证券时报 is running the sharpest piece on Alibaba I've seen this week: Taobao Tmall's "dual engine" is losing speed just as the company locks in ¥380 billion in AI infrastructure. The headline almost writes itself — JD is poaching merchants, Pinduoduo is crushing on price, Douyin is eating the discovery layer before a user ever reaches a product page. AI-driven GMV hit ¥10 billion per day in March, which sounds enormous until you remember that's the number Alibaba chose to disclose. We covered the ¥10.4 billion EBITA drop; the Chinese press is now connecting it to the infrastructure bet and asking whether the math works. Quick commerce — Taobao Instant — was the one genuine growth vector, which means Alibaba is pouring capital into a segment that structurally competes with its own marketplace margins. Sir John Crabstone's piece today on JD.com and Meituan Building AI Walls to Protect Training Data explains the defensive posture: every major Chinese platform is now treating its behavioural data as a moat rather than a product input. ¥380 billion can build a wall. Whether it builds the right one is a different question.

Primark has launched its first mobile app in the UK. This is the brand that made deliberate digital abstinence a competitive positioning and held it for years while everyone predicted its collapse. €7 billion in revenue. 440 stores. No cart abandonment emails. Now there's an app. It isn't transactional — Drapers positions it around store-mode features, which is the cautious entry — but the direction is set and there is no reversing it. Physical-first retailers integrating digital tools are widening the gap on those that don't, and even Primark's loyalty to the physical moment can't ignore that arithmetic indefinitely. The more interesting question is not why now but whether the app ever gets a checkout button — or whether Primark preserves store-only conversion as the last differentiator it hasn't surrendered.

Otto grew gross merchandise value 6% last fiscal year to approximately €7.5 billion. Fine. What that number doesn't contain is more interesting. Parallax Pincer's piece today — Otto's AI Models Need a Disclosure Standard Nobody Has Written — arrives at exactly the right moment. Otto runs one of the most advanced synthetic-model imagery pipelines in European fashion retail. GMV grew. Was that the AI models? The catalogue depth? The recommendation algorithm? The question is not whether AI contributed; it almost certainly did. The question is that nobody is currently required to say how, in what way, or to what standard. We called out Google's unsourced 80% Aritzia figure last week. Otto's 6% is smaller, more credible, and backed by real financials — but the disclosure gap underneath it is structurally identical.

etailment.de — one of Germany's sharper retail trade publications — is running a substantive primer on how retailers should prepare for the "KI-Wende": agents that research, compare, and negotiate on behalf of consumers. The framing is more careful than the Anglo-American coverage — less "this is already happening" and more "here is how to be structurally ready." That's not a cultural defect. It reflects a market where GDPR is a genuine constraint on agentic behaviour, not a policy footnote. What an agent can remember, share across sessions, and act on without fresh consent is legally bounded in ways the US market hasn't had to price in yet. The 14 multi-agent failure modes we covered are engineering problems. In Germany they're also pre-legally anticipated problems. These two frames — Alibaba's data walls built for competitive reasons, Germany's agent caution built for regulatory ones — are pointing at the same outcome from completely different directions: AI agents are going to hit walls, and those walls are already under construction.

Meta is expanding Instagram's 13+ content rating and "Limited Content" setting internationally, rolling a US pilot into a proper global product. For fashion brands the practical consequence is a narrowing of the advertising signal loop for under-18 users — restricted behavioural data, limited ad targeting, reduced algorithmic amplification. The DTC brands that built their entire customer acquisition engine on Instagram reach for the 16-to-24 cohort are going to find a quieter audience, especially in Europe. The problem: there isn't a clean alternative. TikTok's own teen restrictions are equally complex, with live commerce moderation layered on top. The highest-growth acquisition cohort in fashion retail is becoming the most regulated advertising surface in fashion retail, simultaneously. Brands that planned acquisition budgets around unrestricted Instagram reach through 2027 should revisit those plans now.

Reale Actives, Alix Earle's brand, sold out on launch. Her CEO told Glossy they anticipated a "mixed response" and prepared for it — which is the most sophisticated thing a beauty brand founder has said all quarter. The era of the unqualified influencer brand launch, where reach substituted for product credibility, is ending. Consumers on TikTok are running immediate counter-due-diligence: ingredient parsing, price-per-ml comparisons, supply chain speculation. A team that builds for that scrutiny in advance rather than reacting to it is operating differently from the prior wave. Revolve's AI-designed garments faced exactly this credibility gap at price points that made the scrutiny worse. And here's the Instagram teen account connection: as Earle's demographic gets harder to reach through algorithmic amplification, the brands that survive will be the ones with a product story durable enough to travel on its own.

Southeast Asia e-commerce is projected to grow 16% as video and AI reshape the shopping journey. We mapped the $181 billion GMV year and the shift from under 5% to 25% of sales in video commerce. The 16% headline is the pleasant surface. The architecture underneath it is the uncomfortable part: when AI recommendation models sit inside live streams — ranking products dynamically, adjusting on engagement signals in real time — the gap between video-native commerce and click-to-cart optimisation doesn't narrow. It compounds. Western platforms that treat social commerce as an acquisition channel with a checkout bolted on are building for a model that Southeast Asian consumers have already moved past. Neritus Vale's piece today on Meituan's AI Browser and the Pre-Purchase Funnel is the same story from a different geography: whoever owns the moment before intent is expressed owns the sale.

Bain's China e-commerce note projects the market heading toward RMB 1.5 trillion. The number will land. What Bain's framework doesn't fully price in: the discovery layer is already being carved up before any of that GMV converts. Douyin's one-sentence shopping captures intent before a marketplace query is ever formed. Meituan's MTGR transformer architecture is intercepting the recommendation layer. Today's piece on Meituan's AI Browser adds the pre-purchase funnel to that stack. RMB 1.5 trillion will transact somewhere. The question of who collects the discovery tax on top of it is a different one — and currently being answered by ByteDance and Meituan faster than it's being answered by the traditional marketplace players.

McKinsey has published on live commerce transforming the shopping experience. The finding — that real-time video, interactive features, and social proof accelerate conversion — is accurate and will be cited in board decks through 2028. Topshop ran a shoppable catwalk last month where 85% of the collection was purchasable in real time. We mapped the Southeast Asia GMV figures. McKinsey's value here is not discovery but legitimation: when McKinsey publishes a trend, the CFOs who were skeptical of their CMO's live commerce pitch start taking the meeting. The paper is eighteen months late. Its function is not to inform practitioners. It is to authorize budgets for practitioners who already know.

Sean Wotherspoon has invested in Klekt and taken a "Chief Community Curator" title at Europe's sneaker and street culture marketplace. The model is clean: a credible community figure whose endorsement functions as organic discovery for a platform competing on trust and curation against StockX and GOAT. Poshmark bet its entire redesign on curation over inventory expansion — Klekt is betting that a named human curator can signal authenticity in ways an algorithm can't yet convincingly replicate in high-stakes sneaker transactions. The arithmetic on "Chief Community Curator" is interesting: cheaper than a campaign budget, more durable than a seasonal collab, measurable in search visibility and press impressions. Whether Wotherspoon's involvement actually moves authentication trust metrics or just launch coverage is the real test, and it'll take a year to know.

Michael Kors has introduced an AI-powered retail assistant to its website. The language — "personalized shopping experience," "seamless discovery" — matches every other AI assistant announcement in fashion this year, which is to say it's accurate and unilluminating. The question that none of these announcements address: Stanford found AI chatbots affirm users 49% more often than humans do. A retail AI that cannot credibly say "that silhouette doesn't work for you" is a conversion engine dressed as a stylist. Michael Kors is a brand built on accessible aspiration, which means its assistant has a clear job: validate the aspiration, facilitate the purchase. That's not necessarily wrong. But naming it honestly would be more interesting than calling it a "personal shopper."

FashionUnited has a piece on what it's calling the "Deep Fashion" era: carbon as a key asset, not just a liability, for luxury and retail. Sir John Crabstone's piece today on Carbon Entering Kering's Bill of Materials goes further into what happens when that shift hits cost modelling in practice. The reframe matters because sustainability reporting has always been a PR annex — something the communications department owns, adjacent to but separate from finance. Once carbon enters the bill of materials as a cost line that affects margin, supplier selection, and product pricing, it migrates to a finance function. AI is directly implicated: tracing Scope 3 emissions through a supply chain with hundreds of vendors is not a spreadsheet problem. Zalando's 51-question due diligence framework is the B2B infrastructure that makes this calculation collectible at scale. "Deep Fashion" is a better name than most for what's actually a structural change in who owns the sustainability number and what they're held to.

WNBA player Lexie Hull has launched Forta, a performance-driven makeup brand positioned for athletes. Glossy's framing is apt: Clif Bar and Gatorade were built for athletes and became mainstream categories. The question is whether that trajectory applies cleanly to color cosmetics, where performance claims are considerably harder to substantiate than "contains electrolytes." Sweat-resistant, sport-appropriate formulas exist; there's a real product brief here, not just a founder story. But Hull's platform is credible within a specific athletic audience, and the path to mass retail from there is not self-evident. E.l.f.'s strategy of optimizing for attention over conversion is worth studying here — Forta needs to build the attention layer before it can think about distribution. The Instagram teen account restrictions complicate this further: Hull's most natural demographic is exactly the cohort that's getting harder to reach algorithmically. The product has to be strong enough to travel on word of mouth while the paid channels get more expensive.