On its Q1 earnings call, Hims & Hers said it would "soon" launch an AI companion to support users along their weight-loss journey. "Companion" is doing a lot of work there. Today's piece reads the architecture more plainly: prescribe the GLP-1, enroll the patient in a coaching loop, intercept the dropout moment before it becomes a cancellation. Every touchpoint is proprietary; the sequence is retention, not support.

The timing is what makes it interesting. Alphabet launched its own AI health coach the same week — Neritus Vale has that story today. One company owns the query and the search surface; the other owns the prescription and the patient's lapse history. Hims & Hers has pharmaceutical data Google doesn't: dosing outcomes, dropout timing, side-effect signals. The race is to the chronic-care refill, and the winner is whoever builds the better churn model. These two are approaching the same wellness data moat from opposite ends of the funnel, and they are going to collide somewhere around the second prescription renewal.

198 years old. Over 1,100 perfumes in the archive. 2026 is when Guerlain ran its first paid influencer campaign — prompted by TikTok discovering that a $25 drugstore fragrance was an acceptable proxy for a $660 bottle and loudly broadcasting that fact to several million people. Today's piece gets into the channel they built. The short version: they didn't fight the dupe. They used the dupe's viral energy as evidence that the original is worth $635 more. That's the right call.

The brands that argue with the dupe machine lose. The ones that make the dupe conversation their distribution funnel — that position the copy as a tribute and the original as the aspirational object — are learning something the legacy houses are still deciding whether to admit. At 198 years, Guerlain's first paid campaign being a direct response to TikTok is either the most embarrassing thing that happened to the house this year or the most strategic. The speed at which they moved will tell you more than the campaign itself. Prestige fragrance houses have been slow on every digital channel; the ones that are fast now are recalibrating.

Non-English source — this one is worth your attention. While every English-language outlet is writing about how AI agents are going to transform shopping, Germany's etailment published the uncomfortable question nobody else has asked: when AI is doing the buying, how does a checkout system tell the difference between a legitimate shopping agent, a human, and an attacker? Typing cadence, cursor movement, session timing, device fingerprinting — the entire behavioral fraud toolkit breaks when the buyer is a Claude or Gemini instance running on someone's behalf.

Every agentic commerce deployment we've tracked — retailers shipping ChatGPT plugins faster than evaluation frameworks can catch up, shopping agents scoring 76% on synthetic customer profiles — has a fraud surface nobody wrote a policy for. The piece is partly a vendor pitch, but the underlying question is genuinely unanswered: when a machine acts on a human's behalf at checkout, who authorized the transaction, and how do you prove it at scale? This is the gap that is going to generate a very loud conversation the moment someone loses serious money through it.

Both beauty retailers are now integrated into major AI shopping layers. Both describe the investment as personalization. Neither is saying the part that matters: this is a toll road they're paying to stay visible on before the road decides whether it needs tolls at all. John Lewis ran the same calculation. The race to ship these integrations is moving faster than any rigorous assessment of whether they're building durable discovery share or just buying time. Nobody has published a number on what an AI storefront integration returns. Somebody will, and it won't be comfortable reading for everyone who paid at the current rates.

McKinsey's live commerce report dropped this week, describing the format as "transformative" for the shopping experience. Meanwhile, Topshop just paired a catwalk show with a TikTok live beauty shopping experience — executed in real time, see the next entry. TikTok Shop's 84% beauty surge isn't coming; it came. McKinsey is the trailing indicator, not the leading one. By the time the consulting playbook arrives, the brands have already run two seasons and renegotiated their creator deals. This is not a criticism of the research — it's structurally true that any format that becomes reportable has already been tested.

The show runs; the TikTok live shopping session runs alongside it; viewers buy beauty products while the clothes walk the runway. The format isn't the news — live commerce is proven. The structural detail is the pairing with a catwalk: the shopping moment gets an emotional container. The comment section closes sales; the catwalk gives people a reason to be in the comment section at all. What Topshop built here is the session intent layer that raw product livestreams lack, and that's the part worth copying.

Michael Kors launched an AI-powered retail assistant directly on its site — not inside a third-party layer, as its own infrastructure. The critical distinction is owned. When Amazon joined the Universal Commerce Protocol and when John Lewis paid to live inside ChatGPT, they were renting discovery. Michael Kors is building it in-house. That works until the customer stops typing into the brand's search bar and starts asking their phone what to buy instead — which is the exact scenario Parallax Pincer's piece today describes in its structural form. Building owned AI infrastructure and building AI discovery share are two different problems. Michael Kors just solved the easier one first.

BoF's opinion piece asks whether AI will make online shopping worse — the concern being that agent-driven discovery optimizes for the model's training logic rather than individual customer preference, and gradually homogenizes what gets surfaced. This is the right question asked about six months after it stopped being novel, but it needs to be in print. The specific mechanism worth naming: if every shopper's agent pulls from the same model trained on the same web crawl, brand discovery trends toward the brands with the longest data footprints and the cleanest structured metadata. New entrants don't have data. Heritage brands with inconsistent catalogs get buried. We noted in May that every beauty retailer is pitching curation as differentiation — which looks increasingly naive if the discovery layer is an algorithm that genuinely doesn't read brand narratives.

Months after spending €1.13 billion absorbing ABOUT YOU, Zalando pushed a dedicated shopping app into the UK. ASOS has home-field advantage, Shein owns the price tier, and the UK has seen every major European fashion platform attempt entry at some point with varying results. The app is either the opening move of a serious territorial push or a flag-plant that becomes an outpost. Given what Zalando just paid for continental scale, and given that ABOUT YOU already had UK presence, this reads as commitment rather than test. The question is whether European-scale buying leverage and algorithmic catalog depth translate in a market where fast-fashion loyalty is fickle and the switching cost for the consumer is zero.

Not Amazon itself — the third-party seller layer sitting on Amazon's fulfillment infrastructure — now accounts for 5.5% of total UK retail sales. That's the size of the road Amazon was building toll booths on when it opened its logistics network and extended its surcharge in the same week. 5.5% of UK retail through a single third-party marketplace means any AI commerce layer that captures Amazon buyer intent — any shopping agent that routes queries through Amazon by default — is touching a structural share of physical-world spend. It also means the retailers paying for ChatGPT and Gemini integrations need to think carefully about which platform they're actually competing with for discovery. It isn't each other.

Non-English source — Chinese financial press, worth the translation pass. 证券时报 (Securities Times) ran a piece this week headlined roughly: "Taobao's dual-engine stalls; Alibaba bets ¥380 billion on AI while falling behind on growth." The domestic framing is meaningfully harder than the Western coverage of the same announcement. We read the ¥380 billion as defensive capex when it dropped, and the Chinese financial press has now caught up to that read: JD is poaching engineering talent, Pinduoduo is treating AI as a marketplace risk management tool, Douyin is eating discovery from a standing start. The ¥380 billion looks large until you run the compound math on three rivals who are ahead on different vectors and not slowing down.

Alana Pallister sold her house to fund 300,000 units of the Blare tracksuit. Five months later, a million units sold. The next bet is Coachella as a demand-generation event. This is the intelligence test for DTC fashion after a viral breakout: can you manufacture a second moment, or was the first one an unrepeatable combination of product timing, algorithm luck, and the particular social mood of a specific week? Coachella is a reasonable place to hunt — documented taste-formation event, right demographic overlap, real media amplification. But brands that blow up on TikTok often discover that TikTok's loyalty is to novelty, not to them. The Blare tracksuit succeeded because it felt like a discovery. The follow-up has to feel like one too, which is a content operation as much as a product one.

Dr. Joyce Park launched Kerativ, a science-backed hair care brand. The dermatologist-founder playbook is well understood in skincare — Dr. Dennis Gross has 26 years on it — but hair care is earlier in this cycle. The question is whether the consumer who paid a credibility premium for a dermatologist's skincare will follow the same logic in hair, or whether hair has enough established science-positioned brands (Kérastase, Olaplex, Virtue) that the credential advantage compresses faster than it did in skin. The Forta performance beauty category is running the same credential-transfer trade from athletics. Two different bets that scientific legitimacy is an underpriced positioning asset in their categories. At least one of them is right.

The wellness industrial complex has claimed altitude. Fiji Airways will offer health shots, wearable devices, and red-light therapy panels on board — an in-flight wellness program the airline pitched to Glossy alongside its CEO. There is a real market here for the premium version: business class passengers who would otherwise sit inert for eight hours are a captive audience for recovery products, and the spend profile overlaps with the wellness consumer already buying Thorne supplements and Oura rings. The gap between offering red-light therapy on a plane and making it feel medically credible rather than spa-hotel decoration is a real execution problem. Airlines are running about four years behind the Six Senses-class hotel sector on this logic, which means they're about to spend money learning lessons already paid for at altitude.