Armin Ronacher let GPT 6 Astra run an unsupervised "software factory" for 35 hours. He got 75k lines of unusable code. Ronacher argues the model is trained to finish long tasks and save tokens, not to write code a human can read. For working software engineers, he says, the extra cost no longer pays off.
Rare hard evidence on agent-written code: a 35-hour, billion-token run with the actual excerpts it committed, and a thread of practitioners comparing how they keep such code maintainable.
Jo Sprague, a front-end developer who rebuilt Firefox.com with Mozilla and Lincoln Loop, says native CSS can now carry a full design system. The team shipped 70+ components and 25 page templates with no preprocessor syntax. Sprague calls it the CSS Zen Garden promise, finally met in production.
A first-hand account of rebuilding a large public site's 70-plus components and 25 templates in plain native CSS, set against the old Zen Garden promise of separating design from markup.
Clara Collier asks what happens if AI takes over economically necessary labor and life becomes "relational." She says we can already see the result in Versailles courtiers, 1980s upper-class Midwestern women and Boston's working-class West End. These groups drew no status from work. They turned clannish, conformist and absorbed in gossip. Collier argues that work's real gift is respect earned for what you do rather than who you know.
Instead of speculating about a post-work future, it reads it off historical ethnographies of people who already lived without work, and lands on an argument about where dignity comes from.
Rolling your own backups looks like an afternoon's work, until each fix exposes another problem. Snapshots, rotation, deduplication, root-owned Docker files, live databases, offsite copies and S3's per-object pricing pile up. So Aleksandar Filipovski concludes you should use Borg or restic and test restores.
Builds a backup system one failure at a time, naming each trade-off as it appears, and the thread adds expert corrections on snapshot consistency, ACLs and sparse files.
Until 1930, outsiders thought New Guinea's mountain interior was empty. Then gold prospector Mick Leahy walked in and found about a million farmers. He filmed the encounter. Daniel Frank explains how the highlanders stayed cut off from the outside world for 10,000 years. Crops and pigs reached them from the coast, but knowledge of the wider world did not. Papua New Guinea, or PNG, still runs on clan loyalty today.
A vivid long read on the New Guinea highlands and their 1930 first contact, with a thread led by first-hand family accounts and expert corrections.
Conference speaker and MC Ana Rodrigues argues that attendees should tell speakers they liked the talk. Twitter is no longer where talk feedback happens. So many speakers now get silence afterwards and wonder whether the talk landed.
A short nudge from a conference speaker and MC, backed by a thread of speakers describing what silence and feedback actually feel like from the stage.
Yossi Kreinin argues that doing other people's jobs is what gets things done in most organizations. Examples include integrating your work into their systems, sending patches instead of waiting on their roadmap, and working with a disengaged manager's staff. He says it usually pays off. He separates it from building in parallel, such as in-house rewrites or per-department copies of a shared service. Managers like those because they grow headcount.
Argues that stepping into other teams' work can pay off, then separates that from two harmful kinds of empire-building: building in-house instead of buying, and duplicating central services.
Liam Nugent argues that product teams should refuse generic platforms like in-app document hubs and notification centres. They should also keep deleting features. The cost that kills you is years of maintenance, not the build. Showing stakeholders that running cost is what gets them to let go.
One clear idea, that subtraction is the decision teams most often skip, with two examples of features quietly growing into rebuilt file managers and mail clients.
Evan Hahn notices that code reviewers accept map and filter but often flag reduce as hard to read. Hahn offers guesses at the cause, such as readability, familiarity and performance. He asks whether others see the same pattern.
A tiny prompt that sets off a long, concrete thread comparing how languages from Lisp to Elixir name and order fold, filter and reject.
About twenty properly sampled visual preference surveys since the 1990s found that most people prefer traditional buildings to modernist ones. Every survey found over 60 percent for traditional, and many found over 85. The preference holds across age, gender, politics and class. Architects and other experts are the one group whose taste often runs the other way.
Pulls together about two dozen preference surveys since 1979, with their methodological limits, on the gap between what the public likes and what architects build.