Aaron Levie frames enterprise token spend as evidence that AI systems are now being used at scales their early buyers did not price for. The strategic response is not simply choosing cheaper models; it is building an applied layer that knows enough about the workflow to route each task to the right model.

The argument turns cost pressure into a product moat. Frontier models remain necessary for high-end work, but the winners in applied AI may be the companies with domain evals, workflow-specific routing, and pricing aligned to customer economics.

Token costs are becoming one of the hottest topics for any enterprise I talk with right now.

Levels turns Jen Zhu’s “output uptick, flat adoption” chart into a blunt distribution thesis: AI makes app-building easier, but it does not give builders an audience, paid acquisition budget, or taste for original viral execution.

His follow-up sorts likely winners into three buckets: VC-funded startups spending heavily on distribution, influencers trading reach for equity, and a small number of unusually creative operators who can catch the zeitgeist. The useful warning is that “everyone can build” increases competition for attention rather than removing go-to-market work.

I think the challenge is that everyone can now build apps

Edward Feser explains Pope Leo XIV’s comments on just war by separating defensive force from wars pursued for broader aims. His reading is that modern weapons and infrastructure damage make traditional criteria harder to satisfy in practice, not that every act of defense is illegitimate.

His longer follow-up presses the civilian-harm premise. Flagged for relevance, not endorsement: critics of restrictive just-war accounts will dispute whether the doctrine can be narrowed this far, but Feser’s point is that good aims do not erase foreseeable harm to innocent third parties.

It’s the enormous number of civilians who have done nothing deserving of harm and yet are killed, or have their homes destroyed, or whose society’s infrastructure is so devastated that anything like normal life becomes impossible for them.

Nous Research announces Simplified Chinese support across Hermes Desktop: chat, sidebar, settings, command center, cron, messages, profiles, skills, agents, and the rest of the UI. The post is in Chinese, which is itself the product demo.

For agent software, localization is a distribution feature. The change moves Hermes from an English-default developer tool toward a broader desktop surface that can meet users in their own interface language.

Hermes Desktop 现已支持简体中文——聊天界面完整适配简体中文。

Riley Brown predicts Mythos will be able to one-shot almost any full-stack mobile app and submit it to the App Store, but the economics are conspicuous: at roughly five times Opus pricing, a run could cost about $120–$400.

The builder signal is less “mobile apps are solved” than “agent capability is becoming priced like a serious production input.” Usage on Claude Code or Desktop will test whether builders pay frontier-model rates for end-to-end delivery.

Mythos will be able to one shot almost any full stack mobile app end to end and submit it to the app store...

Josh Pigford says an audience helps but does not guarantee that products work or make money; Rob Walling adds portfolio data from 241 SaaS companies, saying less than 5% started with a notable social audience while roughly 22% reached seven- or eight-figure ARR.

The combined point is a useful correction to founder folklore. Media reach can lower one acquisition cost, but SaaS companies still compound through customer discovery, reputation, partnerships, and networks that are not identical to posting.

If you want to do SaaS, build your network, not your audience.

Levels uses Matthew Prince’s Khosla anecdote and the Travis Kalanick/Uber history to warn founders that investment is also a transfer of power. The post is familiar anti-VC terrain, but it names the trade plainly: money arrives with ownership, governance rights, and future leverage.

A second post praises Franz founder Stefan Malzner for rejecting VC, trying a donor-funded team, then returning to a team of one. The cluster contrasts capitalized scaling with speed, control, and lower coordination costs.

when you raise money, you don't just receive millions $$$, you also sell something back

WordPress wraps WCEU with 2,458 attendees from 81 countries, CERN going live on WordPress, WordPress 7.0, and AI running through the event. The official account’s live-session posts and Matt Mullenweg’s remote commentary make the conference the day’s main WordPress cluster.

The practical signal is project positioning: WordPress is tying its open-source community, large institutional deployments, and AI roadmap into the same event narrative.

WordPress 7.0 and AI ran through the whole event

Levels replies to a suggestion that founders should lean on SEO by arguing that AI changed the competitive set. Before AI, search meant competing with a handful of people who had invested effort; now anyone can generate plausible pages at scale.

The practical implication is that text production is no longer the hard part of SEO. Distribution, backlinks, originality, and trust become more important as generated content floods query surfaces.

After AI, you're now competing with 1000x more because anyone can generate pretty human sounding pages that rank on Google

Levels says a judging key appeared in a livestream URL bar, forcing Vibe Jam ratings to be cancelled and reset from the stream recording. The fix is small but instructive: future judging URLs strip the key and set a cookie instead.

This is classic MVP security debt. Hashes in URLs are convenient until screenshares, streams, browser history, logs, or indexing expose them.

Keys/hashes in URL bars are fun for MVP but never good idea long term cause they always somehow get exposed

Matt Mullenweg asks at WCEU whether Cloudflare Emdash has “spiritually succeeded” WordPress, then points to 263 GitHub releases and jokes about a CMS with more releases than users. It is a jab, but the target is a real product question: developer infrastructure companies keep circling publishing surfaces.

The useful reading is competitive posture. WordPress sees CMS challengers not only from traditional website builders, but from infrastructure platforms trying to turn edge primitives into content systems.

Is this the first CMS with more releases than users?

Josh Pigford gives Granite a narrow document-management hook: Google Drive will not remind users when a passport is about to expire, but Granite will. It is a small feature claim, yet it clarifies the product’s positioning.

The practical value is metadata plus workflow, not storage. Important documents become more useful when the system knows deadlines, renewal triggers, and family-continuity needs.

will your google drive remind you when your passport is about to expire? no.

Levels extends his anti-luxury argument from hotels to luggage after a Rimowa suitcase breaks again between Copenhagen and Lisbon. The sharper follow-up is about service: a premium travel product failed while the store told him to seek repair in the city where it was bought.

The claim is polemical, but the consumer test is concrete. If the premium promise is durability and global support, failures during travel expose whether luxury pricing bought service or only brand scarcity.

The point of service is especially when it's a suitcase, you're probably traveling when it breaks

Ross Douthat reacts to Birth Gauge’s new country update by noting a shifting fertility picture: Europe stabilizing after a 2015–2023 drop, South Korea climbing from its nadir, and Thailand and Taiwan taking the bottom slots.

The post is only one line, but the demographic topic is globally consequential. It points away from a single “lowest-low fertility” story toward a rolling map where regional recoveries and declines happen at different times.

Europe stabilizing after the 2015-2023 drop, South Korea climbing from its nadir, Thailand and Taiwan taking its place at the bottom.

Levels uses a Grok explanation to frame hotel eco programs as margin expansion: hotels avoid cleaning rooms, add profit, and market the reduction in service as sustainability. The post continues his recent suspicion of premium travel brands.

Flagged as consumer commentary rather than settled accounting. Hotels may have real resource-saving motives, but the complaint names the obvious incentive problem when “green” programs also reduce labor and service costs.

It's all greenwashing to make more money off of you while giving you much less

Devon Zuegel thanks a San Francisco Chronicle reporter for spending time on Esmeralda Institute’s proposal, interviewing many people, and digging into history and context. The post is mostly gratitude, but it indicates what the project needs: not hype, but locally literate scrutiny.

Her Edge Esmeralda photos remain community texture. The substantive thread is that new-city and temporary-village projects depend on whether outsiders understand the proposal in civic, historical, and neighborhood terms.

She spent her time to really learn about what we're proposing with @esmeralda_inst, interviewed tons of people, & dug deep on history & context

Devon Zuegel quote-tweets David Perell’s announcement of a TV show about charm, character, and architectural sameness, calling it a story that needs to be told. The quoted premise asks why so much recent building feels generic across cities.

This belongs with the list’s recurring urbanism interest. It is not a policy plan, but it marks continued demand for media that treats built-environment quality as a mainstream cultural question.

Extremely excited for this! It is a story that needs to be told

Jon Yongfook punctures the now-standard AI-founder line that “building was never the hard part, distribution is.” The joke lands because the sentence is increasingly repeated by people with little evidence they have done either.

As a companion to Levels’ more serious distribution thread, it shows the phrase becoming both true and overused. Builder culture is already developing antibodies against its own new clichés.

"Building was never the hard part, distribution is" - AI bro who has never built anything, and has 36 followers

Devon Zuegel posts a simple feed-temperature check, saying Twitter has felt more fun and higher quality in recent months. The claim is subjective, but it is a useful meta-note from someone whose work depends on online communities.

Kept low because it is mood rather than argument. It does, however, explain why several community and urbanism threads continue to surface here rather than moving entirely to private channels.

The quality of content on my feed the last few months is way better than it has been for years

Several posts are mostly texture: Chris Arnade reports that “No-Names” is back but still crazy, Edward Feser approves of a physical-media enthusiast, and Jean Lucas Lima points readers to the quote-tweets under a venture-capital anecdote.

They are grouped rather than inflated into separate cards. The day’s bottom tier is pets, books, and side-channel reactions around larger conversations.

Welp, No-Names back (good news!), but still crazy.