Phia's founders knew about cookie stuffing for months

Phoebe Gates's AI shopping startup just lost its defense. Bloomberg reports that Gates and co-founder Sophia Kianni knew Phia's extension was claiming affiliate credit for sales it never drove — for months, and that they pushed for the features behind the practice.

Bloomberg reported Monday that leaked Slack messages and people familiar with the matter show Phia co-founders Phoebe Gates and Sophia Kianni knew their AI shopping extension was "cookie stuffing" — automatically injecting affiliate tracking cookies so Phia got commission on purchases it never influenced — as far back as December, and that they personally pushed for the attribution features that made it possible. The new reporting says the behavior was a purposefully built, switchable feature rather than the bug the company called it in July, that cookie stuffing made up a meaningful share of Phia's sales, and that daily revenue dropped noticeably once it stopped. Nike and Nordstrom were among the retailers whose transactions got misattributed.

The August follow-up lands on top of Bloomberg's July 9 investigation, which found the extension opened a hidden tab at checkout, loaded its own affiliate link, and in some cases overwrote the tracking of the publisher who actually sent the shopper. Independent researcher Ben Edelman, who has documented the tactic since 2004, reviewed the code and said the behavior did not look accidental, estimating Phia claimed credit for up to ten times the purchases it actually influenced. Affiliate network impact.com suspended Phia's account, and the company is now facing the same legal exposure as PayPal-owned Honey, which is still fighting a class action over identical conduct. A Phia spokesperson told Bloomberg the company rebuffed some of the claims but would "learn from this"; TechCrunch says Phia did not respond to its request for comment.

The damage here goes beyond the fraud itself. Puck reported that Phia lost nearly half its full-time staff since the start of the year, that some brands did not know they were listed on the app, and that investors were spooked by how aggressively it leaned into affiliate marketing — this after a November Fortune investigation into how the extension collected sensitive user data. For a consumer AI product whose entire pitch is trust — find me the best price, reliably — a months-long, founder-directed attribution scheme is close to existential. The Honey case shows the courts are willing to take cookie stuffing seriously, and founders pushing for the feature, per Bloomberg's sourcing, turns a technical glitch story into a governance one.

What to watch: whether retailers like Nike and Nordstrom push back publicly, and whether the Honey class action gives Phia's own users a template.

When an AI shopping agent's business model depends on quietly taking credit, how much should founders personally answer for it? Tell us in the comments.

Deep Dive — America's top open model still trails China's best

Meta released the weights for Muse Spark Open on Thursday and called it "the most capable US open-weight model to date." Independent measurement agrees with the first half of that sentence and complicates the second: on Artificial Analysis's Intelligence Index the model scores 39, first among American open weights and behind GLM-5.3 at 45, Kimi K3 at 44 and Alibaba's Qwen3.8 at 40. The qualifier in Meta's headline is doing the work. This is a real release, a real engineering achievement, and not

AI 101 — How do AI video generators work?

An AI video generator is a model that starts with a grid of random noise and repeatedly refines it into a moving picture that matches your prompt — the same trick image generators use, stretched across time. Nothing is filmed, and nothing is stitched together from clips the model has seen. The pixels in the finished video were invented, frame by frame, all at once. Why it matters right now Kling 4.0, Kuaishou's flagship video model, went fully live on October 15 with up to 30 seconds of vid

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Read more

Deep Dive — America's top open model still trails China's best

AI 101 — How do AI video generators work?

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR