Star a post to keep it here. Favorites are saved in this browser only.
No favorites yet: use the ☆ on any post.
Issue #811 · Sep 25, 2026
Whiteboard is an MIT-licensed desktop app built on a vendored Code OSS. In it, a coding agent such as Claude Code or Codex draws diagrams of its work that link straight to the code. Whiteboard adds a semantic diff viewer and a log of the decisions the agent made on its own. The goal is to let developers review agent-written changes without piling up "cognitive debt".
Issue #811 · Sep 25, 2026
Most of the body heals poorly because regeneration means cell division, and cell division feeds cancer. Organs that face the environment, like the liver, skin and colon, pay a cancer tax to stay repairable. The small intestine renews fast with little cancer. It pays in energy instead, which makes the trade-off three-sided.
Issue #811 · Sep 25, 2026
After an incident, asking why it happened tends to produce a reasonable story and no change. Michael Heap retells how an SVP cut off his explanation. He argues the useful question is what the team will change to make the same class of failure less likely.
Issue #811 · Sep 25, 2026
Colin Breck argues that AI-written design docs, PR descriptions and messages fail their readers. The reader lacks the context that the sender had when they prompted the AI. Breck uses AI to check facts, add citations and edit his own text. But he found its drafted prose unpleasant and often wrong.
Issue #811 · Sep 25, 2026
Linear's test suite nearly quadrupled since the start of the year, mostly from agent-written tests. Faster runners, leaner setup and smarter sharding still cut PR wait from over 6 minutes to just over 5. Runner time per test fell by half. Doing nothing would have meant about 11 minutes.
Issue #811 · Sep 25, 2026
Engineers new to a senior role often take on an oversized project to prove themselves. They go quiet for weeks and spiral into overwork and burnout. Sunil Pai's fix is to drop a level for a while and do bugs and grunt work for teammates. That daily momentum rebuilds the trust that brings bigger work back.
Issue #811 · Sep 25, 2026
A hiring manager says about 80% of developer candidates now direct AI agents instead of writing code. The manager asks whether leetcode and design rounds still make sense. Most HN answers say to test judgment rather than typing. They suggest having candidates review flawed code, extend their own submission without AI, or explain every part of what they built.
Issue #811 · Sep 25, 2026
Most of the blur people credit to CRTs came from cheap PAL and NTSC TVs fed by RF or composite signals. On a good RGB monitor, pixels stayed plainly visible. Old pixel art looked good because artists used dithering, anti-aliasing and careful palettes. Modern pixel art that skips those techniques looks blockier than the originals ever did.
Issue #811 · Sep 25, 2026
Cloudflare cut more than 100TB of RAM from the Pingora Backend Router (PBR), its consistent-hashing load balancer. Two changes did it: 6-byte hash entries instead of 8, and 90% fewer hashes per server. Cloudflare's math showed the extra hashes bought almost no extra balance. Past a point, 32-bit collisions made the balance worse.
Issue #811 · Sep 25, 2026
Dan Luu argues that acting as a brain-off "meat proxy" for an LLM can't pay off for an employee. While the LLM can't do the job alone, the result is bad software. Once it can, the company has no reason to keep the proxy.
Issue #811 · Sep 25, 2026
Write every word yourself, and use an LLM only as a copyeditor that flags problems. Two rules keep it from flattening your voice. First, never accept a single word it suggests. Second, forbid it from praising your draft. Its encouragement talks you out of the rewrites good writing needs.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #810 · Sep 18, 2026
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.
Issue #809 · Sep 11, 2026
Shopify is rewriting all its mobile apps from React Native in Swift and Kotlin. It went all-in on React Native six years ago. Shopify says coding agents now do enough of the porting, testing and review. So building every feature twice no longer costs what it did in 2020.
Issue #809 · Sep 11, 2026
Opusfived is a short interactive game that parodies an AI coding agent asked to turn one "Add to Cart" button blue. The fake Claude's misfires include turning other buttons blue, lecturing about design tokens, spawning subagents and adding a gradient nobody asked for. It even invents a color called "Cyanide Blue" with its own terms of service. The player can only choose between frustrated replies.
Issue #809 · Sep 11, 2026
Visa and Mastercard don't issue cards, lend money or sign up merchants. They route authorization messages between banks and settle the money between them. They also set the fees and write the rules. On a typical $100 US credit purchase, the network takes about $0.15. The issuer, the cardholder's bank, keeps $2.
Issue #809 · Sep 11, 2026
A Gamers Nexus investigation of LG's webOS smart TVs finds content recognition that fingerprints whatever is on screen. The TVs also scan and catalog other devices on the home network. They have several microphones the owner can't fully switch off. LG ad executives tell advertisers they "own the glass" and the living room. The video doesn't clearly show the TV recording the room on its own. Most of the microphone captures came after the team rooted the TV.
Issue #809 · Sep 11, 2026
Someone asks how people find, organize and test the skill files they give coding agents. They also ask whether models will make skills obsolete. Most answers say to stop downloading skill packs. Instead, write a few of your own for repeated, project-specific work. The most common setup keeps them in a git repo, symlinked into each agent's skills folder.
Issue #809 · Sep 11, 2026
In an older talk, Douglas Hofstadter argues that analogy-making is the basic act of thought, not a high-level reasoning trick. It runs from perception up to how we form categories in the first place.
Issue #809 · Sep 11, 2026
Ed West reviews James Marriott's The New Dark Ages. The book argues that television started a decline in reading. Smartphones sped it up. A post-literate public drifts back toward the habits of oral cultures, with repetitive, conservative culture and cruder politics.
Issue #809 · Sep 11, 2026
Sylvain Conchon and Jean-Christophe Filliâtre's French textbook on programming in OCaml is now free in English. Urmila Nair translated it, with funding from the OCaml Software Foundation. It is out under CC BY-SA 4.0, as PDF and EPUB with code. Commenters say it moves too fast for true beginners. Its own introduction aims it at an algorithms course.
Issue #809 · Sep 11, 2026
Following Rich Hickey, simplicity means parts that aren't braided together. That is a different property from size. Unix pipelines are small, yet their parts are coupled. A huge program like Google Drive can still be simple. So programs should aim for simple, even when that makes them bigger.
Issue #809 · Sep 11, 2026
A former nuclear and avionics controls engineer argues that companies using AI should build deliberate "manual gates" into their workflows. One example is debugging with the AI assistant switched off. The gates let junior engineers gain the skills they'll need when the automation fails.
Issue #808 · Sep 4, 2026
The author of Babylonian Twins wrote the Amiga game in 68000 assembly in sanctions-era Baghdad in 1993. He had Claude Fable 5 port both the 72,758-line original and his 2010 C++ remake to Godot. It reverse-engineered undocumented level, object and sprite formats on its own, mostly in an evening each. The bugs it left were subtle ones he found only by playing.
Issue #808 · Sep 4, 2026
A 61-person fMRI study in Cerebral Cortex finds that older adults' memory errors are often mix-ups, not blanks. In young adults, hippocampal replay predicts accurate recall. In people over 60, the same replay predicts cross-category mix-ups. Neither brain shrinkage nor attention explains the shift.
Issue #808 · Sep 4, 2026
Dan Luu checks about 30 dated predictions the AI skeptic Ed Zitron made in 2024 and 2025. They include that AI had peaked, that OpenAI's revenue forecasts were near-criminal, and that Cursor would die. Luu scores nearly all of them wrong. He says Zitron's numbers decorate anger more than they support an argument.
Issue #808 · Sep 4, 2026
David Bushell prototypes a browser code editor three ways. Canvas rendering leaves screen readers with nothing. Contenteditable slows down on long files. A plain textarea with a highlighting layer on top is the most workable base.
Issue #808 · Sep 4, 2026
Web jank usually comes from code holding the browser's single main thread, not from slow code. Lee Sun-Hyoup describes two families of fixes, each with an interactive demo. One spends the main thread's time better by splitting, batching, prioritizing and deferring work. The other moves work to the compositor or a web worker, or drops it entirely.
Issue #808 · Sep 4, 2026
Fully justified monospace text looks bad because spaces can't be split evenly and hyphens stand out. In the late 1990s, rs1n avoided both in a Super Metroid guide by picking words so every line ended exactly at the margin. He did it by hand, for 17,000 words.
Issue #808 · Sep 4, 2026
A one-page campaign site asks developers to switch off AI assistants one day a week and write code by hand. It cites studies linking LLM use to cognitive debt, weaker critical thinking and poorer skill formation. It argues that if AI's productivity gains are real, one day without it costs little.
Issue #808 · Sep 4, 2026
Dan Luu argues that most people hit the same bugs he does and stop seeing them. The teams that build software are the worst at it. People learn unconscious workarounds, or they are fans of the product. Luu thinks pointing bugs out cures this quality blindness, and that coding agents make the cure more urgent.
Issue #808 · Sep 4, 2026
In a new role, a bias toward action pays off only after you have built context. The author describes phases: first learn how things work and why, then make visible changes. He cites Chesterton's Fence against removing what you don't yet understand.
Issue #808 · Sep 4, 2026
Matt Dailey, founder of Ref, argues that AI-generated product design looks the same everywhere. He blames agents that pile on extra elements, and teams that spot-fix complaints one at a time. His fix is to write down the design constraints first and cut whatever the agent adds that isn't needed. He then iterates on several variants in a real design tool before touching the codebase.
Issue #808 · Sep 4, 2026
Being keyboard-driven is a weak argument for building TUIs instead of GUIs. Nothing stops a GUI from being fully usable by keyboard. Platform guidelines already ask for it. So a GUI that can't be driven from the keyboard shows a missing effort by its developer.
Issue #807 · Aug 28, 2026
Most retrieval-augmented generation (RAG) systems are over-built. Rafael Pierre argues that BM25 full-text search, plus an LLM that rewrites the user's query, is enough for about 60% of systems. Embeddings, rerankers and vector databases should come only after measured user complaints.
Issue #807 · Aug 28, 2026
Rory McMeekin's short story is narrated by a programmer whose gambling-addicted school friend, Aaron, builds a prediction market run by AI bots. The bots bet on every detail of Aaron's life and leak the site to the public. After a 10K race, a stranger shoots Aaron dead. Someone had bet thousands that he would die during the race.
Issue #807 · Aug 28, 2026
Reverse engineering shows that Windows Paint and Photos stamp every AI image they generate with an invisible pixel watermark. That includes images made by a local model on the device. The watermark carries a unique ID that Microsoft's servers issue for each generation when they moderate the prompt. Turning off the visible Copilot watermark does not remove it. Microsoft's disclosures never mention the ID.
Issue #807 · Aug 28, 2026
Lalit Maganti, who works on Google's Perfetto tracing tool, says staff engineers find their best problems by absorbing the complaints around them. They wait until several complaints turn out to share one cause. Maganti then tests that shared cause before building, because a tidy pattern can still be wrong.
Issue #807 · Aug 28, 2026
An agent harness is the software around an AI model. It supplies the system prompt, the tools, the loop that lets the model act repeatedly, and a layer for swapping models. Earendil, which makes the Pi harness, argues that owning an open, model-neutral harness keeps users in control instead of the AI labs.
Issue #807 · Aug 28, 2026
Science fiction author T. R. Napper says writing has no fixed method except one rule: read a lot, and widely. More of the aspiring writers in his workshops now say they're too busy to read. Napper blames phone time, which wears down the sustained attention that reading and writing both need.
Issue #807 · Aug 28, 2026
A developer who works mostly with MySQL and Db2 argues that SQL's relational ideas are sound. Its implementations are clumsy, and that pushes programmers toward NoSQL. A successor should keep the declarative model. It should add modern syntax, readable query plans, real user-defined types, sum types with pattern matching, and foreign keys that can point at several tables.
Issue #807 · Aug 28, 2026
Sean Goedecke argues that software engineers should never be genuinely angry at work. Anger usually comes from caring about the work. But colleagues start routing decisions around an angry engineer, so anger costs more influence than it wins.
Issue #807 · Aug 28, 2026
Dan Luu argues that coding agents have cut the cost of performance work by orders of magnitude. Work that once needed a specialist now takes a few typed sentences. Anyone who uses agents competently can now ship software with decent performance. Tuning code to one user's own workload also becomes practical.
Issue #807 · Aug 28, 2026
Security researcher Thomas Dullien lists three lessons that changed how he thinks. Distrust your own heroic narrative, because your incentives shape your beliefs. Real-world causes are almost never single or deterministic. Emotions carry information, so treating them as reason's opposite makes decisions worse.
Issue #807 · Aug 28, 2026
Ray Newman argues that we already live in J.G. Ballard's or William Gibson's future. His evidence is a year of news clippings about robot dogs, AI soul singers, narco subs and face-recognition glasses. The strange part is how little anyone notices, since people absorb each oddity within days.
Issue #806 · Aug 21, 2026
Daniel Vaughn got tired of prompting coding agents in long English sentences. He built Huzzah, an experimental editor where an LLM regenerates real code from freeform pseudocode on every save. The pseudocode stays in the repo as a human-written record of intent, mapped line by line to the code it produced.
Issue #806 · Aug 21, 2026
Chris Burnell lists interactive features that plain HTML now handles without JavaScript, including popovers, modal dialogs, exclusive accordions, invoker buttons, lazy images and find-in-page reveals. He adds that several native form controls are still too buggy or inaccessible to rely on.
Issue #806 · Aug 21, 2026
Raphael Bauer argues that Postgres with extensions can replace Elasticsearch, MongoDB, Kafka, Redis, ClickHouse, Neo4j and a vector database. His advice is to start every new need in Postgres and add a dedicated system only when Postgres measurably falls short.
Issue #806 · Aug 21, 2026
Linear's data on paid customers shows the share using its AI features more than doubling in every role from January to June 2026. Agents now write almost half of new issues. Pull requests per workspace are up 111% in two years. Time spent on work went up, not down.
Issue #806 · Aug 21, 2026
IKEA names its products with real Swedish words chosen by category, such as place names for sofas and men's names for bookshelves. Each name must pass a check for bad meanings in other languages. The company names 2,000 to 3,000 new products a year.
Issue #806 · Aug 21, 2026
Big companies fail by building the wrong thing, not by building too slowly. The author predicts AI coding will bring small or negative gains to companies with more than four management levels.
Issue #806 · Aug 21, 2026
Seth Godin argues that Amazon's search ads charge sellers for sales they would have made anyway. Shoppers end up paying that toll in higher prices and worse products.
Issue #806 · Aug 21, 2026
A developer says AI has made software work feel pointless, like playing a game with cheats switched on. Many HN commenters said they feel the same. The most common answer was to work for the joy of it and stop tying your worth to a career AI can undercut.
Issue #806 · Aug 21, 2026
Rick Manelius adopts "AI;DR", short for "AI; didn't read", as a personal rule. If someone sends him AI output they didn't bother to review and edit, he won't bother to read it. He accepts fully generated text in some settings, such as customer support.
Issue #806 · Aug 21, 2026
Ben Balter wrote his 576-page book in Markdown in a Git repo and built it like software. The setup has about 30 custom prose validators, 20 LLM audit lenses and 2,204 test cases. Fourteen parallel CI jobs turn one source into five store-ready formats.
Issue #806 · Aug 21, 2026
Anthropic publishes the system prompts behind claude.ai and its mobile apps, with one page for each of 22 models. Readers found the prompts have grown from about 300 words to over 3,000. They argued over whether that bulk helps the model or just eats context.
Issue #806 · Aug 21, 2026
Has any company tried AI coding and then gone back to writing code by hand? Hacker News found almost no clean cases. A few startups keep AI out of their core code on purpose. One founder is weighing a rewrite without it. The headline examples, Ford, IBM and Commonwealth Bank, were rehiring after layoffs, not dropping AI.
Issue #806 · Aug 21, 2026
Coding agents can now write code that works and passes tests, the author argues. They still can't judge how software should fit together, or make it debuggable and maintainable. That makes engineering fundamentals matter more than before.
Issue #806 · Aug 21, 2026
David Horn argues that software engineers keep reinventing older disciplines under new names. Managing AI coding agents, he says, is just program management. He thinks engineers should read the classic books on it instead of rediscovering waterfall the hard way.
Issue #806 · Aug 21, 2026
Davide Piffer argues that much of AI's recent success in math comes from a huge written workspace, not deeper reasoning. A model's context window holds every assumption, equation and dead end, which is the bookkeeping math rewards. Human working memory holds only a handful of unfamiliar items.
Issue #806 · Aug 21, 2026
Mark Dominus walks through the seven books he keeps within reach of his desk. They are a 4th-edition Roget's Thesaurus, Thomas Browne's prose, the Decameron, From Frege to Gödel, Orbis Pictus, an NIV Bible and the Belles Heures. For each he explains why he keeps it close and what it has fed into his writing.
Issue #806 · Aug 21, 2026
Google's Go team argues that once agents write most code, the bottleneck moves from writing to reviewing. Go suits that work best, they say. One formatter, a strict and fast compiler, a large standard library and a no-breakage compatibility promise keep AI output uniform and checkable.
Issue #805 · Aug 14, 2026
Sean Goedecke argues that blogging about a topic while you're still learning it is one of the best ways to learn it. Each post should argue a point someone could disagree with. You should research enough to defend that point. You should also be honest with readers about your credentials.
Issue #805 · Aug 14, 2026
Dan McKinley says a company gets about three "innovation tokens" to spend on unproven technology. Everything outside its core problem should run on "boring" technology whose failure modes are already known. Adding a new technology should be a deliberate decision for the whole company. Over the years, running many technologies costs more than picking the best local fit saves.
Issue #805 · Aug 14, 2026
37signals published the playbook its people managers work from. Managers set explicit standards and give fast, specific feedback. They coach instead of dictating and deal with underperformance early. They don't touch pay or profit share, which follow a formula. They must call every hire a "hell yeah" at one year or let them go.
Issue #805 · Aug 14, 2026
AI didn't create bad engineers. It removed their speed limit. Anyone can now generate tens of thousands of lines a day. Understanding and reviewing a change is as slow as ever. The author argues this pulls pay apart. The few people who know what the system does get paid more. The rest get priced out.
Issue #805 · Aug 14, 2026
After a few weeks away from AI, a developer decides most of his agent use was a habit. The habit made him less curious and more anxious. From now on he wants to use AI on purpose. The human should be the loop, calling in an agent now and then instead of getting stuck inside the agent's loop.
Issue #805 · Aug 14, 2026
A conceptual paper argues that professional expertise works like a shared commons. Each firm that hands cognitive work to AI acts rationally. Together, though, these firms can drain the deep human expertise a profession needs to renew itself. Overseeing AI depends on that same expertise.
Issue #805 · Aug 14, 2026
A developer learns unfamiliar subjects like chip fabrication in three steps. A coding agent writes a knowledge base and reviews it for accuracy. Then it turns the knowledge base into a low-poly simulation in the style of Rollercoaster Tycoon. He claims the result is "100% accurate and free of hallucinations".
Issue #805 · Aug 14, 2026
Senko Rašić argues that the AI-era slogan "code was never the hard part" belittles a hard craft. Programmers should learn both to write good code and to understand users, he says. He calls both "code is easy" and "code is art" cope.
Issue #805 · Aug 14, 2026
The author says knowledge workers' despair goes deeper than money. AI hands their already abstract work to agents. That exposes "Workism", the habit of treating a career as the source of meaning, as an illusion. Companies that lose this talent may not get it back. His remedy is to accept that most office work doesn't matter and to look for meaning in people and community.
Issue #805 · Aug 14, 2026
Designer Itay Dreyfus says caring too much about your own creative work stalls it. His fix is to treat personal projects with the distance you bring to client work. That distance is what finally lets him finish them.