Aaron Levie reads the latest JOLTs surprise as evidence against the simplest AI-layoff story. Engineering is his core case: AI expands the number of software projects companies can attempt, but humans still need to understand, maintain, secure, and upgrade what gets built.

The thesis travels beyond engineering into sales and marketing, where agents may increase throughput and therefore create more work to manage. It is a bullish counterclaim, not a settled labor-market result, but it is the day’s clearest strategic argument.

AI is going to have the opposite effect that lots of people thought on jobs.

Levie’s second enterprise-AI post turns Uber’s reported $1,500-per-month cap on vibe-coding tools into a market-sizing argument. If companies tolerate token spend that dwarfs old per-seat software licenses, the addressable market for “intelligence” is larger than the old SaaS market.

The constraint is budget discipline, not lack of demand. Caps may slow waste, but the willingness to pay hundreds or thousands per employee signals that productivity software is being repriced around output rather than access.

This shows you how big the TAM for intelligence is in the enterprise.

Edward Feser answers the Iran-war headline with Washington, Jefferson, and Madison on Congress’s power to declare war. The quoted OSINTdefender post says the House passed a War Powers Resolution attempting to limit President Trump’s military powers and send the measure to a contested Senate vote.

The substance is not a prediction about the resolution. It is a reminder that American foreign-policy fights keep returning to the same constitutional stress point: executives move quickly toward war, while the legislature is supposed to deliberate first.

The constitution supposes, what the History of all Govts demonstrates, that the Ex[ecutive] is the branch of power most interested in war, and most prone to it.

Nous Research follows the previous day’s desktop push with a browser-admin post: the Hermes Web Dashboard is described as a feature-complete panel that can be managed entirely from the browser.

For agents, the administrative surface matters. Provider credentials, tasks, gateways, logs, and safety controls need a place that non-terminal users can operate without turning every change into a CLI session.

The Hermes Web Dashboard got a major overhaul: it is now a feature-complete admin panel that you can manage entirely from your browser.

Riley Brown’s strongest product post says AI-built software will depend on reusable building blocks with extensive instructions. Users may never know which microservice or component their agent assembled in the background.

That reframes opportunity away from obvious app wrappers and toward “unobvious,” “unsexy,” or “unverifiable” domains where agent-friendly primitives do not yet exist. The platform layer becomes less visible but more valuable.

Many users will have their AI build things for them, and be completely unaware of these building blocks.

Riley’s Codex Sites example adds a Convex database that an agent can write to, turning a hosted page into a small persistent app. His example is a todo list, but the pattern is broader: give the agent a skill, a database, and a deploy target.

The interesting point is not the todo list. It is that Codex Sites can become a lightweight surface for agent-managed tools that persist across conversations.

You can make a todo list app that is hosted on Codex Sites, then tell Codex to make a skill that allows the agent to WRITE/EDIT to that todo list db.

Levels praises Lassie’s launch video because the startup frames its $47M funding around founders leaving Robinhood and Superhuman to work inside doctors’ offices. The quoted launch says Lassie now serves 700+ practices and claims 30 hours of autonomous labor per month.

The distribution lesson is practical: AI-for-operations stories are stronger when they show domain immersion, not just automation promises. Funding announcements increasingly need evidence that the team learned the workflow firsthand.

This is how you do a great startup video

Riley Brown backs X’s anti-aggregator push and repeats his older argument that curator accounts should be labeled and demonetized. His claim is platform-economic: automated reposting is a short-lived tactic because networks eventually treat it as spam.

The builder lesson is blunt. In an AI content environment, synthetic scale makes trust and originality more valuable, while platforms try to stop paying twice for the same recycled post.

The only sustainable content strategy in the AI future is to be real, authentic, and to build trust.

WordPress.com announces WordPress 7.0 “Armstrong” as a step toward the world’s largest CMS becoming AI-native, while Jamie Marsland asks users what they like about WordPress Desktop Mode and points to a writer-focused posting interface.

Taken together, the posts show WordPress experimenting at three layers: core release positioning, desktop workflow, and a simpler writing surface. The stakes are whether a legacy CMS can absorb AI-native expectations without losing its publishing identity.

it’s a major step toward the world’s largest CMS becoming AI-native.

Jon Yongfook credits Claude with reducing an AWS bill through Lambda optimization and S3 Intelligent-Tiering. The post is short, but the operator signal is concrete: LLMs are already useful for the unglamorous cost-audit work that teams postpone.

The practical caveat is that cloud optimizations still need human review. But as a first-pass investigator across usage patterns and service docs, an assistant can turn infrastructure hygiene into a cheaper recurring habit.

AWS bill. Thank you Claude!

Josh Pigford follows up on his “BBQ” subject-line experiment with a skull emoji and “this worked phenomenally.” It is anecdotal, not a controlled study, but it continues his practical thread about outbound email response mechanics.

The narrow lesson is that distinctive subject lines can matter because inboxes are crowded and overly polished sales language is easy to ignore. The broader claim still needs data beyond one operator’s campaign.

this worked phenomenally ☠️

Carlo Lancellotti asks what higher education would look like if accreditation agencies had spent 25 years pushing instructional-faculty-to-administrator ratios instead of diversity metrics. A follow-up on an $11B endowment sharpens the target: wealthy institutions can carry large administrative classes for years.

The post is polemical, but the operational question is legible. Universities face pressure not only over ideology or tuition, but over the ratio of teaching labor to bureaucratic overhead.

what if the major accreditation agencies instead of obsessing about diversity for a quarter of a century had tried to promote a reasonable instructional faculty to full-time administrators ratio?

Levels turns a Danish store’s sofa, coffee, and Wi‑Fi corner into a small market observation: boyfriends and husbands waiting during shopping are an underserved audience. The follow-up imagines the dormant market in plainer terms.

It is mostly lifestyle commentary, but it doubles as product thinking. A retail environment can increase dwell time by serving the person who is not shopping yet still controls patience, transport, or shared spending.

Danish stores respect guys

Ross Douthat reacts to the Washington archdiocese removing Monsignor Stephen Rossetti as an exorcist and ending affiliation with the Saint Michael Center. His frame is less ecclesial analysis than genre shock: the UFO/demon angle makes the institutional news feel unusually 2026.

Without more context, the briefing should not adjudicate the dispute. The relevance is cultural: Catholic authority, paranormal discourse, and public communications collided in a way that made even a seasoned religion columnist reach for a bingo-card joke.

Did not have "exorcist sacked for calling UFOs demons" on my bingo card

Ricardo Almeida posts through Renan Santos’s Flow appearance, first tracking the audience and then reacting sharply to reports that Renan discussed Eric Voegelin. The punchline compares the candidate’s level to Cleitinho and Flávio Bolsonaro.

This is inside-Brazilian-politics discourse, but it matters because the list has been following Missão/MBL as a third-way project. Ricardo’s criticism shows that intellectual branding can become a liability even among sympathetic observers.

Acho que o nível do candidato é similar ao de Cleitinho e Flávio Bolsonaro.

Ricardo also pushes back on Bruna Frascolla’s contempt for Voegelin, arguing that intellectuals should not dismiss world-famous authors so casually. A later post softens the personal attack: he calls Bruna talented and intellectually competent while saying her fixation on Olavismo and MBL led her astray.

The cluster is a culture-war footnote, but it shows a recurring pattern in Brazilian intellectual politics: imported philosophers become proxies for local factional identity.

Ela é bem talentosa, na verdade, uma intelectual competente.

Ricardo reacts with “Tercei… kkkkkk” to a Metrópoles report that Romeu Zema left room to compose with Flávio Bolsonaro at the right moment. The joke lands because “third way” positioning often collides with coalition arithmetic.

Flagged for relevance, not endorsement. The tweet is slight, but it continues the week’s Brazilian politics thread: anti-polarization branding has to survive the ordinary incentives of alliances, runoff math, and party bargaining.

Tercei… kkkkkk

Levels responds to Grok’s Imagine 1.5 Preview with a single heart-eyes emoji. The quoted post says the preview is available in the API, so the substance belongs mostly to the product announcement rather than the reaction.

Included as a narrow AI-tooling signal: image-generation releases still move quickly enough that a tiny reaction from a builder account can be a useful pointer, even when it does not contain analysis.

😍

Two lighter Ricardo posts sit below the political arguments: one explains that he is wearing a Brazilian national-team shirt while also running a small tailoring business, and another simply points followers to a “great work of art.”

They are useful mostly as texture. The same account moves between MBL politics, Hobbes research, taste, and side-business promotion inside the same daily window.

Nas horas vagas do mbl e da minha pesquisa sobre Hobbes eu toco uma alfaiataria.

Several posts are too thin for standalone treatment: Riley Brown points agent-interested followers to a YouTube video, Josh Pigford posts a “fellow kids” joke, and another Pigford post is only an eyes emoji plus a link.

They round out the day without changing its center of gravity. The substantive clusters are enterprise AI economics, agent infrastructure, platform incentives, and a handful of cultural arguments.

Posted a YouTube video. If you like agents you’ll like it.

WordPress posts two travel-and-community notes for WordCamp Europe 2026, welcoming attendees to Kraków and pointing them toward the June 4–6 event at ICE Kraków Congress Centre.

The content is mostly event logistics and community atmosphere. It belongs at the bottom because the more consequential WordPress item today is the AI-native release positioning, not the conference welcome copy.

Before WordCamp Europe 2026 begins, take time to enjoy the city, connect with the community, and settle in for an unforgettable few days together.