important thing in our #feed was a pair of items nobody made a fuss about. It deserved one. Giorgi Gagoshidze shared ngrok's Compression is prediction in August. The idea is simple and slightly vertiginous: compression and LLMs are virtually solving the exact same problem — predicting what comes next — and better prediction is better shrinkage.
The long-watch half of the same idea is Reinventing Entropy — Compression & Intelligence Part 1, which I dropped into the channel and which builds the entropy story from the ground up. Watch it after the read.
The Maximalist Position
Steve Yegge's The Shape of Things to Come — the maximalist position: loops and graphs, the end of human code review, and the land rush that follows.
Yegge actually got a double billing this summer: I shared his The Flat Curve Society myself, with a flat recommendation to just read it. Two ideas. First, that after "Fable" we get maybe one more iteration before nuclear-weapons-grade control descends on the frontier models — and for the rest of us, the curve flattens artificially right about where we're standing now :) Second, that even if they handed us the frontier models anyway, for 99% of tasks we couldn't tell the difference — because we can no longer pose problems hard enough that we'd still be able to verify the answers without AI. And it's full of other interesting things beyond those two 🙌
The Calm Rebuttal
Around the same axis, and from the opposite temperament: antirez wrote two essays this summer that Giorgi Gagoshidze surfaced — Control the ideas, not the code and Not just development, distribution of software may change as well. Where Yegge announces a thunderdome, antirez calmly relocates the thing worth owning: not the code, the ideas — and maybe not even the software, but how it reaches people. And for the long watch, I shared How AI will change software engineering – with Martin Fowler, the calmest voice in the room on the same question.
Three temperaments, one question. Nobody's pretending the answer is in.
Also on Our Radar
The long pieces that didn't make the main sections but deserve a slot on your shelf — no short links this time, only things that take an evening.
- Inference Engineering — Baseten's book, free to download. For anyone tired of shepherding agents and ready to look under the floor.
- DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux — Lex Fridman #501 — several hours of DHH being DHH. You already know whether you want this.
- I was laid off by Atlassian — Temuri Takalandze's find: a developer let go after eight years recorded forty minutes on what he'd actually been doing there. Quieter than the usual layoff content.
- Mechanical Watch — Bartosz Ciechanowski's interactive essay, surfaced by Giorgi Gagoshidze. The gold standard for explaining a machine on a web page.
- Old and new apps, via modern coding agents — Terry Tao on what agents did for his own tooling. When the strongest living mathematician writes up his workflow, you read it.
This is what Omedians actually read and watched at length in June–August 2026. Not everything that happened. Just what made it through our prism.