Adam Skye Jones

From linkThe post on Om's site is a stark family notice rather than a retrospective essay: it says Om Malik died on June 24 after nearly two years of treatment for stage 4 lung cancer, and it frames the loss through the traits people around him knew directly, calling him brilliant, funny, curious, generous, fiercely loyal, and impossible to replace. It also points readers to his long-running about page and shares practical details for a celebration of life in Tiburon on August 22 plus memorial donations, which makes the page feel intimate and grounded instead of performative.

From commentsThe HN thread is full of people treating Malik as one of the rare tech writers who could both understand the industry and keep enough distance to stay honest about it. Readers remembered his early blogging voice, his patience with founders and operators, and the way he could be sharp without turning cynical, while others shared small personal stories about mentorship, dinners, and unexpectedly kind check-ins. The overall tone is not abstract praise for a public figure so much as a community realizing that one of the people who helped narrate Silicon Valley for two decades is suddenly gone.

From linkThe Vesuvius Challenge update says researchers have crossed from deciphering fragments to reading a complete sealed scroll: the first full Herculaneum roll has now yielded substantial continuous text without being physically opened. The project describes eleven readable columns, roughly two thousand Greek characters, and a probable title and author attribution to Philodemus's On Vices, which matters because it turns the effort from an impressive imaging demo into an actual pipeline for recovering classical literature from a library carbonized by Mount Vesuvius.

From commentsThe HN discussion mixes amazement with a surprisingly practical curiosity about why these texts stayed unread for so long. Some commenters focused on the emotional weight of recovering voices from antiquity that looked permanently lost, while others talked through preservation constraints, asking why archaeologists cannot simply unroll the papyri and how much confidence scholars should place in machine-assisted reconstruction. The thread reads like people recognizing a real humanities breakthrough while still wanting to understand the physical and interpretive limits around it.

From linkThe post is a compact field report from an intentionally hostile launch: Fernando Borretti invited roughly two thousand people to attack his AI assistant and then wrote up what broke, what surprisingly held, and what he changed afterward. The useful part is that it stays concrete, covering the app's model, tool, and background-job architecture, then documenting outcomes like six thousand sign-up emails, no successful credit-card grabs, and repeated pressure on prompt boundaries and isolated job execution. It reads less like growth hacking and more like a rare public postmortem on how agent systems behave once curious strangers actually start probing them.

From commentsThe HN thread rewards the honesty of the experiment more than the product itself. Readers appreciated seeing someone publish the ugly parts of prompt-injection risk, tool abuse, and sandbox design instead of claiming hand-wavy safety, while others argued that the exercise mostly proves how hard it is to make autonomous agents robust once they can read arbitrary input and call external systems. The consensus is not that the problem is solved, but that transparent write-ups like this are more valuable than another polished demo because they expose the real attack surface people keep underestimating.

From linkJeff Geerling uses Framework's new 10GbE expansion card as a way to explain how messy USB-C really is once you leave the marketing diagrams and start chasing actual bandwidth. The module works by squeezing an Aquantia controller and a full RJ45 port into Framework's swappable-card form factor, but the result is constrained by how the laptop surfaces PCIe lanes through USB4, so you get a clever piece of modular hardware that still cannot deliver desktop-class 10-gig performance in every configuration. The post is really about the compromises hidden inside elegant standards: form-factor constraints, thermals, protocol layers, and the mismatch between what a connector implies and what the system behind it can truly sustain.

From commentsThe HN comments split between admiration for Framework shipping something this nerdy at all and skepticism about how useful 10GbE is on a laptop once the real bottlenecks are laid bare. Some readers were delighted by the sheer modular-hardware energy of the thing, while others argued that 2.5GbE already covers most plausible mobile workflows and that the card mostly demonstrates how overcomplicated USB-C has become. Even so, the thread stays constructive because the hardware is interesting precisely as a case study in where open, swappable designs run into the ugly physics and standards baggage of modern I/O.

From linkThe FIRE essay argues that age-verification laws are turning ordinary web access into an identity checkpoint problem, where reading or viewing lawful content increasingly means proving who you are to a third-party gatekeeper before a page will load. The piece treats that as a structural privacy failure rather than a narrow compliance annoyance, because the same systems meant to keep minors away from some content also normalize retention of sensitive identity data, create attractive breach targets, and make anonymous access feel like a suspicious exception instead of a default feature of the open web.

From commentsThe HN comments are predictably alarmed, but not in a shallow way. People connect the essay to broader KYC creep, pornography age-gating, app-store policy, and the general habit of solving every messy online problem by adding another identity tollbooth. A few readers argue the policy direction is politically inevitable or technically easy to route around, yet the stronger current is that even flawed workarounds miss the deeper issue: once governments and platforms get comfortable tying access to credentials, the internet starts inheriting the bureaucratic posture of border control.

From linkThe essay's core claim is simple and uncomfortable: free and open-source software has become critical infrastructure, but the people maintaining it are still too often expected to do so on volunteer energy, intermittent sponsorships, or outright personal sacrifice. Rather than treating this as an abstract fairness complaint, the piece frames it as an ecosystem risk, because companies, governments, and end users now rely on deep dependency chains whose weak points are often exhausted individuals with little bargaining power. It is a reminder that the romance of open collaboration does not magically solve payroll, burnout, succession, or institutional accountability.

From commentsThe HN thread turns quickly into the usual but still unresolved fight over whether the problem is corporate freeloading, licensing strategy, or simply the economics of making shared tools. Some commenters argue maintainers should choose business models, dual licensing, or foundations earlier instead of hoping goodwill scales, while others push back that the market keeps externalizing costs onto unpaid labor because permissive ecosystems make that easy. The discussion does not land on a neat fix, but it does show broad agreement on one point: the software world keeps talking about open source as infrastructure while funding it like a hobby.