Nous Research adds a live product milestone to the list: Hermes Agent v0.15.0, branded as the “Velocity Release.” In a day otherwise dominated by Claude/Opus speculation, this is the concrete builder news — an agent platform release with enough engagement to become one of the day’s highest-reach posts.

Hermes Agent v0.15.0 - “The Velocity Release”

The expanded list brings in a second axis on the day’s AI discussion: not model benchmarks, but governance. Douthat uses Pope Leo’s encyclical to ask whether neutral transnational oversight can actually restrain AI, while The New Yorker frames the Church’s critique as a direct challenge to Silicon Valley’s salvation narrative.

A.I. leaders like to say that they are saving the world. The Pope fears they are destroying it.

Douthat points readers to a conversation with Anduril’s president about the immediate future of warfare. It broadens the list’s AI-heavy day into defense and autonomy: not just what agents can do in browsers and offices, but how fast software-defined systems are changing military assumptions.

This week's Interesting Times is a conversation about the (immediate) future of warfare with the president of Anduril:

The New Yorker surfaces the next phase of the Alex Murdaugh saga: the court clerk accused of trying to influence the murder conviction is now on trial herself. It is more legal-culture than builder news, but it adds a high-stakes institutional-accountability thread to the expanded list.

The court clerk Becky Hill allegedly did everything she could to get the lawyer Alex Murdaugh convicted of murdering his wife and their son.

WPTuts adds a practical design critique from the WordPress edge of the list, objecting to new seven-color combinations. Narrow, but useful signal for readers tracking WordPress product polish and community feedback.

More of those wonderful @WordPress 7 colour combinations that any junior level designer would shudder at!

Detailed performance benchmarks comparing Opus 4.8 against 4.7 across enterprise use cases. Box tested the model on real-world knowledge work: report drafting (87% vs 77%), legal NDA review with better clause detection, and financial analysis with 8-point improvement on accuracy metrics.

"Opus 4.8 outperforms on a majority of report drafting tasks, producing more complete and accurate analytical reports."

Riley follows up on yesterday's major AI agent announcement tease with a correction. His initial prediction of a "VERY big day" drew 437K views but the timeline shifted. Context: this relates to Codex browser improvements and OpenAI's planned multi-tab sessions.

"I hope you know tomorrow is going to be a VERY big day in the world of AI agents..."

Riley demonstrates Codex's browser persistence feature that maintains login sessions across chats. The agent can now create Notion docs, send him links, and work collaboratively in real-time within the same interface.

"This update to Codex is 10x more important than any marginal improvement to the models."

His thesis: agent platform improvements matter more than raw model capability gains.

Riley argues for prioritizing interface improvements over model capability advances. He's more excited about Codex and Claude Desktop enhancements than new model releases.

"Unless it's a major breakthrough in model capability I'm much more excited for super-app updates."

The implication: Claude has significant catching up to do on user experience.

Shpigford's satirical take on the Opus 4.8 launch claims he's been using version 4.8 "for the past 2 years" and declares it a startup killer. The joke plays on AI hype cycles and over-the-top launch rhetoric.

"🚨 i've been using 4.8 for the past 2 years and it's a GAME CHANGER. absolutely CRUSHES it. this kills EVERY STARTUP. just dead. all of them."

Shpigford follows Levels' approach and deploys Claude AI on his server for system administration tasks. The result: successful debugging and patching of a memory leak issue.

"pulled a @levelsio and installed @claudeai on my server to help sysadmin that junk."

Practical validation of AI-assisted server operations.

Levels adds a filter to his hotel site showing where US Presidents have stayed, reasoning that presidential security standards might indicate hotel quality.

"New filter added where [ 🇺🇸 US Presidents stayed ] Because if the President stays there it might be a good hotel"

The day's highest-reach tweet (357K views). Levels chronicles his frustration with Dutch hotel policies: AC limited to 23°C, locked windows, no room cleaning by default, and sustainability messaging on coffee machines. Quote-tweets his own earlier rant about EU regulations.

"Western Europe is so fucking tiring"

This continues yesterday's theme of European climate policy criticism.

Feser pushes back against claims that school voucher advocacy equals "Christian nationalism," noting that free market economists like Milton Friedman developed the concept decades before the term existed.

"the school voucher concept was developed by free market economists like Milton Friedman and has been a mainstay of conservative politics since at least the 1980s"

Levels connects EU climate policies to broader concerns about funding allocation. He claims the AI fund promised GPUs to startups but delivered to cronies instead, calling it "taxpayer money burnt."

"It's all cronyism and a lot of people getting rich from trillions in EU funding which never arrives to the actual people who it is made for"

After posting about a bomb threat incident near his hotel, Levels explains why Dutch tactical teams wear jeans and sneakers: they're on 24/7 call and need to respond in minutes.

"Jeans and Nikes is standard wear for Dutch tactical operations teams because they're on call 24/7"

WordPress extends a conference invitation to Dave Winer after he applied to speak at WordCamp US. Winer's bio claimed credit for "most of the tools and standards behind WordPress" including blogging, RSS, OPML, and XML-RPC.

"Dave, you invented and promoted many of the core standards that we still use today! You're welcome to have a spot of honor at any of our conferences."

Feser announces he's finished reading Pope Leo's latest document and teases upcoming commentary, dismissing existing "hot takes" as uninformed.

"Just finished reading through Pope Leo's Magnifica Humanitas. It's a rich and very important document, and many of the hot takes about it are dumb."

Riley marvels at Anthropic's social media influence, suggesting they could launch any product and receive massive attention.

"dude what is going on Anthropic could launch a T-shirt and it would get 10M views."

Simple social coordination tweet from Riley looking for viewing companions in NYC.

"Who wants to watch the Spurs game in NYC?"

Shpigford promotes his Granite product launch on ProductHunt with characteristic low-key enthusiasm.

"hey chat! granite is on @ProductHunt today!"

Standard content creator announcement about new video upload.

"New YouTube Video up."