60 stories
Ioannou, an engineer at Bold.org, pulled his GitHub history and found his median weekday had gone from 3 or 4 contributions to 7. Days with no activity fell from 40 to 46 a year to 18 so far. He blames cheap starts: agents make every new project easy to begin and hard to put down.
The thread's sharpest reply came from commenters who think he misread the deal from the start.
The top HN threads: facts from the discussions, both sides, insider voices. Email, Telegram, or RSS.
The JIT, merged in 2024 without a standards-track PEP, was paused in June 2026 until one was accepted. Its maintainers drafted PEP 836, but the council wanted concrete support, performance and security plans, and objected to a compatibility review not due until August 2028. On October 19 the authors gave up.
Readers split between mourning and shrugging, and one pointed at a pattern.
Voytko lists what the old days actually looked like: blue screens from drivers, no autosave, screensaver passwords anyone could bypass, 20-minute boots, scratched CDs and corroded NES cartridges you could never patch. He still concedes that today's software is bloated and worries about delegating it to AI agents.
Commenters accepted the history but disputed one metric.
Every logical step of the classification, the theorem that every finite simple group is cyclic of prime order, alternating, of Lie type or one of 26 sporadic groups, was checked by Lean's kernel. The formalization has 2.05 million declarations and 100 milestones, and the team published an alternative route for the affected case.
Mathematicians in the thread were surprised the count was so low.
A pre-release NT 3.1 CD player linked on 16 November 1992 also runs, using the system's crtdll.dll. Earlier NT builds crash because they used the cdecl calling convention, where Windows 11 expects stdcall.
With the otvdm 16-bit emulator, even Windows 1.01's Reversi from 1985 launches, which the author admits feels like cheating.
The course improves a pre-trained Qwen3 0.6B model rather than training from scratch, runs early sections on a CPU and ships code in a GitHub repository.
Commenters split between readers of his LLM book who loved it and skeptics who asked who the course is for.
The test was a price tag on the claim that AI makes PHP exploitation trivial. Within hours strangers had topped the wallet up to $2,000 and the prize was still unclaimed.
One commenter had an LLM-generated proof of concept that crashed the site. A crash is not a shell.
His case rests on Michael Malis's pgrust, a Postgres rewrite that has already surfaced real Postgres bugs, including a money overflow bug.
The thread split on whether people are bored or just not enjoying it.
On the Silesia corpus, roc-deflate trails libdeflate by a median of 8.9% on a Ryzen 9 3900X and 2.6% on an M1 MacBook Pro.
The repository claims about 3.5x the speed of idiomatic C++ and 1.1x the best safe Rust on 16 of its own benchmarks. The README credits an AI model with compiler implementation, benchmarking and docs.
ClickBench's maintainer removed pgrust, a Rust reimplementation of Postgres, saying its entry measured "a lookup table of the benchmark, not a database." Probes with slightly altered queries were refused with errors.
pgrust's creator showed up in the thread to say the accusations are false. Commenters wanted to know why the paths were in the binary.
Alex Scott's post argues the three properties people credit to UUIDv7, monotonic order, a trustworthy creation time and unguessability, are not guaranteed by RFC 9562. Even Postgres warns that an extracted timestamp may differ from the creation time.
Commenters said that misses the real reason to use it: keeping new rows together in the index.
He follows the four-weekly PHP release cycle and compiles from source, so he wanted to warn people like him to update. The disclosure schedule changed after his post.
A commenter's rule: if you got the vulnerability details without the embargo, someone upstream broke it, not you.
The property is modest: no crash or failed assertion for any input under 64 MB. A typical proof is about 600,000 lines for a binary of roughly 18,000 instructions, and the process found hundreds of bugs, including one needing a ~20 MB input.
The proofs cover static musl builds on RISC-V, not the x86 and glibc most people run.
The author wrote earstrip, a C++ header with a WebAssembly build, adding a convex shortcut, strip output and branch-free counting to ear clipping. Native, it is about 14% faster than earcut.hpp.
Self-touching polygons fall back to Delaunay, as do those with 128 or more holes.
An essay at hrih.si says AI follows FORTRAN's pattern, only faster: a METR trial found experienced developers 19% slower with AI while believing they were faster, and Faros AI data showed more pull requests but much longer reviews.
Commenters attack the analogy itself. One says compilers were first called "automatic coders"; another says a compiler maps input to output roughly 1:1, and an LLM does not.
The fix, tracked as CVE-2026-111943, filters out chains containing the distrusted intermediates; backports were opened for Go 1.26 and 1.27.
Cloudflare's experimental Rust port of the TypeScript compiler claims up to 8x speed and about 70% less memory than TypeScript 7.0, with the code written by frontier models in days.
A TypeScript team member says much of the speed is not language choice, and some of it is already landing in TypeScript 7.1.
Paul Battley, a London developer, archived all his Ruby gems and wrote that he no longer wants to feed a commons that big companies strip-mine for machines. He points at Ruby's own core team, which he says now builds the language largely with LLM agents.
In the thread, the founder of an email company described a friend on an 80K-star project who is disillusioned too: the robots work, but the joy of collaborating is gone.
An Ask HN thread asked how engineers cope with shipping LLM-written code they haven't fully read. A Hasura engineer says his own system became fuzzy once he moved from IDE autocomplete to Claude and Codex CLIs, and argues the answer is formally verifiable code.
The loudest counter: you already ship code you haven't read. Principal engineers always have.
LWN calls Linux 7.3 a record-breaking release: 18,765 non-merge commits from 3,021 developers, 807 of them first-timers. The headline changes include a new fchroot() system call, per-CPU global data for BPF, and a finished sched_ext subscheduler.
There is also failfs, a filesystem that deliberately fails every operation, built for testing.
A parser here is a function from a string to a value and the unread remainder. Small ones chain into larger ones with pSeq, pMap and pRepeat, and the typed result is a tuple or object rather than a match array.
The finished example parses the header, per-packet lines and summary of a real ping run into one typed object.
1,111 changes carry Assisted-by tags, up from 301 in 7.1, and 252 commits credit the Sashiko review system with the report. LWN says the surge in Fixes tags looks like people using LLMs to hunt bugs in old code.
Its worry: if the flood of patches continues, the kernel could end up with code nobody fully understands.
Each compiler stage was compared against tsgo on TypeScript's own test corpus of roughly 12,500 files, and the project reports matching results on the tested cases. No benchmark results are published, only a script to produce them.
A commenter who tested it against the Go build confirmed it is faster than the Go version, but warned the code targets the subset of TypeScript that scriptc supports, so the result may be tuned to the toolchain.
Anthropic says merged PRs for Claude Code more than doubled from June to September while reverts fell by over 40%. A classifier now sorts changes into no-human, one-human or two-human review, and an agent's approval never counts as a human's.
Hacker News mostly doubted that throughput measures anything.
The Xena Project post names a 13-million-line AI formalization of Fermat's Last Theorem and a sphere-packing attempt as adding little lasting value, while Mathlib's 950 contributors and 65 reviewers keep the quality bar.
Contracts let a C++ function state what callers must guarantee and what it promises back. A violation can be ignored, observed, enforced or quick-enforced, chosen per build with no portable per-assertion switch, and the guide warns they are not input validation.
Walter Bright, who says he added contracts to Zortech C++ in the 1990s and to D later, wrote that they "just turned out to not be very useful". An embedded developer wants the opposite trade: checks on at the workstation, off on weak production CPUs.
The author used the shape and perspective clues in a photograph to narrow the world, then used CUDA to compare geometry across possible locations.
Its value is methodological: the answer matters less than making the chain of inference reproducible enough for someone else to challenge.
PostgreSQL can cover relational data, full-text search, queues, and extensions in one familiar operational surface, which is attractive to small teams.
The counterpoint is not that Postgres is weak. It is that a default becomes cargo cult when workload shape, scale, or failure modes have not been examined.
The complaint is not only stylistic. Readers said the prose makes it harder to see the actual decision, so they add banned-word lists and ask for plain-language explanations with examples.
The sharpest irony was an Anthropic reply reportedly written with Claude in a thread criticising recognisable Claude phrasing.
Forgejo and Codeberg kept coming up for people who want a lightweight forge they can host or use as a public service; GitLab was the answer for organizations that need a familiar enterprise-shaped platform.
The useful distinction was not hosted versus self-hosted. It was whether a team needs GitHub Actions and App integrations, or merely Git repositories with dependable CI.
The author linked the resulting [hp-laser-1008a-macos](https://github.com/Kuberwastaken/hp-laser-1008a-macos) project after describing the machine as a useful case study.
Other developers replied with similar work on Xbox adapters, e-paper screens, and industrial printers: specs, captured traffic, and prior art become enough material for a model to help bridge an unsupported platform.
Anthropic's multi-agent research treats coordination as its own engineering problem, not a capability that simply appears when each model gets stronger.
HN readers focused on the awkward practical consequence: clear roles and constrained communication may matter more than unleashing a roomful of equal agents.
Buf released a production-grade Language Server Protocol server for Protobuf, bundled with its CLI and available through its VS Code extension or editor configuration.
The real upgrade is incremental, query-driven analysis that can diagnose schema mistakes while developers work, rather than treating .proto files as second-class text.
The claim is concrete: a system can keep far more intermediate material active than a person. That makes it able to search many more paths before it commits.
HN's important turn was about verification. A result can be useful while becoming too intricate for a human to audit without structured checkpoints.
The experiment paired an agent with benchmarking, profiling, research, and correctness verification. That made optimization a constrained search problem.
HN readers reported that agents work best where a system supplies a hard answer. They fail more readily when success depends on credentials or a human-operated dashboard.
The leadership analogy points to a real workflow: define outcomes, expose constraints, and make decisions explicit before implementation starts.
The counterargument is equally useful. Agent work is specification, review, and feedback design, not people management with its consent and motivation work.
The essay warns against treating intellectual labor as interchangeable output. HN readers countered with supervised workflows that let them do work they otherwise could not.
Both views point to the same test: local task speed is observable, while organizational transformation must survive quality costs, oversight, and work omitted from benchmarks.
The essay follows characters that entered reference works through transcription errors and historical ambiguity. Once encoded, they become part of the compatibility burden that keeps old text readable.
HN readers described the conservative trade-off: extra invalid characters are often safer than deleting a character that later proves to be real.
The RISC-V debate is not about whether an open instruction set is valuable. It is about whether the base specification makes the right compromises for cheap controllers and broad implementation.
HN’s reply was strikingly practical: even critics who recognize the flaws still prefer an open ISA to a future built only on proprietary ones.
The preview supports Ubuntu 24.04 and 26.04 LTS, Debian 13, Fedora 43 and 44, on x64 and ARM64 through deb and rpm packages.
Early users immediately tested the boundary: Wayland input methods, downloads from another OS, and whether desktop projects can see existing CLI work.
DeepSeek Harness is a developer preview for coding-agent workflows, but its sparse initial material made its architecture clearer than its everyday use case.
HN readers were curious about first-party harnesses, yet the most repeated question was simpler: what is this tool for and why should a developer switch?
In today’s GitHub Trending: free-for-dev (a catalog of developer free tiers), OpenBB (an open data platform for analysis), awesome-go (a curated Go ecosystem), Playwright (cross-browser testing), and Web Dev for Beginners (a 24-lesson curriculum).
The unifying signal is reuse: each project shortens the path to a capability that would otherwise require searching, building infrastructure, or assembling a learning plan from scratch.
The essay argues that implementation is only one part of programming: requirements, constraints, integration, debugging, and responsibility do not disappear when generation gets faster.
In a 364-comment HN thread, the productive disagreement was not whether tools can write code, but which parts of software work remain hard after they do.
The comparison centers on power efficiency, not a single benchmark score. CPU, GPU, display, operating system, and application behavior all change the result a buyer experiences.
HN readers used the article to challenge simple architecture narratives and to ask which workloads still make the trade-off worthwhile.
A Universal Android Debloater issue reports that non-rooted Android 17 no longer allows ADB to uninstall certain system apps for a user.
The thread found a less dramatic workaround—disable rather than remove—while warning that carrier and vendor apps make the difference meaningful to privacy-minded owners.
In today's GitHub Trending: Svelte, a compiler-led web framework; Nuclei, a template-driven vulnerability scanner; Awesome, a directory of curated lists; System Design Primer, interview study material; and Summer2027-Internships, a daily-updated opportunities board.
The list favors tools that organize a developer's existing work—learning, finding opportunities, checking exposure and choosing components—over a single shiny new platform.
The release includes additions from animated WebP handling to new Vulkan and VideoToolbox acceleration paths. Each is a small capability until it becomes the missing piece in someone else's product.
HN readers celebrated the project as unusually durable open-source infrastructure: technical work that is visible mostly when it is absent.
JavaScript dependency trees turn a tiny convenience library into a path for a compromise. The practical task is not merely to remove one package, but to find where it entered the build.
HN treated the episode as another reason that package provenance and update discipline belong to ordinary engineering work.
The project turns a vague request for more inclusive visuals into a technical input developers can use when generating avatars, illustrations and examples.
HN's interest reflects a practical appetite for tools that make a design choice repeatable instead of requiring every team to rediscover it.
The announcement presents ten advances and says researchers helped prepare manuscripts and formalize proofs in Lean. That turns a model result into a claim about a research process, not just an answer.
HN’s sharpest question was what the release leaves uncounted: attempted problems, failed runs, repeated attempts, and the cost of the surrounding harness.
The proposed workflow does not reject coding assistants. It turns their output into a prompt: read it, type it, and notice each assumption before it enters the project.
Commenters agreed on the diagnosis more than the ritual: deliberate attention matters, while typing alone does not magically produce understanding.
The critique of Tailwind is not that CSS disappears; it is that layout decisions become a dense string of tokens beside the component they style. HN developers split between teams that value a shared configuration and those who find the markup harder to audit.
A retrospective on SwiftUI argues that declarative syntax did not erase the need to drop into UIKit or AppKit. The HN thread added a more practical split: solo developers praise the speed, while complex applications expose the framework’s seams.
F* is a general-purpose, proof-oriented language: code can carry specifications that a verifier checks. Its appeal is not replacing ordinary programming, but making high-assurance pieces of a system expressible in the same development workflow.
Kakehashi is an experimental userspace project for running macOS binaries on Linux ARM. The value is in the boundary it probes: compatibility layers reveal which parts of a platform are portable and which depend on its private operating-system contract.
The ripgrep issue reports crashes in musl-based binaries during very large searches. The thread quickly moved beyond a simple application bug and linked to an external analysis of the kernel behavior.
That shift matters because a fast command-line tool can be the first visible symptom of a defect beneath its own code, so reproductions and build details become evidence rather than noise.
NetBSD 11.0 is now available as a formal release. Users in the HN thread highlighted smooth upgrades from release candidates and the project’s detailed accounting of open issues.
The discussion’s real question was not whether NetBSD is fashionable as a desktop, but where a conservative Unix still earns trust: systems that value compatibility, documentation, and predictable maintenance.
Zig’s incremental path updates a debug binary in place instead of rebuilding every dependency after a small edit.
The discussion’s sharper point: semantic-analysis boundaries and language choices determine whether that shortcut stays tractable.
The source asks how the Bun rewrite is progressing; the thread answered with a more immediate story about AI doing narrowly scoped maintenance in the new codebase.
Some readers treated source-to-source rewriting as a familiar transformation. Others argued that passing integration tests is not enough evidence when an agent chose both the patch and its explanation.
