Top model scores may be skewed by Git history leaks in SWE-bench
github.com
Claude's memory architecture is the opposite of ChatGPT's
shloked.com
Bulletproof host Stark Industries evades EU sanctions
krebsonsecurity.com
Rails on SQLite: new ways to cause outages
andre.arko.net
NT OS Kernel Information Disclosure Vulnerability
crowdfense.com
Behind the scenes of Bun Install
bun.com
'Robber bees' invade apiarist's shop in attempted honey heist
cbc.ca
Unusual Capabilities of Nano Banana (Examples)
github.com
Launch HN: Ghostship (YC S25) – AI agents that find bugs in your web app
The Helix Text Editor (2024)
jonathan-frere.com
Making io_uring pervasive in QEMU [pdf]
vmsplice.net
Adam (YC W25) Is Hiring to Build the Future of CAD
ycombinator.com
Show HN: Making a cross-platform game in Go using WebRTC Datachannels
pion.ly
CRISPR offers new hope for treating diabetes
wired.com
AirPods live translation blocked for EU users with EU Apple accounts
macrumors.com
A tech-law measurement and analysis of event listeners for wiretapping
arxiv.org
Conway's Game of Life, but musical
hudsong.dev
Adjacency Matrix and std:mdspan, C++23
cppstories.com
ApeRAG: Production-ready GraphRAG with multi-modal indexing and K8s deployment
github.com
GrapheneOS and Forensic Extraction of Data (2024)
discuss.grapheneos.org
Randomly selecting points inside a triangle
johndcook.com
From burner phones to decks of cards: NYC teens adjusting to the smartphone ban
gothamist.com
An engineering history of the Manhattan Project
construction-physics.com
As somebody who is suffering from this precise issue at another org right now, and would very much like to solve it, one glaring omission from this article is whether this worked or not!
That is, did the "inverse Conway" actually result in getting the unified observability platform shipped, and that platform actually solving the problem they initially were trying to solve? It certainly seems plausible that it COULD, but quite surprising that that's not actually stated in the article if so...