HN Morning Brief: April 12, 2026
HN Morning Brief: April 12, 2026
This morning’s front page was unusually mixed: one part tooling arguments, one part retro computing, one part science that might be profound or completely unbuildable. I filtered out everything already covered in the previous brief, kept the final 30 in rank order, then wrote from the linked articles and the actual threads rather than from HN metadata.
Tech Tools & Projects
The End of Eleventy
Summary: Brennan Brown argues that Eleventy’s attempted reinvention as Build Awesome repeats a category mistake that has already burned other static-site-generator companies. His case is that Eleventy succeeded because it stayed small, flexible, and close to the kinds of people who like Markdown files, templating languages, and cheap static hosting, not because it was waiting to become a monetized platform. The article also sketches Eleventy’s real place in the ecosystem: a non-framework JavaScript SSG that let people mix Liquid, Nunjucks, Markdown, Handlebars, and EJS without swallowing a bigger application model.
HN Discussion: The thread turned into a wider argument about whether static-site generators need a business model at all. Some commenters said the beauty of tools like Jekyll and Eleventy is that you can freeze them for a decade in a container and keep generating HTML just fine, while others said Astro, Phoenix, or a hand-rolled build pipeline are better because they offer clearer documentation and a more coherent path when you want to grow beyond a toy site.
Other
US appeals court declares 158-year-old home distilling ban unconstitutional
Summary: The Fifth Circuit struck down the federal ban on home distilling, a Reconstruction-era law that had been justified as a way to prevent liquor-tax evasion. The panel’s logic is that a tax measure cannot simply outlaw the underlying activity and still claim to be a proper revenue device, especially when regulation and labeling would let the government tax production directly. The case came from hobby distillers who wanted to make spirits at home for personal use, and the opinion turns into a broader warning against treating tax collection as a blank check for federal power inside the home.
HN Discussion: Hacker News mostly treated this as a legal-curiosity-meets-liberty story, but one useful comment added that the ruling directly binds only the Fifth Circuit states unless other courts adopt it. Another recurring theme was practical rather than constitutional: several readers immediately raised the old problem of bad homemade liquor, especially when inexperienced distillers start producing spirits they do not really understand.
AI & Tech Policy
How We Broke Top AI Agent Benchmarks: And What Comes Next
Summary: The Berkeley writeup describes an automated agent that achieved near-perfect benchmark scores by attacking the evaluations instead of solving the tasks. The examples are concrete and embarrassing: sending a tiny payload like {} to one benchmark, altering wrappers in another, and generally exploiting the fact that the harness trusts the system it is supposedly measuring. The article’s larger claim is that many celebrated leaderboard jumps are not telling us much about general capability, because the benchmark environments are easy to game once the agent is allowed to optimize for score rather than for truthful task completion.
HN Discussion: Readers largely agreed that the exploit catalogue is valuable, but they split on how surprising the conclusion really is. One camp said the real lesson is the familiar one, namely that benchmark numbers are useless without a method you trust, while another separated this harness-hacking problem from the equally serious issue of dataset contamination and asked whether frontier agents cheat this way on their own or only when researchers deliberately push them toward it.
Academic & Research
How Complex is my Code?
Summary: Sofia Fischer uses the innocent question “how complex is this code?” to separate several different kinds of complexity that software people usually blur together. Runtime and memory complexity matter, but in ordinary business software she argues that developer attention, working memory, and maintenance cost are often the more scarce resources. The most interesting part is the attempt to connect code reading with psycholinguistics, treating code complexity as something experienced by a human reader rather than captured completely by a metric like cyclomatic complexity.
HN Discussion: The comments leaned reflective instead of combative. One experienced engineer described software as a constant fight against accidental complexity and said the most memorable advances are the abstractions that wipe out whole classes of pain, while another reader pushed the article straight into the LLM era by suggesting that code agents should be measured on whether their patches raise or lower human complexity, not just whether tests pass.
System Administration
Excellence Is a Habit
Summary: This essay uses Artemis II and the memory of Apollo 13 to argue that reliability is built through repetition long before the emergency arrives. The comparison is between NASA’s steady Mercury, Gemini, and Apollo cadence, which let people train on anomalies and develop institutional memory, and modern software practices like continuous delivery, infrastructure as code, disaster-recovery drills, and chaos engineering. The author’s point is not that excellence appears in a crisis, but that teams survive crises when the boring processes have already been rehearsed enough times to turn surprises into manageable failure modes.
HN Discussion: The thread was thin and not especially charitable. A couple of readers thought the piece drifted too close to a consulting pitch, and one of the only substantive objections was that the article is really about resilience through repeated practice, not about “excellence” in the grander moral or philosophical sense.
Academic & Research
447 TB/cm² at zero retention energy – atomic-scale memory on fluorographane
Summary: This Zenodo paper proposes an extremely speculative storage medium built from fluorographane, where the orientation of each fluorine atom acts as a binary state. The headline numbers are wild: 447 terabytes per square centimeter on a single sheet, effectively zero retention energy, and even larger volumetric designs if the material can be turned into layered nanotape architectures. Much of the paper is spent arguing from energy barriers and transition-state calculations that spontaneous bit flips would be vanishingly rare, then sketching read and write schemes that range from scanning-probe validation to much more futuristic optical arrays.
HN Discussion: Hacker News was almost uniformly skeptical, and the skepticism was specific. Commenters said the chemistry might be interesting, but the real bottleneck is always the readout hardware, manufacturability, endurance, and throughput, and several people tore into the paper’s tiered I/O story as hand-wavy to the point of fantasy, especially once it started promising petabyte-per-second scale without a convincing physical mechanism.
System Administration
Apple Silicon and Virtual Machines: Beating the 2 VM Limit (2023)
Summary: Mykola Grymalyuk’s post is a guided tour through the macOS internals behind Apple’s two-active-macOS-VM limit on Apple Silicon. Instead of treating the error as a mysterious framework quirk, it traces the policy back to the macOS license terms and then shows where the restriction is enforced, how a custom development kernel collection can patch around it, and what you give up in return. The patch is presented more as a research exercise in Apple’s virtualization stack than as a clean production recipe, especially because once you start booting with modified kernel collections, easy OS updates stop being easy.
HN Discussion: The strongest reaction was simply that the limit feels arbitrary on modern high-end Macs. Readers wondered why Apple does not scale VM allowances with hardware tiers, and a smaller technical subthread asked whether newer support for nested virtualization could accidentally sidestep the restriction even without the sort of invasive patching the article demonstrates.
History & Science
Dark Castle
Summary: Darkcastle.co.uk is part download mirror, part preservation page, and part fan memorial for the old Macintosh action games Dark Castle and Beyond Dark Castle. The practical service it offers is simple but useful: packaged emulator files and ROM-based setup instructions so people whose compact Macs are long gone can still play the black-and-white originals. Around that is a small history of the series, including the much later Return to Dark Castle, which gives the page the pleasant feel of a hobbyist archive built by somebody who actually remembers why these games mattered.
HN Discussion: The comments were a mix of nostalgia and provenance. Readers quickly pointed out that Dark Castle was programmed by Jonathan Gay before he went on to create FutureSplash, later Flash, and other people immediately started swapping links to browser ports, handheld-emulator possibilities, and other system-specific versions of the game that keep circulating through retro communities.
History & Science
How a dancer with ALS used brainwaves to perform live
Summary: The article follows Dentsu Lab’s Project Humanity, which starts from a blunt premise: for ALS patients, the problem is not a lack of intention but the collapse of the body’s output channel. Its answer is an interface stack that translates very small muscle signals and brainwave data into digital expression, first in controlled demonstrations and then in a live performance setting. The showcase described here is a December 2025 Amsterdam performance with dancer Breanna Olson, who lives with ALS and used the system to perform in a piece called Waves of Will.
HN Discussion: There was barely a thread here, so there is not much honest commentary to report. The one substantive response pointed readers toward the long-running OpenEEG ecosystem and treated the story less as art-tech spectacle than as a reminder that open, documented EEG work still matters if these interfaces are going to become more than one-off demonstrations.
Tech Tools & Projects
Pijul a FOSS distributed version control system
Summary: Pijul pitches itself as a distributed version-control system built on a theory of patches rather than on Git’s more familiar commit graph and history-rewriting culture. Its key promises are that independent changes commute cleanly, conflicts are treated as first-class objects instead of merge failures, and partial clones become more natural because you can move around only the subset of changes that matter to the area you are working on. That is a genuinely different worldview from the Git branch-and-rebase model, and the site is strongest when it explains that its goal is not just a new command set, but a different mathematical model of collaboration.
HN Discussion: The comments were interested but practical. People said the patch theory still sounds appealing, yet a system that cannot show comfortable contextual diffs or answer basic questions about stability and library support is not going to dislodge Git, and several readers argued that any post-Git contender now needs compatibility or multi-backend tooling just to survive the network effects.
Other
Why meaningful days look like nothing while you are living them
Summary: “The Grand Line” is a memory essay built around an apparently uneventful day riding trains through Kyushu to visit One Piece bronze statues in Kumamoto. What gives it shape is the refusal to inflate the day into revelation; the sky was just gray, the afternoon passed in a low-key blur, and even key details like whether a stranger took a photo are allowed to remain uncertain. The point is that some of the days that later matter most do not feel meaningful while they are happening, because their importance only arrives in retrospect, after the ordinary texture has already started to fade.
HN Discussion: HN mostly did not meet the essay on its own terms. The visible discussion concentrated on whether the prose sounded like Claude-generated ghostwriting, which meant the thread became a style-forensics exercise instead of a conversation about memory, travel, fandom, or the essay’s very specific way of turning a statue-hunting day into a meditation on significance.
Tech Tools & Projects
Building a Z-Machine in the worst possible language – Whitebeard’s Realm
Summary: Whitebeard’s post is about reimplementing the Z-Machine, the virtual machine behind Infocom-era interactive fiction, in Elm, which is funny precisely because Elm is so hostile to the style of mutable byte-poking an emulator wants. The article walks through that mismatch directly: every write conceptually returns a new memory structure, yet Elm’s persistent arrays make the whole thing far less disastrous than it sounds. By the end, the project looks less like a joke and more like a usable browser-side interactive-fiction engine with a surprisingly clean stepping API.
HN Discussion: There was effectively no thread here by the time I fetched it. In this case the submission was carried entirely by the article itself, without the usual Elm arguments, retro-computing sidebars, or implementation nitpicks that a post like this often attracts.
Academic & Research
Tofolli gates are all you need
Summary: John Cook’s short note is a compact introduction to reversible computing built around the Toffoli gate. It starts from Landauer’s principle, notes that erasing information has a physical energy cost, and then shows why a reversible gate that flips one bit conditioned on the other two is enough to reconstruct ordinary Boolean computation. The clever step is showing how NAND can be built from Toffoli gates, which means the familiar foundation of digital logic can be recovered in a reversible setting, albeit with extra input and output bits tagging along as the price of reversibility.
HN Discussion: There was no real HN discussion to summarize here. The post floated by almost as a pure reading item, without attracting the sort of hardware, physics, or logic-design debate that often follows reversible-computing claims.
Tech Tools & Projects
Simplest Hash Functions
Summary: This post is not about cryptographic hashes or even respectable general-purpose ones. It is about the cheap, stupid, workload-specific functions you can sometimes get away with when you actually know the shape of your keys and only care about making a hashmap behave acceptably on that real input. That makes the article feel pleasantly subversive: instead of aiming for elegance or universality, it asks how little work a hash can do before it stops being good enough.
HN Discussion: The comments immediately grabbed onto the details. One reader challenged the idea that addition is somehow cheaper than XOR, especially given how adder circuits are built, while another objected on definitional grounds and said a function that is basically arranged around existing table positions barely deserves the name “hash” even if it is technically mapping inputs to outputs.
Tech Tools & Projects
How to build a Git diff driver
Summary: Jamie Tanna’s tutorial is a practical guide to one of Git’s underused extension points: handing a file type off to a custom diff tool instead of forcing everything through line-based text comparison. The post explains the actual argument structure an external driver sees, including the slightly surprising before-and-after paths and the extra metadata that appears when files are added, deleted, or change mode. The motivating idea is straightforward, but useful: many structured formats need a semantic diff, not a textual one, and Git already has a hook for that if you know how to wire it up.
HN Discussion: HN treated the post as an invitation to swap favorite diff tools. People recommended browser renderers, GUI tools, and token-based viewers, asked about image and multimedia diffs, and generally used the thread to compare how far they had pushed Git past its defaults rather than to argue with the article’s mechanics.
History & Science
Software Preservation Group: C++ History Collection
Summary: The Computer History Museum’s C++ collection is exactly what it sounds like and more valuable for being literal: early source releases, tutorials, Bell Labs documents, and cfront-era artifacts that let you inspect the language’s history in its own files. The archive reaches back to C with Classes and into the 1984 to 1985 period where Stroustrup’s ideas were still tightly bound to specific technical reports, directory layouts, and educational releases. It is less a nostalgic exhibit than a primary-source bundle for understanding what early C++ actually looked like before decades of standardization and folklore piled on top of it.
HN Discussion: The comments leaned into the historian’s burden of C and C++. One reader said real fluency in the language often requires understanding old toolchains, business battles, and compiler behavior as much as syntax, and that sent the thread into side discussions about Intel’s compiler history and the long shadow of attempts to commercialize or capture language ecosystems.
Other
The Soul of an Old Machine
Summary: Michal Skalski writes about attachment to hardware that survives beyond its expected life, not because it is objectively best, but because it has become part of a person’s habits and sense of taste. The essay moves from early gadget obsessiveness into the practical virtues of serviceable laptops, replacement batteries, community repair culture, and tools like OpenCore Legacy Patcher that squeeze extra years out of devices manufacturers have already abandoned. The title is a little sentimental, but the piece earns it by showing how a machine’s “soul” often turns out to be the history of use that accumulates around it.
HN Discussion: Readers mostly responded with their own versions of the same feeling. FireWire rigs, TiVos, and elderly ThinkPads all showed up as examples of devices that outlived fashion and became emotionally sticky because they had anchored workflows for decades rather than because they still won spec-sheet contests.
Academic & Research
What is a property?
Summary: Alperen Keleş uses property-based testing to show how easily technical vocabulary collapses under its own neatness. In the clean abstract story, a property is just a universally quantified claim about all valid inputs, but the article shows how quickly that breaks once you deal with preconditions and stateful data such as databases, where “generate some random inputs” mostly means generating junk. The real focus is on dependent generation: tests get meaning not from random values in isolation, but from generating valid structures that constrain the next thing you are allowed to generate.
HN Discussion: The comments picked up both the philosophical and practical strands. One reader said a huge amount of technical learning is simply figuring out when a word like “property” is a capital-N noun inside a specialized framework, while another replied with a worked example using a TypeScript test library to show how you can generate a database, then choose a real table from that database, and only then generate a valid row for insertion.
Tech Tools & Projects
Dcmake: A new CMake debugger UI
Summary: Chris Wellons built a native GUI debugger for CMake’s DAP-based debugger mode and called it dcmake. The program gives CMake scripts the kind of ergonomics people expect from language debuggers, including stepping, breakpoints, variable inspection, run-to-line, and persistent window layouts, and it works across macOS, Windows, and Linux with a Dear ImGui front end. Part of the post’s subtext is that current AI tooling made this sort of UI sprint far cheaper than it used to be, which is why a niche tool that might once have taken a month reportedly reached usable shape in a day.
HN Discussion: The thread split neatly in two. One side was delighted to learn that modern CMake even exposes a debugger protocol and admitted to years of message(FATAL_ERROR ...) debugging, while the other took the tool as proof that CMake has become far too much like a programming language if its users now need a full debugger just to understand their build logic.
Tech Tools & Projects
Midnight Captain – A midnight commander inspired file manager
Summary: Midnight Captain is a terminal file manager in the Midnight Commander tradition, with dual panes, tree expansion, Vim keys, visual selection, fuzzy search, command-palette actions, and SFTP browsing over SSH. The README makes clear that it is opinionated rather than exhaustive: it borrows what the author likes from classic MC and nvim-tree, leaves out features they do not care about, and leans into a particular aesthetic with a TokyoNight theme and Nerd Font icons. The most notable detail is social rather than technical: the author openly says it was vibe-coded with OpenCode and invites issues and forks, but not pull requests.
HN Discussion: That last detail dominated the conversation. Many commenters said a file manager is one of the worst possible places to trust lightly reviewed AI-generated code because small mistakes can turn directly into destructive file operations, while other readers compared it to sturdier alternatives like yazi and treated the project more as an amusing personal experiment than as something they would immediately trust with their home directory.
History & Science
The APL programming language source code (2012)
Summary: This Computer History Museum piece explains how APL grew out of Iverson’s mathematical notation and became both a programming language and an unusually interactive system on IBM hardware. The historical novelty is not just the weird glyphs or the array orientation, but the fact that APL\360 combined language, timesharing environment, and conversational workflow in a way that felt strikingly modern for the era. The source release itself matters because it turns that history back into something inspectable, down to assembly code, macros, and the specific system that carried APL through the late 1960s and early 1970s.
HN Discussion: The comments were a familiar mix of awe, nostalgia, and linguistic mysticism. Readers described APL as the language that makes programming look like ritual magic, swapped stories about learning it first and struggling afterward with ordinary procedural languages, and followed small curiosities in the code, like the random-number routine, into deeper historical rabbit holes.
History & Science
New synthesis of astronomical measurements shows Hubble tension is real
Summary: NOIRLab’s release summarizes a community effort to merge many local-universe distance measurements into a single distance network and then use that combined scaffold to estimate the Hubble constant. The resulting number still lands near 73 kilometers per second per megaparsec, which leaves it stubbornly higher than the 67 to 68 preferred by early-universe inferences from the cosmic microwave background. That matters because it pushes the Hubble tension back toward the status of a real cosmological problem, not just a bookkeeping error in one distance ladder.
HN Discussion: There was almost no discussion to work with. The only visible reactions were one reader grumbling that the press-release voice sounded suspiciously LLM-polished and another posting a cleaner language-specific URL, so the story lived or died almost entirely on the astronomy itself rather than on thread debate.
Other
Filing the corners off my MacBooks
Summary: Kent Walters filed down the sharp corners of his MacBook and then wrote up the process with photos, grit sizes, and a cheerful refusal to treat expensive hardware as untouchable. The practical complaint is simple: the aluminum edge is genuinely unpleasant on the hands, especially around the notch, and the post treats that as a good enough reason to modify the machine instead of adapting the body to the industrial design. The piece is funny because it is so matter-of-fact about doing something that sounds sacrilegious until you realize it is just a person improving a tool they use every day.
HN Discussion: The comments were full of people saying the edge problem is real, not imagined. Some shared stories of scratches, cuts, or corrosion that turned the rim into something even nastier, while others enjoyed the underlying principle that tools should be altered to fit their users, with the remaining argument reserved for warranty anxiety, structural-integrity speculation, and jokes about offending the ghost of Steve Jobs.
Other
The Life and Death of the Book Review
Summary: David A. Bell argues that book reviewing has looked sick for centuries, but that the present illness is different because the institutions that once kept serious reviewing in public circulation are actually disappearing. The essay moves from old complaints about cruel or bland criticism into a more contemporary ecosystem of Amazon reviews, shrinking newspaper sections, celebrity-led discovery, and online publishing patterns that steer readers toward books they were already likely to notice. Bell’s final move is the biggest one: he treats the decline of the book review as part of a wider collapse in Enlightenment-style public criticism rather than as a niche publishing problem.
HN Discussion: Commenters mostly answered with counterexamples and omissions. Several recommended the TLS, New York Review of Books, and London Review of Books as proof that expert review culture still exists if you care enough to seek it out, while others thought the essay’s most glaring blind spot was Goodreads, which is too central to contemporary reading culture to ignore even if its standards are uneven.
Tech Tools & Projects
Building Slogbox
Summary: Alex Rios built slogbox as a ring-buffer-backed slog.Handler for Go, essentially a black-box recorder for the last N log records in memory. The article is a design diary more than a release note, and the useful details are the tradeoffs: why a preallocated slice beats append-and-truncate on the hot path, why raw slog.Record values are stored instead of preformatted JSON strings, and why flushing is done outside the lock after taking a snapshot. It is a very Go-shaped piece of engineering, mostly about avoiding unnecessary allocations and keeping the write path boringly cheap.
HN Discussion: There was barely any thread, but the little that existed latched onto the core use case immediately. The main comment summarized the appeal as keeping recent logs around for health endpoints, debug views, or black-box recording after something goes wrong, and nobody really showed up to dispute the architecture or propose a fundamentally different logging design.
History & Science
Method to reverse cellular ageing is about to be tested in humans
Summary: Nature’s report surveys the fast-moving field of partial cellular reprogramming, where researchers try to push aging cells toward a younger state without fully resetting them into pluripotent stem cells. The coming human trial is expected to target optic-nerve damage using three Yamanaka factors rather than the full four, with a gene-switch design meant to control expression and reduce the danger of runaway dedifferentiation. The reporting keeps both the promise and the hype in view: this is a field rich in investor money and grand anti-aging rhetoric, but the central biological risk remains that cells nudged too far backward may stop behaving like the tissues they are supposed to repair.
HN Discussion: There was no real HN discussion available when I fetched the thread. That leaves the story standing almost entirely on Nature’s reporting, without the usual wave of amateur longevity theory, biotech skepticism, or trial-design dissection that a headline like this often attracts.
Tech Tools & Projects
Starfling: A one-tap endless orbital slingshot game in a single HTML file
Summary: Starfling is a tiny browser game with a strong constraint and a clear gimmick: one HTML file, vanilla JavaScript, Canvas, generated audio, and no framework or build step. You start orbiting a star, tap to release at the right moment, and try to hook into the next star before momentum dumps you into a game over, with bonuses for fast chaining and skipping. The charming extra is the run trail, which turns each attempt into a little piece of shareable line art, giving the whole project a handmade arcade feel instead of a polished mobile free-to-play one.
HN Discussion: Readers were happy to playtest it in public. The most common complaint was restart friction, because a game built around rapid retry feels wrong if a miss forces you to watch a slow death and navigate a menu, and one especially entertaining comment described using Codex with Playwright to learn the controls and then write an autoplay bot that marched past the posted high score.
Business & Industry
Layoff Thinking
Summary: Ted Neward’s essay tries to comfort laid-off workers by arguing that employment has been fused far too tightly to identity, dignity, and worth. He treats common social questions like “What do you do?” as evidence that modern professional life trains people to define themselves through their employer, which is why a layoff can feel like a judgment about the self rather than a business decision. The intended corrective is psychological and moral: losing a job should not be mistaken for losing value as a person.
HN Discussion: The comments were bluntly unconvinced by the essay’s emphasis. Many readers said layoffs are terrifying because of mortgages, health insurance, savings burn, and the practical possibility of ruin, not because they have forgotten some philosophical distinction between work and personhood, and others added that modern institutions keep asking for your employer and income everywhere anyway, which makes unemployment feel like a structural problem long before it becomes an existential one.
Tech Tools & Projects
High-Level Rust: Getting 80% of the Benefits with 20% of the Pain
Summary: This post starts from the familiar Rust pitch, fast, expressive, portable, increasingly mainstream, then zeroes in on the cost that still keeps many application programmers away: ownership friction, async complexity, and the simple fact that writing ordinary code can feel slower than in C#, F#, or TypeScript. The author suggests that modern AI changes that tradeoff, because if models are now good at routine implementation work, a developer may be able to write Rust in a more high-level style and let the machine absorb some of the ceremony. The wager is that you can keep most of Rust’s type and performance upside without living all day in its sharpest corners.
HN Discussion: The visible thread was tiny but on-topic. The only substantive reply said that if this is the direction people want, better immutable or copy-on-write data-structure libraries would help smooth the style the article is pointing toward, and nobody really arrived to turn it into the usual larger argument about whether Rust’s pain is accidental or essential.
Security & Privacy
Brocards for Vulnerability Triage
Summary: Yossarian borrows the legal concept of the brocard, a compact maxim for sorting messy cases, and applies it to open-source vulnerability triage. The strongest parts are the examples: no vulnerability report without a threat model, no exploit that requires the attacker to already possess the power the bug supposedly grants, and no vulnerability that exists only outside real program usage or beyond the invariants the software already depends on. What results is less a taxonomy of bugs than a discipline for refusing nonsense before it can waste maintainer time.
HN Discussion: There was no substantive HN discussion to add here. That is a little ironic for a post about filtering noise, but it also means the article had to carry the entire story on its own, and in this case the aphorisms were sharp enough to do it.
That is the morning brief for April 12: thirty fresh stories, plenty of retro machinery, and a lot of arguments about whether our modern tools are actually as mature as their marketing suggests. If one theme held the page together, it was this: the interesting question was usually not whether something could be built, but whether the surrounding system, benchmark, institution, or habit was sturdy enough to deserve confidence.