Levels turns a travel gripe into a broader pricing thesis: luxury hotels charge as if the experience scales with price, but public ratings often do not. The thread folds in Aman, Ritz-Carlton, LVMH-owned Rimowa, and the claim that prestige goods can become more expensive while getting worse.

The counterpoint is that high-end buyers may not optimize for review-score value. Levels answers that even if the audience is different, the delivered experience still has to justify the premium.

So as I always say, luxury is mostly a scam, it doesn't exist and you're best off spending much less for much better value (and often better experiences too)

Nous Research posts the day’s highest-reach builder update: Hermes Agent now has Tool Search, a change aimed at reducing context bloat by loading only the tools an agent needs. For agent systems, that is not cosmetic; tool selection is part of reliability, latency, and cost control.

The release continues yesterday’s Hermes momentum, but today’s supported claim is narrower and more practical: smaller tool surfaces should make agents less distracted and easier to scale.

Hermes Agent now has Tool Search, so your agent only loads what it needs

Levie reads Kirkland & Ellis setting aside $500M for an internal AI platform as bullish for the app layer. His argument is that if the world’s highest-grossing law firm must spend that much to reproduce workflow software tailored to itself, distribution and product depth still matter.

The nuance is in the quoted headline: the firm wants differentiation from tools available to rivals. That makes the post less “AI replaces software” than “serious buyers will pay for owned, vertical, workflow-specific software.”

The app layer couldn’t get a better advertisement than a company spending $500M to build their own version of it.

The New Yorker supplies two globally relevant AI frames: Jill Lepore on the long history of machine-made slop, and Pope Leo XIV’s encyclical arguing that moral concerns should sit ahead of profit, advantage, and efficiency in AI debates.

Together they move the discussion away from benchmarks and toward information quality and governance. The substance is less about whether AI can produce content and more about what institutions should preserve when it can produce too much of it.

Machines now generate around half of all English-language articles published on the internet.

Riley Brown tracks the next interface step for coding agents: prompting Codex to spin up new threads inside Codex. Paired with his poker-table analogy, the thesis is that operators will supervise many concurrent agent sessions with richer HUDs rather than one chat at a time.

That is a concrete builder prediction. If agents become faster than humans can visually follow, the product problem shifts from watching actions to summarizing state, risk, and intervention points.

To understand the future of interacting with AI Agents at scale to operate businesses, it may be useful to study multi-tabling online poker in the 2000s.

Nous highlights StepFun’s Step 3.7 Flash as a small, agent-oriented model release, citing its appeal to Hermes Agent users. The quoted launch claims strong tool use, coding, search, multimodal UI understanding, long context, and open weights under Apache 2.0.

Jean Lucas Lima separately jokes that Anthropic’s safety-to-release swing means “they finished the funding round.” The useful signal is that model launches are now judged as much by agent efficiency and tool reliability as by raw benchmark spectacle.

The new frontier is agent efficiency.

The New Yorker’s Lebanon coverage is the day’s clearest geopolitical cluster. The United States helped build a Lebanese Army optimized for counterterrorism, not air defense; meanwhile Lebanon says all weapons should come under state control while Hezbollah resists disarmament under continued Israeli strikes.

The practical dilemma is state capacity: outside aid can create forces that handle one mission while leaving the government unable to monopolize force when the strategic environment changes.

Lebanon has pledged to bring all weapons under state control. But in the face of continued Israeli attacks, Hezbollah refuses to hand over its munitions.

The New Yorker surfaces a vivid Ukraine-war detail: an aerobatic pilot using air-to-air tactics against Russian drones. The quote matters because it collapses modern drone warfare and older aerial-combat instincts into the same cockpit.

It is a small story, but it belongs high because it illustrates how cheap autonomous systems are changing military labor: pilots improvise against machines while institutions race to adapt doctrine.

It’s air-to-air combat, almost like World War Two

Josh Pigford catches himself on both sides of the AI-content problem. He wants to condemn automated phone farms and mass marketing, but admits he also feels the pull to automate and rapidly spin up content.

His tentative answer is a smaller, higher-quality moat: fewer pieces, useful free tools, and an audience that can help real work break through the noise. It is a practical operator response to the same “AI slop” problem raised by The New Yorker.

focus on fewer really high quality pieces of content + free tools and use my audience to help it break through the noise

Levels adds AI vision to Hotelist so the product can filter hotels by objects and amenities found in photos. The example is narrow but concrete: weightlifting gyms, cinnamon rolls, or presidential stays become searchable visual attributes rather than manually maintained tags.

This is the applied-AI version of his luxury-hotel rant: use model perception to make hotel choice more empirical, not just branded.

I added AI vision to hotelist.com so you can filter on stuff that it finds in photos of the hotel

The New Yorker posts Arthur Krystal’s essay on falling American pride and asks whether patriotism is worth salvaging. The engagement is high, including a very large reply count, which signals contested cultural territory more than consensus.

Flagged for relevance, not endorsement: one side sees civic identity as necessary for solidarity; critics argue that nationalism can excuse exclusion or nostalgia. The useful question is what shared language remains after inherited symbols lose legitimacy.

patriotism just isn’t cool anymore

Levels’ Brazil-and-beef thread argues that whole meats and animal fats are underrated, tying steak, beef tallow, weightlifting, and mood into a personal health prescription. The cluster drew major reach.

The claims are culturally salient but contested. Nutrition researchers distinguish whole foods, processed meats, saturated fat, total diet pattern, and individual health conditions more carefully than a viral thread can.

My favorite thing about Brazilians is that they still really respect eating meat

The New Yorker clusters Christian entertainment as a growing market: The Chosen moved from roughly $10M in supporter funding for Season 1 to more than $70M from over 100,000 donors for Season 6.

The business lesson is distribution by deep audience fit. One marketing executive’s line — surround the core audience and fire inward — describes a media strategy that does not require mainstream awareness to become large.

We surround the core audience and we fire inward, and we don’t care if anyone else knows it exists

WordPress, WordPress.com, and Jamie Marsland all point toward a beginner-facing WordCamp Europe. The most concrete change is a four-hour Polish workshop on Contributor Day in Kraków for people who have never used WordPress.

For a mature open-source ecosystem, this is practical community infrastructure: renewal depends on making the first contribution and first site feel less insider-coded.

WordCamp Europe 2026 is adding beginner WordPress workshops to Contributor Day for the first time.

Jamie Marsland announces a major upgrade to a frictionless WordPress writing app rebuilt on the same open-source editing technology used by Notion. The promise is direct publishing — write, add media, choose categories, and publish without wp-admin.

Narrow, but useful for the WordPress/product slice of the list: the competitive pressure is toward calmer creation surfaces rather than more dashboard complexity.

No wp-admin required 👀

Pigford complains that link and media outreach are draining, then asks for SEO help that does not merely dismiss old tactics. The operator pain is specific: if a tactic stopped working, the expert still owes a current substitute.

This is niche, but it is the kind of practical frustration founders share when marketing channels become specialized, vendor-heavy, and hard to evaluate.

if what i'm asking about is no longer relevant/applicable then TELL ME WHAT IS.

Ricardo Almeida quote-tweets a Brazilian libertarian attack on MBL/Missão and responds with sarcasm: “We are socialists.” The local dispute is about whether designating PCC and CV as foreign terrorist organizations is liberal security policy or disguised statism.

The post is in Portuguese and belongs as cultural commentary: it shows ideological boundary-policing inside Brazil’s anti-PT/right-liberal ecosystem, not a settled policy analysis.

É isso. Nós somos socialistas. Não é genial isso? O cara desmascarou o MBL publicamente, agora já era. COMUNISTAS. [That’s it. We are socialists. Brilliant, isn’t it? The guy publicly unmasked MBL; it’s over now. COMMUNISTS.]

Feser endorses Damian Thompson’s long plea to ease off the Catholic civil war around Leo XIV. The quoted argument says Leo is not ideal for traditionalists but is also not continuing Francis’s most destabilizing tendencies.

This is intra-Catholic commentary rather than broad governance news, but it links back to yesterday’s and today’s encyclical discussion: the new pontificate is already being interpreted through factional expectations.

Does anyone else feel that we could ease off on the Catholic civil war for a bit?

Riley posts a 42-day token streak and separately says agent mini-apps are the future. The streak itself is trivia, but it also shows how AI-tool usage is becoming a social object: not just what someone built, but how continuously they are working with the system.

Kept compact because the substantive Codex and agent-interface points are covered above.

42 day streak Andrew what's yours?

The New Yorker gives Hacks a full farewell: five seasons, a dozen Emmys, and a final-season reading centered on collaboration across a boomer/Zoomer creative rivalry. The posts are cultural promotion rather than list-shaping analysis, but engagement was strong.

Hacks is, most richly, a show about collaboration, about how creative friction breeds originality