Nous Research’s biggest-reach post says Hermes Agent was built to work smoothly on NVIDIA’s new RTX Spark superchip and the OpenShell runtime tied to Microsoft security primitives. The post matters because it places an agent product inside a hardware, runtime, and enterprise-security story rather than only a chatbot story.

The Computex framing is distribution-heavy: if local AI PCs become a default buyer category, agent software that already speaks the platform’s runtime and security language gets a more credible path into everyday desktop work.

We have been working closely with @nvidia to ensure Hermes Agent works smoothly on their new @NVIDIARTXSpark superchip and integrates with the new OpenShell runtime, which connects Hermes to @Microsoft's security primitives.

Nous follows the NVIDIA announcement with a product milestone: Hermes Desktop is now in public preview after a GTC keynote demo. The pitch is straightforward but strategically important: move the agent from a remote interface into the user’s own machine.

That shift changes what agents can plausibly do. A native desktop agent can connect files, apps, local context, and permissions in ways that web-only workflows struggle to match, while also raising the standard for security and user control.

Introducing Hermes Desktop: everything you love about Hermes, now native on your machine.

Aaron Levie answers a hard enterprise AI question: if everyone can buy similar frontier models, what stays defensible? He argues that the moat shifts to internal institutional knowledge, existing data assets, and domain-specific workflows connected to AI.

He is reacting to Kirkland’s reported internal legal-AI spend, but the argument travels. Every industry has a version of the same build-versus-buy question: whether proprietary workflows become durable advantage or get standardized by vendors.

The companies that are able to best harness their internal institutional knowledge, existing data assets, and domain-specific workflows -- connected with AI -- will be those that are able to stay ahead in the future.

Levie frames OpenAI’s Bedrock availability as a distribution event. AWS already has large enterprise commitments, so making OpenAI frontier models and Codex available through familiar security, compliance, and governance workflows can broaden model adoption.

The point is less about one cloud press release than about how AI consumption is sold. Procurement channels, governance wrappers, and existing cloud spend can decide which models enterprises actually use.

AWS has massive enterprise traction, with large committed contracts from enterprises.

Levels argues that telling a public story is effectively free user acquisition, but it charges the founder in personal brand and time. He contrasts that with SEO, which AI search may weaken, and paid ads, which require cash.

His follow-up is a useful builder constraint: a normal story rarely travels. The public narrative needs a modern edge that makes the project legible and memorable without relying on yesterday’s novelty, such as unfunded backpacker startups.

So telling your story on X (and other social media) is essentially free user acquisition But you pay with your personal brand (and time)

Riley Brown’s most substantive cluster is a thesis about super-apps and agent-native software. He says standalone applications should pivot toward agent-native apps because a full browser inside a desktop agent can make separate design or app-building surfaces feel redundant.

The Sites example extends the same idea: Codex can turn work and plans into a shareable site, while automations can keep a personal site updated. The common claim is that agents become the orchestration layer and apps become surfaces they open, modify, and publish through.

If you're building a standalone application pivot to making an agent native app.

A separate Riley post shows Paper inside Codex being used as a visual ideation board. It can pull thumbnails from YouTube or other sites, place them on a board, and use the built-in image model to mix visual concepts.

The signal is that coding agents are becoming multimodal workbenches. The workflow is not only code generation; it is browsing, collecting references, arranging assets, and turning rough concepts into editable artifacts.

Every image is an element, and i'm using the built in gpt- image 2 model that's built into codex to mix and match.

Nous says it integrated NVIDIA’s official Agent Skills catalog into the Hermes Skills Hub. The concrete value is procedural: skills for CUDA-X libraries, Omniverse and Physical AI workflows, NeMo training and inference, and other platform components.

For builders, that points to a durable layer in agent products. If agents need to operate specialized platforms, packaged skills can become the difference between generic reasoning and repeatable execution.

These skills teach your agent how to use CUDA-X libraries, Omniverse and Physical AI workflows, NeMo training and inference tools, and other platform components.

Josh Pigford quotes an Ahrefs research thread on AI search optimization and calls it “interesting and brutal.” The cited finding is narrow but actionable: “Best X” blog listicles are reportedly the most prominent content format cited by AI chatbots.

That does not mean every operator should spam listicles. It does mean AI-search visibility may inherit and amplify a specific kind of comparison-page content, changing how companies write for discovery.

Interesting and brutal.

Pigford keeps turning an email problem into operator diagnostics. He asks whether low-volume wholesale outreach should move from a newer warmed domain to the primary shop domain with more history, then shares example emails to active customers that drew no replies.

The practical lesson is that outreach performance is not just copy. Domain reputation, spam placement, recipient context, sequence length, and the channel’s own decay all interact.

example emails. these are to active customers. have sent variations of these emails to probably 30+ customers. not a single response.

WordPress.com announces WordPress Workspace for Mac as a way to bring editing, notes, screenshots, media, and AI tools into one desktop workflow. Another post points to the platform’s response after more than 2,000 sites were targeted in a supply-chain attack.

Together the posts address two unglamorous site-owner problems: fragmented production and security response. Both matter more to daily operators than another abstract AI claim.

Managing a @WordPress site can mean jumping between your editor, notes, screenshots, media library, and AI tools. But it doesn’t have to.

Pigford also ships a small Granite workflow feature: users can forward emails and attachments directly to the product and have them parsed and organized. He pairs that with a reminder that supporting tax documents often need to be retained for years.

This is narrow, but the operator value is clear. Document tools win when intake is easy, classification is automatic, and retention rules are visible before a shoebox of files becomes a problem.

you can now forward emails + attachments directly to https://t.co/kVQhsVK4Qn and have them automatically parsed/organized!

Ricardo Almeida reacts to polling and press coverage around Renan Santos, the Missão/MBL figure being framed as a Brazilian “third way.” He says the consolidation is necessary for debate access and describes Santos as the only pre-candidate building an alternative antagonistic to both Lula and Bolsonaro.

Flagged for relevance, not endorsement. Early polls, ballot access, media incentives, and the instability of anti-polarization lanes are all counterweights to a strong “third way” reading.

O único pré-candidato que construiu uma via alternativa igualmente antagônica a Lula e a Bolsonaro.

Edward Feser points readers to his essay on Pope Leo XIV’s Magnifica Humanitas and later supplies a longer teaser. He argues the document is not primarily about AI, does not break with traditional teaching, and belongs within Catholic social teaching’s larger account of human dignity.

The useful signal is how technology debates are being routed through theological categories. AI appears as one part of a broader argument about modern power, slavery, just war, and the meaning of humanity.

There is no break with traditional teaching. While artificial intelligence (or AI) gets significant attention, the encyclical is actually devoted to a much larger theme, of which AI is only a part.

Levels attacks a European Commission post about “the battle of narratives,” calling the institution Orwellian and accusing it of recasting observable conditions as narrative warfare. The quoted Commission post frames Europe as a recurring target and calls for telling stories that protect democracy.

This is a contested cultural-politics item. Critics will see bureaucratic narrative management; defenders will see a normal response to disinformation and hostile influence. The briefing should keep both frames in view.

I think what's crazy is just how much the European Commission flat out lies While calling the truth "a narrative"

Chris Arnade says New York City World Cup games are technically walkable, then immediately undercuts the point: nobody should or will walk them. His broader claim is that the World Cup is a television event with stadiums attached, and FIFA follows the richest media market.

The post turns a local logistics gripe into a media-economics claim. Fan experience matters, but venue selection is shaped by broadcast money, market size, and institutional incentives.

The World Cup is a television event that happens to have stadiums attached, and the U.S. is the richest media market on earth.

Ross Douthat quote-tweets Mark Levin’s long Independence-era thread with the understated line, “This post is a journey.” The substance sits more in the rhetorical arc than in Douthat’s own sentence: a 250th-anniversary reflection moves quickly through war, legitimacy, and national memory.

As a briefing item, it is mainly a signal of political-cultural framing around the semiquincentennial. The short commentary does not make a full argument, but it points to the mood of the conversation.

This post is a journey.

Pigford’s lower-stakes commentary names two pathologies of online attention. The “ragebaiter,” he says, speaks from a desire not to be forgotten rather than from principle; a bot-detection screenshot labels “almost no original posts” in a way he reads as thoughtlessness.

The posts are not a full theory, but they capture a useful cultural distinction: platforms reward reaction volume, while users still use originality and principle as informal tests of whether an account is worth hearing.

the ragebaiter never speaks from a place of principle, only a desperate desire to not be forgotten.

Riley Brown’s Wispr Flow outage post is personal, but it says something about workflow dependence. Once dictation or voice-to-text becomes part of a creator’s default operating system, a temporary outage feels less like inconvenience and more like losing an input method.

The same background chatter includes a Codex Browser Use good-morning post and a weekly Codex summary. None is a major topic, but together they show how tightly his public posting loop is tied to agent tools.

bro wispr flow went down for me... i'm so cooked lol i hate typing.

The rest of the day is texture: Pigford marvels at multi-nozzle 3D printing and posts a small win; Jamie Marsland shares a WordCamp Asia speed-build photo; Antonio Risério promotes a YouTube program; Chris Arnade posts a spare media reaction.

Portuguese-language side chatter rounds it out, including Ricardo Almeida pointing followers to Paulo Cruz on Euphoria and Jean Lucas Lima agreeing with an AI-boom concern. These are kept compact because they preserve the list’s background without pretending every post is a major signal.

multi-nozzle printing is freaking witchcraft. also GRADIENT MIXING?!?!?!?!?!?