2026W25
A few interesting articles I read over the past few days
This is the output of an automated process. Every Sunday, a script retrieves articles I've saved and read, uses AI to expand my quick notes into something more coherent, then publishes them. This post is one of those articles.
- There Are No Instances in atproto — overreacted — The RSS analogy is what makes the whole thing click. Counting Bluesky instances misses the point the same way counting RSS readers would have missed the point of blogs. What matters is whether you can swap your hosting without losing your identity, and whether new applications can build on top independently. Dan literally moved his hosting to a different provider the same day he wrote this, which makes the argument more concrete than any architecture diagram could.
- What the Fuck Happened to Nerds — The core claim is that tech leaders are liquidating 40 years of accumulated “boring nerd” credibility into celebrity, and the exchange rate will catch up with them. The Founders Fund Mafia video is the perfect exhibit: billionaires with weapons contracts and a line to the White House playing a deception game on camera, produced like reality TV. The author’s right that if any of them faces scandal, the footage of them being “good at hiding how good they are at deception” writes the headline itself.
- Adaptive compression codec — A 96.9% storage reduction on sequential integers is striking, but the design philosophy matters more than the numbers. The system tests compression candidates on sample blocks and picks the winner per column, so users never choose between LZ4 and ZSTD. The idea that “the system should test, measure, and adapt where it has enough evidence” extends well beyond compression — if the database can figure out the right codec, why are we still manually picking primary keys?
- How to Earn a Billion Dollars — The math is the easy part — 15% monthly growth for five years turns $10k/month into $526M/year. What makes Graham’s version of this argument interesting is the second half: the best startup ideas sound terrible initially, so deliberately searching for them filters out the winners before you even start. Airbnb got funded despite YC thinking the idea was bad. Not sure I fully buy “just build cool stuff with friends” as universal advice, but the filter problem is real — if an idea sounds obviously good, someone with more resources is already doing it.
- Leaving Mozilla — The diner analogy carries the whole piece. Firefox users are people who walked past every McDonald’s to find the mom-and-pop place, and leadership keeps responding by trying to serve Big Macs. After 15+ years, JR Conlin’s frustration reads less like bitterness and more like watching something you helped build get managed by people who don’t understand why it worked. The question “Who am I doing this for?” hits hard when the answer he kept arriving at was “so that someone else can get a gold star on their resume for the next gig.”
- Introducing RawTree — The “schema before data” problem is real — I’ve worked on enough projects where the schema was wrong by the time the first real data showed up. The dynamic type system handling mixed types in GROUP BY and aggregations is the technical detail that separates this from “just throw JSON in a column.” Whether automatic primary key generation and self-materializing projections hold up at scale is the open question, but the premise that the system should observe actual queries and optimize accordingly feels like the right direction.