Fast Retailing's H1 numbers: revenue up, operating profit climbing 14%, guidance raised. Crabstone reads the full results today — the short version is that this record wasn't built on any AI narrative Fast Retailing has been pushing. It was built on inventory velocity and supply chain precision compounded over a decade. The contrast with Alibaba's ¥10 billion EBITA drop over the same period is stark: one company cashing in on boring operational excellence, the other spending through a ¥380 billion AI wager that hasn't paid out yet. Both strategies are coherent. The scorecard so far favors the boring one.

The Chinese financial press is asking what Western analysts won't: is Alibaba's ¥380 billion AI wager working while the core business loses ground? Securities Times frames it directly — Taotian's dual-engine (Taobao + Tmall) losing thrust as JD.com poaches traffic, Pinduoduo pulverizes on price, and Douyin hoovers attention. We covered the RMB 10.4 billion EBITA drop at the quarter's end. The follow-on question from Chinese financial media: does ¥380 billion buy enough model advantage before the operational lead evaporates? Meanwhile JD.com and Meituan are blocking rival AI tools specifically to protect the data moats that would make any AI advantage compound. The data war and the spending war are the same war.

E.l.f.'s CDO told Glossy that "culture moves at the speed of swipe" and the brand competes for attention before it competes for conversion. It's a principled position — until the consumer's next step is asking ChatGPT for a drugstore bronzer and the assistant recommends whichever brand has better catalog data, not better cultural heat. Crabstone's piece today closes the argument. We covered the structural gap in the earlier e.l.f. piece: an entertainment strategy has no conversion defense layer when discovery moves off the feed entirely. The question isn't whether Gen Z watches e.l.f.'s content — they do. The question is whether they cart from an AI query instead, and whether e.l.f.'s product data is strong enough to win that query.

Reale Actives cleared inventory in hours. The CEO told Glossy they were "prepared for a mixed response" — which is a considered way of saying they anticipated the ingredient scrutiny that followed. Parallax Pincer breaks it down today in full: the 24-hour post-launch window is now commercially as significant as the launch itself, because the audit arrives at the same speed as the sell-through. And Crabstone's piece today on Gen Z running brands through AI before the cart closes the loop — sell-out velocity is table stakes. What the AI ingredient check returns is the actual brand equity moment.

Meta is rolling out age-appropriate 13+ content ratings and a Limited Content setting for teens internationally. Crabstone maps the funnel math in today's piece: a significant slice of beauty's most enthusiastic buying cohort will now see fewer product posts, hauls, and tutorials on Instagram. The timing is uncomfortable — fragrance just had a breakout moment on TikTok Shop and the whole beauty sector has been betting that social-native discovery is permanent infrastructure. TikTok Shop's live AI ranking layer increasingly looks like the alternate channel brands are going to need sooner than they planned.

FashionUnited's AW26 survey documents the corset-and-crinoline wave running through Paris — Nina Ricci and others at price points where structural complexity is the entire value proposition. Parallax Pincer makes the economic argument in today's piece. The connection that matters: Neritus Vale's piece today on the garment reconstruction problem virtual try-on left behind runs directly into this. The categories where AI imagery is weakest are precisely the categories where material engineering IS the product. A generated corset image tells you the silhouette. It cannot tell you the boning. A prompt cannot build what decades of atelier practice developed — and the AW26 corset moment is the market reminding everyone of that in the same season that AI fashion imagery is everywhere.

Michael Kors has added an AI-powered retail assistant to its website, with the expected language about personalized recommendations and thin implementation specifics. The specifics matter: which model, what training data, whether it pulls from live inventory or a static product dump. Because the catalog has to be right for the AI to be right, and fashion catalog data at this tier is notoriously inconsistent. When the assistant gets a product wrong, Kors absorbs the complaint. The AI gets none of it. The first mid-market brand to publish honest conversion numbers here sets the benchmark for the whole tier. Worth watching in six months.

Topshop's AI-generated catwalk is now pairing with TikTok Live beauty shopping — Shark Beauty and LOOKFANTASTIC joining to make the spectacle transactional. The model is legible: AI-generated visual content as the container, live commerce as the payload, clothes as a format rather than products anyone will wear off a rack. Topshop's post-physical relaunch has essentially bet on synthetic production plus live social distribution as the full brand stack. Southeast Asia built this architecture years ago. Parallax Pincer's piece today on 70,000 consultants replaced by a campaign provides the cost argument for why this model keeps accelerating. The lingering question: does a catwalk that was generated convert at the same rate as one that was filmed? Nobody has the data yet.

Onton has raised $7.5M for AI search and discovery in fashion ecommerce. The space is genuinely crowded: Algolia owns the budget line nobody cuts, Constructor is going hard on B2C, and every major platform has its own layer. At $7.5M you are not fighting those budgets. The viable path is a vertical the big players haven't fully staffed — independent boutiques, specialty resale, a non-English market with complex taxonomy. The fact that investors are still funding dedicated search infrastructure despite the "agents will replace search" narrative is its own signal. That gap in expectations is probably worth money to the right team.

16% ecommerce growth for Southeast Asia, with video doing the structural work. Consistent with the data we covered when video expanded from under 5% to 25% of SEA GMV in a short window. The point Western platform teams keep missing: it's not that video works as a format add-on, it's that the entire discovery-to-purchase logic is rebuilt when video is the default interface. The AI layer being bolted onto Western checkout pages is solving for browsing behavior already obsolete in the fastest-growing markets. And Neritus Vale's piece today on Vietnam's $110 billion export target completes the picture from the supply side: the manufacturing infrastructure being built here is explicitly designed to cut out the brand layer.

Otto grew GMV to around €7.5 billion — 6% up year-on-year, a solid result in a grinding European platform market. We covered their MOVEX platform generating photorealistic model imagery at 60% lower cost in our piece last week. The GMV growth and the AI model deployment are running in parallel and neither number references the other. What's still missing: a disclosure framework for synthetic model imagery that neither Otto, nor any regulator, nor any industry body has produced. A growing catalog of AI-generated visuals attached to a growing GMV figure, no framework for what shoppers are owed. That combination doesn't stay stable indefinitely.

Glossy traces dupe fragrance from peripheral novelty to category phenomenon. The "now what?" is the right question. The mainstream moment for dupes typically arrives just as the originals start winning on AI-mediated discovery — because AI assistants recommend from product data and brand authority, not olfactory experience. Fragrance is doing very well on TikTok Shop right now through social proof and creator amplification. But AI scent description is still primitive, and the dupe market's survival depends on the original staying culturally visible — visibility increasingly mediated by platforms the originals are only beginning to understand. If the originals get the AI discovery layer right before the dupes do, the category dynamics shift fast.

German trade press etailment.de has a readiness primer on agentic commerce — worth flagging not for the content, which covers familiar ground, but for the signal: when the German-language trade press publishes "how to prepare" explainers, the concept has crossed from early-adopter discourse to planning horizon for the DACH market. Douyin had a ship date in March. Germany's wholesale channel is still asking what the vocabulary means. The failure mode research is already published. The preparation window is shorter than this primer's framing implies.