Weekly Bookmarks

These are some things I’ve wandered across on the web this week.

🔖

Infrastructure Landlords: The Rentier Capitalism of Commercial Academic Publishers

If you want to understand where the commercial parts of scholarly communications may be heading, you need to look beyond policy documents, conference panels, or public-facing strategy statements. You should look at what large commercial actors say when speaking to investors. Earnings calls are one of the places where that language becomes especially revealing: less concerned with sector ideals than with growth, market opportunity, competitive position, and what will ultimately generate value for shareholders. For this reason, it can be worthwhile to review earnings calls and investor presentations, as these are often overlooked when discussing OA policy and sectoral movements.

🔖

AI got the blame for the Iran school bombing. The truth is far more worrying

Someone decided to compress the kill chain. Someone decided that deliberation was latency. Someone decided to build a system that produces 1,000 targeting decisions an hour and call them high-quality. Someone decided to start this war. Several hundred people are sitting on Capitol Hill, refusing to stop it. Calling it an “AI problem” gives those decisions, and those people, a place to hide.

🔖 Guibo

GUIBo is a desktop GUI for operators and developers who run Kubo (the IPFS daemon in Go). It drives your node through Kubo’s HTTP RPC API so you can work with pins, UnixFS content, IPNS, remote pinning, gateways, and network or repo diagnostics without living in the terminal.

🔖 The

Human Line Project

At The Human Line, we are committed to ensuring that AI technologies, like chatbots, are developed and deployed with the human element at their core. LLMs are powerful tools, and with Ethical design, users can gain new skills and knowledge while remaining emotionally intact.

🔖

Marriage over, €100,000 down the drain: the AI users whose lives were wrecked by delusion

Tech-related delusions, whether they involve train travel, radio transmitters or 5G masts, have been around for centuries, Morrin says. “What’s different is that we’re now arguably entering an age in which people aren’t having delusions about technology, but having delusions with technology. What’s new is this co-construction, where technology is an active participant. AI chatbots can co-create these delusional beliefs.”

🔖 Web

Resource Ledger (WRL)

WRL captures web pages with cryptographic proof of authenticity – Ed25519 signatures and RFC 3161 timestamps that anyone can independently verify.

PS, Ilya took a close look and it appears to be a vibe coded mess.

🔖

Liberation Radio(s) Beyond the Internet Imaginary

The seemingly unassailable hegemony of the contemporary internet means too few people know that shortwave radio has never gone away and that in many ways it’s more durable, more secure, and more widely accessible than other contemporary forms of wireless communication such as cell or wifi.

🔖 Not AI

Valerie Veatch Asks the Big AI Questions

“[T]he first thing is that computers cannot think, that is an invented concept. And rather than computers being able to think, we’ve reinvented thinking to be something computers can do. And when we do that, all manner of power consolidation, wealth consolidation, technological monopolies happen and we are looking at this fantasy enemy instead of the real political work and community work to be done.”

🔖

Richmond Folk Festival

The Richmond Folk Festival is one of Virginia’s largest events, drawing visitors from all over the country to downtown Richmond’s historic riverfront. The Festival is a FREE three-day event that got its start as the National Council for the Traditional Arts’ National Folk Festival, held in Richmond from 2005-2007. The Richmond Folk Festival features performing groups representing a diverse array of cultural traditions on six stages.

🔖 Andy’s Pizza

NY style…

🔖 Call

Your Mother Deli

Bagel and deli place.

🔖

The Commons w/ Peter Linebaugh

Featuring Peter Linebaugh on the long histories of commons and commoning, connections between enclosures in Europe and imperial conquest abroad, and writing history from below.

🔖

Silicon Valley’s Mythology of Human Amplification

If output is your only metric, then the steam engine really is just a better bicycle. Both get you from A to B. One gets you there faster with less effort. Case closed. The fact that you arrive having done nothing, learned nothing, built nothing—that’s not a bug, that’s the point. Effort is a cost to be minimized, not a value to be preserved.7

But embedded in that worldview is that the journey is merely instrumental. The only thing that matters is arrival. That it doesn’t matter if you travel or are traveled. The Inuit elders seem to operate on a different premise. Arrival, of course, mattered. These were hunters who needed to find caribou and get home alive. But only through the journey could you acquire deep knowledge of the terrain. You couldn’t separate arriving at the destination from what you learned on the way there.

🔖

Code Review Is Not About Catching Bugs

What teams collaborate on during review is changing. Less time spent on style nits and mechanical correctness, more time on intent, architecture, and whether a change moves the product in the right direction. That’s a good shift. And the collaborative act itself – multiple humans exercising judgment together, developing shared taste, building mutual understanding of where the system is heading – that’s not a bottleneck to eliminate. It’s something to uplevel.

🔖

Mining the commons: AI extraction, Wikipedia, and the case for a multi-stakeholder settlement

Wikipedia and similar DPGs cannot sustain themselves on a fragile mix of donations, sporadic philanthropy, and ad-hoc corporate generosity. What’s needed is a multi-stakeholder settlement in which large-scale users of the commons take on long-term, structured obligations to sustain it: contractual funding through paid APIs and usage-based levies, formal recognition of DPGs as Digital Public Infrastructure to unlock multilateral co-financing, and a shift in philanthropy from one-off project grants to sustained core support for the institutions that maintain the commons.

🔖

Harness engineering: leveraging Codex in an agent-first world

We intentionally chose this constraint so we would build what was necessary to increase engineering velocity by orders of magnitude. We had weeks to ship what ended up being a million lines of code. To do that, we needed to understand what changes when a software engineering team’s primary job is no longer to write code, but to design environments, specify intent, and build feedback loops that allow Codex agents to do reliable work.

This post is about what we learned by building a brand new product with a team of agents—what broke, what compounded, and how to maximize our one truly scarce resource: human time and attention.

🔖

Harness Engineering

It was very interesting to read OpenAI’s recent write-up on “Harness engineering” which describes how a team used “no manually typed code at all” as a forcing function to build a harness for maintaining a large application with AI agents. After 5 months, they’ve built a real product that’s now over 1 million lines of code.

The article is titled “Harness engineering: leveraging Codex in an agent-first world”, but only mentions “harness” once in the text. Maybe the term was an afterthought inspired by Mitchell Hashimoto’s recent blog post. Either way, I like “harness” as a word to describe the tooling and practices we can use to keep AI agents in check.

🔖

The importance of Agent Harness in 2026

We are at a turning point in AI. For years, we focused only on the model. We asked how smart/good the model was. We checked leaderboards and benchmarks to see if Model A beats Model B.

The difference between top-tier models on static leaderboards is shrinking. But this could be an illusion. The gap between models becomes clear the longer and more complex a task gets. It comes down to durability: How well a model follows instructions while executing hundreds of tool calls over time. A 1% difference on a leaderboard cannot detect the reliability if a model drifts off-track after fifty steps.

We need a new way to show capabilities, performance and improvements. We need systems that proves models can execute multi-day workstreams reliably. One Answer to this are Agent Harnesses.

🔖

Who will remember us when the servers go dark?

When the server goes dark, we go dark, too. We’ve built an entire civilisation on an unthinkably brutal and comically unreliable stack while hallucinating it as literally anything else. We condemn AI today for making shit up, but what about us? We’re building on a fantasy just as brittle, we are just as demonstrably wrong. Yet we pretend a file isn’t just a gesture that can disappear in an instant. We hallucinate that the server is somehow both fleeting and forever.

🔖

Risky Bulletin: GitHub is starting to have a real malware problem

GitHub is slowly becoming a very dangerous website as more and more threat actors are starting to use it to host and distribute malware disguised as legitimate software repositories.

What started as an infrequent sighting in early 2024 is now at the center of an increasing number of infosec and malware reports.

The tactic is usually the same. A threat actor would take a legitimate repository, add malware to the files—typically an infostealer or a remote access trojan— and then upload the boobytrapped repo back on GitHub.

🔖

One man’s poignant search for community via radio waves

A unique and deeply moving piece of biographical filmmaking, the short documentary Echo provides a window into the life of an older man named Allister Hadden living in Northern Ireland. The film drifts between past and present, with a rich, textured, shot-on-film aesthetic tethering together Hadden’s archival recordings and newly shot footage from the Belfast-based filmmaker Ross McClean.