Hi Shopifreaks
Big thanks to everyone who entered my 300th Edition Giveaway last week! I also appreciated all your kind messages about how much you love the newsletter. To choose winners, I added all entries to a spreadsheet, assigned a sequential number to each, and then used a random number generator to pick 20 names.
If you received an e-mail from me this morning asking for your mailing address… congrats! You’re a winner!
As for everyone else, you may not have won a free t-shirt, but you can still purchase one on Amazon if you’d like. I priced these shirts at $13.38, which is my exact cost through Amazon Merch, meaning I make $0.00 per shirt (other than potentially a small commission from using my Amazon Associates link).
This project isn’t about making money for me. It’s just a fun way to spread some joy in our industry, while offering some inexpensive gifts for your nerdy e-commerce colleagues this holiday season.
Here’s the initial lineup of Shopifreaks T-shirts:
- ROAS Before Bros – my personal favorite.
- Hi First Name – celebrating the most recognizable mistake in e-mail marketing.
- CAC: Can’t Afford Customers – how is anyone supposed to make money?!
- Is This Still Available? – the all-too-familiar conversation for anyone who’s ever sold something on Facebook Marketplace.
- Be The First To Review – the painful moment when your new product goes live.
- Zero Sales Day – that one frustrating day when absolutely nothing sells.
Hope you enjoy the shirts! If you’ve got any suggestions for other t-shirts for e-commerce nerds, send them my way and we’ll continue to build the collection with your ideas. If I end up using your idea for a design, I’ll send you a free shirt.
In this week’s edition I cover:
- The security breach that hit Shopify apps
- Walmart welcomes third-party AI agents
- Wonder wants to build influencer restaurants
- Walmart’s brilliant membership bundle
- Google and Anthropic launch competing agents for SMBs
- Olo drops an app to compete with DoorDash and Uber Eats
- Meta offers exclusive promotional moments for brands
- Google is using your apps to personalize ads
- Meta lost big to New Mexico
- Amazon’s check is coming with President Trump’s name on it
All this and more in this week’s 301st Edition of Shopifreaks. Thanks for subscribing and sharing!
Stat of the Week
The number of Shopify apps is growing twice as fast as the number of stores, according to research from Barn2 founder Katie Keith. App store counts are growing 28% per year, while the number of Shopify stores grows just 14%, leaving 8.8 apps per 1,000 stores compared with 5.5 in July 2023.
1. The Internet went crazy, revealing a glimpse of what’s to come
Shopify and its app developers spent the last two weeks cleaning up after attackers hijacked the checkout buttons on an unknown number of merchants’ stores, beginning on October 10.
The attackers didn’t touch Shopify’s own systems. Instead, they broke into third-party apps and used them to send shoppers to fake payment pages built to steal card details. Within two weeks, about 480 apps had disappeared from the App Store, and Shopify still hasn’t publicly explained most of the removals. There’s also been at least one attack on a Shopify app unrelated to the hijacked checkout buttons.
I’m no security expert, but here’s what I’ve been able to piece together from vendor statements, developer trackers, and Shopify’s changelog. The full picture is still a bit blurry, so don’t take this information as 100% fact, though I’ve done the best I can to fact check.
What we know so far:
- Shopify itself wasn’t hacked. Attackers got into the back ends of third-party apps and changed settings that the apps’ scripts load from the vendors’ own servers. Nothing was written into merchants’ themes, so you won’t find anything weird digging through your code.
- The scam was a fake checkout. When a shopper clicked Checkout, the injected code sent them to a lookalike payment page, grabbed their card details, showed a payment error in some cases, then bounced them to the real Shopify checkout, leaving the shopper none the wiser.
- It started October 10. Developer Roman Zipp caught it while shopping on a mechanical keyboard store and traced it to Hulk Product Options. He’s been running a public tracker since.
- Three vendors have confirmed their apps were used: HulkApps, Logbase (Selleasy), and Swym, which later paused its app for every merchant after seeing more injection attempts.
- Shopify disabled Boost AI Search & Filter on October 17 over a “security issue.” Boost’s status page confirms the app was switched off for about a day and has since been restored, but Boost hasn’t said whether it was part of the checkout hijacking or something else entirely.
- Nobody knows the scale. None of the vendors has said how many stores were hit or how many shoppers handed over card numbers — even though Selleasy alone runs on more than 49,000 stores. We also don’t know whether this was one attacker or several running the same playbook.
- Wallet payments mostly dodged the issue. Shoppers who typed a card number were exposed, while Shop Pay, PayPal, and Google Pay weren’t hit the same way, since they log shoppers in through the wallet’s own system and never hand the card number to the page.
- Then came the “App-ocalypse.” About 480 apps left the App Store between October 15 and 23, including Rebuy, Privy, Consentmo, 17TRACK, Bold, and Rokt’s Aftersell and Upcart. About 41 have since come back.
- Delisted doesn’t mean hacked. Shopify hasn’t explained the removals, and most probably have nothing to do with the hijacking. Aftersell, for example, says it had a separate attack that exposed upsell settings for 22 stores, with no checkout redirects, and it’s still waiting to get relisted.
- Delisting doesn’t automatically switch an app off on stores that already have it installed. Those apps just can’t be installed by new merchants, which is extra devastating to app developers as we approach Black Friday.
- Shopify’s response so far: On October 20 it told app developers to harden their apps “this week” and added a warning before shoppers leave a store for an outside site. Three days later it walked that back, so now it only blocks sites it has flagged as malicious and asks for confirmation on ones it considers suspicious. More on that later in Section #9.
Meanwhile, in the same week…
While Shopify merchants were busy running incognito checkout tests, the rest of the Internet got its own scare. On October 21, the PHP Foundation pushed emergency patches for “php2shell,” a pair of critical flaws that let an attacker run code on a web server with one crafted request and no login. Palo Alto Networks counted 22.6M instances potentially exposed. PHP runs a huge part of the web — WordPress, WooCommerce, Magento — so the vulnerability is far-reaching.
Here’s the wild part: the bug had been sitting in PHP for 14 years, but Calif, OpenAI, and Wiz only recently each found pieces of it independently using various AI models.
October 2026 will forever be remembered as a glimpse of what’s to come.
AI just found a bug that humans missed for 14 years, and attackers have the same tools. The same week, The Information reported that a hacking group called CyberCartel used AI agents built on Claude, Grok, and Kimi to automate attacks on 400-plus Latin American financial firms with little human effort. The tools that find holes now also exploit them, and they’re getting cheaper and faster by the day.
2. Walmart opens its doors to approved third-party AI agents
Walmart announced a new framework that lets AI agents shop Walmart and Sam’s Club, while giving customers control over what the agents can buy, how much they can spend, and what data they can see. CEO John Furner unveiled it last Wednesday at the Masters of Scale Summit in San Francisco.
It comes in three parts:
- Agent Profiles: Each AI agent a customer uses gets its own profile inside their Walmart account, so its purchases show up separately from the customer’s own orders or from other agents’ orders. Customers can either approve every agent purchase one by one or let the agent buy on its own under a spending limit, such as $50, $100, or $200, which Walmart compares to adding an authorized user to your credit card.
- Walmart Agent Gateway: An MCP server that lets agents search inventory and place orders directly instead of having to click and scroll through Walmart.com like a primitive two-eyed human. It can also help an agent find your usual choice of sourdough bread, brand of milk, or other shopping preferences without handing over your entire purchase history. It might even recommend a comfy chair for the corner of your bedroom so you can watch while it pleasures your wife.
- Rules for AI agents: Agents can’t act beyond what the shopper intended, can’t see payment card information or change payment methods, and can only access the information needed for the task. Walmart says that rule-breaking agents will be blocked, but it hasn’t said how it’ll know when an agent crosses the line, especially on something as ambiguous as “acting beyond what the shopper intended.”
Here’s what else we know:
- Agent Profiles launch within the next month. The Agent Gateway is still being built and will start as a beta for companies and developers.
- It’s free for every Walmart account holder, with no Walmart+ membership required.
- Walmart’s normal returns process applies even when an agent screws up an order.
- Walmart may require additional approvals, even for purchases under an agent’s spending limit. I like that better-safe-than-sorry approach to permissions.
- Agents see the same prices as human shoppers. Walmart says, “We price products, not people,” and I guess that extends to agents in this context, which is good, because my agent is rich and spends frivolously.
- Walmart already has conversational commerce integrations with Meta’s Muse, ChatGPT, and Gemini, but it hasn’t said which agents will be approved for the Gateway or how it will decide.
Walmart’s approach to agent acceptance sits somewhere between Amazon’s and Shopify’s.
Amazon took the closed route. Last month it cut off Meta’s Muse from Amazon.com, saying Meta never told it Muse would be shopping its store and that the agent doesn’t identify itself while browsing. Amazon also sued Perplexity last year over its Comet browser’s shopping agent, and that legal battle is still ongoing after a court refused Amazon’s request to ban the tool in August. Amazon would rather shoppers use its own agent, Alexa for Shopping.
Shopify went the other way. Two weeks ago it let any AI agent shop its catalog and check out with Shop Pay, with no partnership or registration required. Agents just have to identify themselves with a profile URL.
Here’s how Shopify and Walmart differ:
- Shopify lets any agent in, while Walmart only lets in “approved” agents. I’m going to guess that Shopify changes its tune on this one eventually.
- Shopify agents can search more than a billion products across Shopify’s merchant base and even buy from non-Shopify stores using a single-use Shop Pay card. Walmart’s Gateway covers only Walmart and Sam’s Club.
- Shopify agents can read a shopper’s past orders to learn their sizes and preferences, while Walmart deliberately keeps purchase history away from the agent, just providing insights instead.
- Shopify agents pay in one of two ways: through Shop Pay’s stored wallet at Shopify stores, or with a single-use card number limited to one checkout at non-Shopify stores. Walmart agents never handle payment credentials at all, because Walmart simply charges the card saved in your account.
- Shopify already lets agent developers earn a 0.3% commission through Promoted Placements. Walmart hasn’t announced any way to pay agents or to sell them ad placements, yet. Given that Walmart runs an affiliate program, this is likely coming.
- Both solutions default to requiring approval for every purchase and both offer spending limits, but Walmart’s separate Agent Profiles give shoppers a cleaner record of what each agent bought.
Walmart is betting that shoppers want agents, but not ones they can’t control. Shopify is betting on open access. Amazon is betting that if it keeps its gates shut, you’ll just use its agent instead, and to be fair, you probably will.
🔥 Partner News
ReFiBuy launched AdGentic, a show and newsletter about advertising in the age of AI agents, hosted by retail media commentator Kiri Masters of the Retail Media Breakfast Club podcast. Masters’ debut episode and essay, “Ensh*ttification is Coming to Personal AI Agents,” argue that agents like Meta’s Muse and OpenAI’s Dots feel great right now partly because nobody is selling through them yet, and that Google, WhatsApp, and ChatGPT all added ads despite early promises or warnings against them. She argues ads are still a better business model than subscriptions, which most users won’t pay for, or transaction fees, which reward agents for steering purchases, as long as they’re clearly labeled and not disguised as the agent’s own recommendations.
3. Wonder wants to turn every influencer into a restaurant owner
Wonder, the owner of Grubhub and Blue Apron, is launching Wonder Create on November 17, a new service that lets influencers, entrepreneurs, and brands run their own “restaurant” out of its commercial kitchen locations for $10 a month and keep a share of the revenue.
Creators enter a prompt for a description of the restaurant they want, and the AI generates a name, logo, recipes, and prices, which they can tweak by editing the prompt. From there, Wonder’s kitchen staff and robots cook and deliver the food, so creators don’t need to put up capital, pay overhead, or oversee a kitchen to launch their dream restaurant and post “OMG, so good!” videos about it.
The catch? Every dish has to come from the roughly 700 ingredients Wonder already stocks, and the influencer is just a face on the menu while Wonder controls the kitchen, the staff, and the delivery.
This is quite possibly one of the worst ideas I’ve ever seen, and I’ve seen my fair share of terrible ideas. Where to begin…
- Confusion for consumers. It’s hard enough to navigate truthful restaurant reviews and find hidden gems in your city. Now the restaurant market is about to get flooded with AI slop ghost kitchens that are all just the same pig wearing different lipstick.
- The sterility of restaurant dining. It’s bad enough that most restaurants are already getting all their food from the same supplier, now it’s all going to get cooked by the same chef?
- Been there, done that. This is the restaurant equivalent of an influencer launching their own beauty line, nutrition label, or coffee brand by simply sticking their logo on a white labeler’s product. Let’s not bring that low-effort, be-your-own-boss, slap-your-name-on-it model to restaurant delivery.
- It’s misleading. Now if a diner orders food from MrFeast’s new restaurant, hates it, and says, “I’m never ordering from there again,” they may unwittingly do so from Jim Fartdashian’s ghost kitchen the very next day.
When I think about a list of my favorite restaurants, it includes the BBQ spot that slow cooks its pork for 12 hours, the hole-in-the-wall seafood restaurant that makes the best ceviche and encebollado I’ve ever had, and the Thai restaurant whose owner used to run restaurants in Chiang Mai.
What I don’t think of is Applebee’s, Chili’s, or other casual dining restaurants that reheat the same Sysco appetizers and serve them up in a slightly different front-of-house — effectively the same bland, pretentious, commercialized experience that Wonder is about to bring to delivery at scale.
I hate it. If this takes off, we’re going to need a new law that requires these “influenceraunts” to disclose that they’re merely affiliates of a larger ghost kitchen, with a label like “Cooked by Wonder” or “Don’t Eat Here.”
4. Walmart bundles Walmart+ and Sam’s Club Plus into one $120 membership
Walmart pulled the ultimate membership move last week, bundling Walmart+ and Sam’s Club Plus into one $120 annual membership that gives customers free delivery from both chains, member prices at Sam’s Club, a choice of Peacock or Paramount+ (the Walmarts of streaming services), gas discounts, and 2% cashback in Sam’s Cash.
Anyone who already has Sam’s Club Plus gets Walmart+ at no extra cost, while annual Walmart+ members can switch to the bundle for $22 more a year, and people already paying for both can get a prorated refund.
Walmart CEO John Furner said:
“We’re giving people a reason to expect more from their memberships. For $120 a year, we’re raising expectations for what a membership can deliver: more money in their pockets and more time to call their own.”
He went on to pitch the bundle against paying $139 a year for delivery from a retailer and $130 for a warehouse club, which are the exact prices of Amazon Prime and Costco’s Executive membership. Burn!
Uh oh, are Costco and Amazon about to team up to compete against this new Walmart Membership Goliath?!
Not likely, LOL. The two companies have very different philosophies when it comes to corporate culture, employee compensation, and how they treat suppliers. Amazon would have to go after BJ’s or some other smaller warehouse chain, same as it went after Whole Foods a decade ago instead of attempting to acquire Kroger.
It’s a brilliant move by Walmart and Sam’s Club to combine the two memberships. The company recognizes that the memberships themselves aren’t about revenue, but about creating sticky shopping habits and offering a value that’s hard to rival anywhere else.
I’ve criticized Amazon heavily for gutting the value of their Prime membership over the past few years. It’s been death by a thousand cuts to the point that I called Amazon Prime “an ad-supported trial of the company’s premium services.” All the while, Amazon Prime has gotten 40% more expensive during the past decade, and technically 100% more expensive if you include paying for the ad-free add-on to Prime Video.
Such a shameful downfall for a retailer that effectively built the modern shopper membership.
5. Google and Anthropic go after small business owners with competing AI assistants
Google and Anthropic launched competing AI tools the same week that help small business owners run their online stores, manage their finances, and market their brands. However, the two companies are taking different approaches, at least for now.
Google added a small business hub to the Gemini app where owners can work on branding, review finances, and see customer insights in one place. Owners can connect it to Google’s own business tools plus outside apps like Shopify, Canva, and Xero, and Gemini saves their brand kit so every image it generates matches the company’s look. It can also turn receipts and spreadsheets into charts and write product descriptions, newsletters, and social media posts. However, it’s still a prompt-and-respond tool — meaning, you ask, and it answers. It doesn’t take the initiative on its own.
Anthropic launched Claude Duet, an agent app that flags what needs your attention without being asked, starting with a morning brief from your connected apps and check-ins every few hours. Unlike Google’s tool, Duet can take action itself on tasks like managing a Shopify store, running payroll in Gusto, chasing unpaid invoices in QuickBooks, and buying things with Shop Pay, asking for approval before it sends, pays for, signs, or deletes anything. Duet is a fully agentic tool — meaning, it doesn’t wait for you to ask, and it doesn’t stop at just giving you an answer. It can go beyond suggestions and actually perform the work.
Google’s hub is rolling out worldwide to every Gemini user, including free ones, while Duet is a waitlist-only research preview for US subscribers on Claude’s Pro and Max plans.
The Different Approaches: Google is launching a smaller set of tools to everyone, while Anthropic is going fully agentic from day one, but only for a small group of users. However, I’d imagine that both paths eventually end up in the same place, with tools that offer business owners a similar set of functionalities and benefits.
If both of those tools sound familiar, you might be remembering that OpenAI entered the space in July with ChatGPT Work, an agent mode that handles multi-step tasks and scheduled jobs, which it pitched to small businesses with a dedicated program called ChatGPT for Small Business a couple of weeks later.
6. Olo launches a consumer app without menu markups to rival DoorDash and Uber Eats
Olo launched a nationwide consumer app for restaurant pickup and delivery, letting people order from more than 250 brands, including Five Guys, Waffle House and Dave’s Hot Chicken, at the same prices they’d pay in the store. Unlike DoorDash, Uber Eats, and Grubhub, which take a 15% to 30% commission on every delivery order that many restaurants cover by raising their prices on those apps, Olo doesn’t take a commission from restaurants. It instead charges customers a flat 5% fee on the order, plus delivery when a third-party courier brings the food.
Quick History About Olo: The company launched in 2005 as GoMobo, letting people pre-order coffee by text message so they could skip the line, and later pivoted into offering ordering software that restaurants could use on their own websites and apps, instead of passing their customers to third-party ordering apps. I was sold on the concept from day one, bought a bunch of stock in Olo at $31 a share just after its 2021 IPO, which subsequently tanked, and then Thoma Bravo took the company private last year at $10.25 per share, so I never got to experience the upside of my investment. I’m secretly still rooting for them though.
Who delivers the food when you order through Olo?
Even though Olo itself isn’t taking a commission from restaurants, aren’t DoorDash, Uber Eats, and other delivery apps getting their slice either way by delivering the food?
Olo’s Dispatch tool pulls quotes from the delivery services available to each restaurant and picks the best one, so the driver often still comes from DoorDash or Uber through their white-label services, DoorDash Drive and Uber Direct, or from the restaurant’s own staff. So yes, DoorDash and Uber still get paid in some cases, but through a delivery fee rather than a 15% to 30% cut of the order, and without the menu markups. Olo also opens the door for other delivery services to compete against those major players for delivery of the order.
More than 800 restaurant companies already use Olo’s software to take orders on their own websites and apps, and Olo wants to eventually bring every one of them onto the new app, along with restaurants that aren’t customers, meaning restaurants won’t have to use Olo’s ordering system to connect their menus and sell through its app.
It’s about time that someone knocked these delivery apps off their pedestal.
I’ve been a proponent of a direct-to-restaurant solution for years, as it’s unfortunate what Uber Eats, DoorDash, and the other delivery apps have done to the cost of restaurant delivery. Back in my day, you’d just call the restaurant directly and they’d send their guy. Now, some rando delivers your food, eats some of your fries, and takes a sip of your drink, and you pay a premium for it.
Will Olo be the disruptor that the industry needs? It’s certainly got the technology and the financial backing to do so. Now we’ll just see how they execute.
7. Meta launches Ads Super Burst for exclusive promotional moments
Meta introduced Ads Super Burst, a new ad format that lets brands reserve between 3 and 24 hours of Facebook and Instagram inventory around a product launch or a big event, with the option to block competitors in the same category from advertising during the same window. For example, an advertiser may purchase an Ads Super Burst to promote a professional sports game or awards show to hype up the event, or to dominate the feed during the first few hours of a new product launch.
Meta wrote in its announcement:
“The brands people remember are often the ones that show up inside the most memorable moments: the game everyone is discussing, the show everyone is watching, and the trend that goes from niche to everywhere in a weekend… Most brands have a few moments a year that matter more than the rest.”
Meta says the median brand in its tests reached 34M people per window, or about 14% of U.S. adults online in a day, which is pretty impressive. In comparison, Super Bowl commercials are estimated to reach upwards of 120M viewers, so an Ads Super Burst from Meta gets you more than a quarter of the way there.
Brands book the slots through Meta’s Reach Planner, which quotes estimated reach and price in advance, and the format is opening to all advertisers in phases this month. Meta is also testing a second reserved placement in the U.S. called First Impression Ads, which gives one brand the first ad slot in the feed for a whole day.
In April, I wrote on LinkedIn:
“There’s an advertising trend happening in front of our eyes right now… Brands don’t want to be AN advertiser, they want to be THE advertiser. They want to take over a user’s entire session, not appear alongside other forgettable one-off ads as the user scrolls. It’s the only way to filter through the advertising noise that the platforms themselves created.”
This was made evident by the launch of Snapchat’s Total Snap Takeovers, TikTok’s Logo Takeovers, Prime Time sequential ads, and TopReach, and Spotify’s Branded Playlist Takeovers. It was obvious that Meta would soon offer a similar format, not just because they’re notorious for blatantly copying the most popular features from all their competitors, but also because exclusive attention is where advertising is headed.
8. Google to use your Gmail and Photos for personalized advertising in AI Mode
Google started testing ads in AI Mode for people who opt to link their Google apps to Personal Intelligence, referencing summaries, excerpts and inferences from these sources to personalize their experience and ads “so they are even more helpful for you.”
Google has been running ads in AI Overviews and AI Mode for over a year, so the ads themselves aren’t new, but it had originally promised to keep users who connect their apps ad-free. Now it just promises that it won’t show personalized ads based on sensitive categories and that advertisers won’t get access to anyone’s connected apps. The privacy goalpost keeps moving.
What’s funny / scary / disturbing is that Google says that “ads will not be based directly on your Gmail inbox or your Google Photos library” — just from the “summaries” and “excerpts” derived from those sources.
So what’s the difference?
For you or Google, there really is no difference. Google AI reads your e-mails and views your photos, creates a summary of its findings, and then personalizes your ads based on that summary. If Morty Smith were here, he’d say, “That just sounds like privacy invasion with extra steps!”
If I start e-mailing my parents to plan our upcoming trip to Miami, and Google reads those e-mails and creates a summary that says, “Paul is planning a trip to Miami” — it’s arrived at the information either way. In fact, I don’t really see how that “extra step” is any different from how it has summarized users’ search and browsing histories for years and served ads based on it. Google isn’t referencing 20,000+ search queries every time it serves an ad to you. It’s often just referencing the persona it has built for you based on those queries.
Remember back in 2017 when Google made a big show of promising that Gmail content “will not be used or scanned for any ads personalization”? Well, the company quietly updated that blog post on October 16 with a note saying that data from connected apps may now be used to show you ads. It appears that old habits are hard to break.
What are your thoughts on Google serving ads based on your Gmail and Photos? Hit reply and let me know or join the conversation on LinkedIn.
9. Other e-commerce news of interest
Meta was ordered to pay New Mexico $25B for misleading Facebook users about how it protected their data after the Cambridge Analytica scandal. The ruling by Judge Francis Mathew comes a month after a jury found over 43M violations of state consumer protection law, with New Mexico seeking $35B to $40B in damages, while Meta argued for a cap of $3.4B and said anything 10 times that or more would violate the Constitution. The jury had agreed that Facebook misled users about how it policed the quiz app that harvested roughly 87M profiles on behalf of Cambridge Analytica, but rejected the state’s case over how it took down harmful posts. Meta chief legal officer C.J. Mahoney said the ruling makes it “all but impossible to do business” in New Mexico and that Meta plans to appeal. The ruling comes two months after Meta agreed to pay up to $17.1B to 47 states to settle a trial over teen social media addiction.
OpenAI is turning on creative enhancements by default for every self-serve ChatGPT Ads campaign, starting October 30, according to an e-mail sent to advertisers. The feature lets OpenAI rewrite headlines and descriptions to fit a user’s conversation or interests, translate ad text, resize images, and send clicks to a different page on the advertiser’s site, based on the original ad and landing page, similar to Google Ads’ AI enhancement features. The variations can be generated in real time and run without the advertiser approving each one, though OpenAI’s ad terms say it isn’t responsible for errors in any AI-generated creative its system produces. Well, isn’t that convenient?
In other ChatGPT Ads news… OpenAI expanded ads to 26 additional countries, bringing the program to 89 markets in total including Argentina, Chile, Colombia, Peru, Nigeria, South Africa, Ukraine, and Cambodia. Advertisers in those markets can now buy ads through technology partners or set up campaigns themselves in OpenAI’s Ads Manager, reaching users on ChatGPT’s Free and Go plans. OpenAI said that more than 100k advertisers have bought ads since the program launched a little over eight months ago. Lastly, OpenAI is asking brands for $300k minimum commitments and $40 CPMs to run ChatGPT ads next to AI-generated images in its pilot that I reported on two weeks ago. The minimum is higher than the $250k it asked for in its first pilot in February, even though brands had trouble spending their commitments because ChatGPT didn’t have enough inventory to fill them.
Speaking of AI-generated slop… Amazon is testing AI-generated main images in search results that place products in staged lifestyle scenes instead of on white backgrounds, according to a screenshot taken by Talal Asad, president of Amazon consultancy Brand Together. The screenshot showed Amazon putting bath products into AI-made bathroom scenes, even though Amazon requires sellers to use a pure white background for main images. Asad argued the test will hurt rankings for the affected brands because shoppers need to be able to read a product’s label in the main image. Personally, I’m all for primary images without white backgrounds, which work for some products and not for others, though sellers should have the choice of doing either themselves, not just Amazon’s AI.
Apple published a blog post explaining that it began blocking certain domains on its browsers earlier this month because they link a person’s e-mail, phone number, IP address, and device details to track them across sites, a practice it calls “identity resolution.” The blocks began without warning and hit The Trade Desk, LiveRamp, ID5, Permutive and Audigent, briefly stopping The Trade Desk from serving any ads at all until Apple fixed it. Apple said people who give a site their email to log in don’t expect it to be passed to other companies for tracking, and it is blocking entire domains so those companies can’t get around the rule. Ad buyers told Business Insider they expect less reach and higher prices on the open web as more people update to iOS 27.
ChatGPT referred only 0.62% as many visits as Google organic search across 100 websites, according to SEO consultant Marie Haynes, who analyzed a year of GA4 data covering more than 145M sessions. All AI chatbots combined only raised that percentage to 0.73% of Google’s organic traffic, which indicates that ChatGPT is leading the way among chatbots in referral traffic. The study notes that ChatGPT visitors convert more often, at 13.8% versus 3.02% for Google organic visitors on the 55 sites that tracked conversions, but Google organic still produced 27 times more conversions than all AI chatbots combined. Haynes, who called GEO a “distraction,” found that ChatGPT referrals stopped growing quickly in August 2025 and have since declined.
Heineken and Amazon launched Pubscription, a free Fire TV app now in a UK pilot that lets adult users pick a bar or pub the same way they’d pick a show on a streaming service. Users can sort through participating Star Pubs venues by “genre,” such as sports pubs or quieter neighborhood spots, and then invite friends to meet them at whichever one they choose. Each month, subscribers get four vouchers on their phones, each good for one Heineken or Heineken 0.0, which they can redeem once per week. Amazon Ads’ creative unit, Brand Innovation Lab, built the app with Heineken, which owns the Star Pubs chain. Very novel idea! However, I’m curious if it’ll actually lead to more sales, or if users will simply cash in their voucher for a free Heineken and then spend the rest of the evening drinking good beer. I guess it’s a win for the participating pubs either way.
Commerce partnered with Cloudflare to add WebMCP support to BigCommerce stores, allowing AI agents running in a shopper’s browser to search a store’s catalog, compare items, and build a cart. The agent doesn’t complete the purchase but instead hands the shopper a link to check out on the store’s own site, with prices and stock pulled live from the merchant’s catalog. Technically, BigCommerce could have built WebMCP support into its storefronts on its own, but it was faster and more efficient to go through Cloudflare. Since BigCommerce stores already run through Cloudflare, it can inject the script at the network edge for every store at once, so no merchant has to edit a theme or install an app. The feature is live by default for eligible merchants, who can turn it off in a new AI settings menu.
Meta’s Muse agent was set up by over 10M people across the US and Canada in its first six weeks, and about a quarter of them use it daily, according to Mark Zuckerberg on the company’s Q3 earnings call. Shopping is the agent’s most popular use, with over 25% of users using it to make purchases with an average order value above $200, and Zuckerberg said purchase volume in the first week ran so far past expectations that Stripe raised its credit limits. Muse is expensive to run, so Meta has held off on a global rollout and plans to start charging for it sooner than it usually would with a new product. Zuckerberg said the company needs compute efficiency to improve more than tenfold over the next year for Muse’s economics to work at scale.
You know all those idiots who looted stores in France earlier this month during the protests? Well, not only were they dumb enough to immediately list the items for sale online, some of them were even stupid enough to tag them with a #October17 hashtag! Now that’s funny! French ministers subsequently held a video call with Vinted, Leboncoin, Amazon, eBay, Meta, and TikTok, demanding that they pull listings for goods looted from over 3,000 businesses during the protests, after cosmetics, skincare, and handbags began appearing on resale sites within hours of the looting. The platforms committed to flagging suspicious listings for police, watching for red flags like unusually low prices, bulk lots, and items still wearing security tags, and reminding buyers that purchasing stolen property is illegal.
Amazon gained additional share of US food sales this year after it began offering same-day delivery of fresh groceries in thousands of US cities and towns, while traditional grocery stores lost share, according to NielsenIQ data. Amazon’s seafood sales jumped 888.2% YoY over the 12 weeks ending in late August, while pork sales rose 1,100% and egg sales rose 750%. The shoppers buying fresh food on Amazon skew older, often have household incomes above $100,000, and already order and spend more on Amazon than the average user. Two weeks ago I reported that Whole Foods CEO Jason Buechel said groceries, which brought Amazon more than $150B in sales last year, can turn the company from the “everything store” to the “everyday store” by getting customers to shop more often across the whole site.
Flipkart is testing a food-delivery service with more than 8,000 of its employees before a public launch that would take on Zomato and Swiggy, according to TechCrunch. The service, called “Eat-in by Minutes,” lists restaurants from the government-backed ONDC network alongside ones Flipkart signed directly and uses a dedicated fleet of Ekart riders to deliver the meals. Flipkart told restaurants it plans to open the service to everyone before November and expand into five or six cities before the year ends, according to Sagar Daryani, who heads the National Restaurant Association of India. Daryani said Flipkart offered restaurants a base commission of roughly 10%, while Zomato and Swiggy typically take 20% to 25%. That’ll be great for restaurants! I hope it forces competitors to bring their fees down too.
Google search advertisers paid 4% more per click in Q3 than a year earlier, up from 1% growth in Q2, according to Tinuiti’s latest Digital Ads Benchmark Report. Total spending on Google search ads grew just under 14%, the same pace as the prior quarter, but advertisers got fewer additional clicks for their money. Click prices had barely moved for the previous four quarters, which Tinuiti attributes to Amazon leaving almost every US Google Shopping auction in July 2025 and thinning out the competition. Tinuiti also tied the third-quarter increase to an August 17 Google change affecting campaigns that use target-based bidding and hit their budget limits.
TikTok shifted its ad spending toward YouTube, Pinterest, and app install networks after Meta stopped allowing all ByteDance ads on its platforms earlier this month, according to Sensor Tower estimates. Well, where else can they really go at that scale? YouTube took 17% of TikTok’s ad budget in the week after the ban, compared with 9% in early October, and Pinterest’s share rose to 9% from 3%. The biggest chunk, 58%, now goes to app networks such as Google’s AdMob, versus 19% before Meta shut TikTok out. TikTok had been one of Meta’s biggest advertisers before it was banned across all Meta platforms in the US, Canada, Egypt, Indonesia, Japan, Thailand, and Vietnam. Doesn’t Meta kind of need that money right now, given that it just raised its AI spending commitments to $825B? Oh, and don’t forget all those multi-billion dollar settlements and fines it just got hit with.
eBay will begin automatically enrolling eligible US sellers in eBay Balance on November 4, holding their sales proceeds on its platform instead of sending the funds straight to their bank accounts, according to Liz Morton at Value Added Resource. Sellers on daily, weekly, or monthly payouts will be switched to on-demand withdrawals and have to restore their schedule by hand, and the new terms let eBay drop scheduled payouts entirely in the future with notice. eBay’s terms already let it pool seller funds into investments and keep the interest, so more money sitting in balances could mean more interest for the company. Honestly, what a piece of shit move. I feel like eBay’s only move lately has been to squeeze every possible nickel of margin from sellers, rather than grow the pie.
Meta cut the word “train” out of its new commercial terms, which take effect October 30, after publishers objected that the terms let the company crawl any website linked to a business page and use what it found to build AI models. The revised terms instead only let Meta pull updates from those sites to “develop” its advertising and business products. A Meta spokesperson said the crawling language only ever covered its ads and business tools, and that site owners can still block its crawlers with robots.txt, but who believes anything Meta says at this point? Legal experts told Nieman Lab that the word “develop” can still cover training, and that Meta could ignore robots.txt because the terms never promise to honor it. Publisher groups in the US, UK and Europe are asking regulators to stop the terms before they take effect.
Ready for some more Meta bullshit? Meta paused Muse’s conversation-sharing feature after over 6,800 chat links were discovered to be searchable and accessible to the public. Muse put each conversation’s title in the link’s subdomain, so once a recipient opened it, the address landed in passive DNS databases that anyone can search. In the few hundred links that security firm Claroc opened, it found a small-business loan application, a company revenue review prepared for its CEO, a bank fraud alert showing part of an account number, and an office’s full staff vacation calendar, among other sensitive chats. How is it that literally every fucking major AI company has made the same mistake? Is it that hard to keep shared conversations private between creators and recipients? It’s purely negligence at this point. No other way to describe it.
Shopify began stopping visitors in their tracks after clicking a link to an outside website, displaying the address they were headed to, and asking whether they’d like to continue or return. The company added the prompt a week after attackers changed settings in apps from HulkApps, Selleasy, and Swym to redirect checkout shoppers to fake payment pages, but its changelog doesn’t mention the attacks. After the change, merchants immediately began complaining in Shopify’s community forums that the alerts popped up on intentional links to partners, and one said Shopify Support told them it couldn’t be turned off. Three days later, Shopify narrowed its external link confirmation so that it only appears for websites Shopify considers suspicious. So what happens now if/when a normal non-dangerous website gets incorrectly flagged as suspicious? Are they fucked or does Shopify offer a way to appeal the label? Shopify only says to contact Support with a screenshot if your website gets blocked, without indicating what comes next in the process.
Just Eat paused a trial program that automatically produced AI-generated menu photos for UK restaurants that hadn’t uploaded their own, after the images came out horribly wrong. For example, one Northern Ireland restaurant’s soda bread breakfast sandwiches were shown as sausage, bacon, and fried eggs floating in glasses of soda. In other examples, a burger came out shrunken and sitting inside a sauce pot, sushi rolls were shown stuffed with cheese, and wine glasses were filled with chunks of watermelon. Damn, I’d hate to guess what images the AI came up with for Spotted Dick, Toad in the Hole, and Cock-a-Leekie! (Yes, those are real dish names in the UK.) Just Eat apologized and said it’s pulling the inaccurate photos and won’t resume the rollout until the tool improves.
Amazon renamed Kindle Unlimited to Kindle+ and expanded the audiobooks in the $12 per month subscription to include AI-narrated versions of self-published books. Authors in KDP Select, Amazon’s exclusivity program for self-published e-books, can now turn their books into audiobooks for free with Amazon’s Virtual Voice AI narration tool (much to the dismay of voice actors) and have them automatically included in both Kindle+ and Audible Plus. Virtual Voice, which is still an invite-only beta limited to the US marketplace, pays authors a 40% royalty on its audiobooks, the same rate Audible pays on exclusive human-narrated titles. They should partner with ElevenLabs so I can get my book narrated by Matthew McConaughey. Alright, alright, alright.
Speaking of AI voices… Block released a research preview of Dessa, a speech model from its AI lab that can listen and talk at the same time, as well as pick up sarcasm, laughter, sighs, and sadness from speakers. You should listen to some of the recordings of Dessa in action because it’s wild! It’s the most realistic AI voice I’ve ever heard. Dessa was trained from scratch on recordings of people talking, with no text or pretrained models involved, and runs on-device on a MacBook Pro with an M4 Max chip or better. Block is letting a handful of AI safety researchers test the model first to find risks it hasn’t anticipated. Square already offers AI voice ordering to its sellers, but Block hasn’t said whether Dessa will power it.
Pirate Ship said it will no longer e-mail customers about routine carrier rate changes because the carriers have started changing prices several times a week as they move to algorithm-set dynamic pricing. The company said in an e-mail to sellers that it will still flag major service changes, holiday surcharges, and shipping deadlines, but otherwise customers have to check its live-rate calculator. The move reverses Pirate Ship’s own policy, which promised customers advance notice of carrier rate increases so they could prepare, and leaves sellers who offer free or flat-rate shipping to discover cost increases on their own, often after they’ve already priced their products.
In lawsuits this week…
- Australia’s eSafety Commissioner is preparing to sue Meta, TikTok, Snap, and YouTube for failing to keep under-16s off their platforms after the country’s social media ban took effect last December.
- Also in Australia, Meta was hit with a class action on behalf of Australians who say Facebook and Instagram harmed them as minors through addictive features like infinite scroll and autoplay and by showing them harmful content.
- In other Meta troubles, Rhode Island became the first US state to sue Meta over scam ads, alleging it knowingly profited from fraudulent ads on Facebook, Instagram, and WhatsApp and that its AI ad tools helped scammers build them.
- Temu agreed to pay Texas $3M to settle a lawsuit claiming it exposed Texans’ data to the Chinese government, and must now keep that data on US servers that people and businesses in China can’t access.
- A judge gave final approval to Google’s $68M settlement over claims that Google Assistant devices recorded conversations without a “Hey Google” prompt, with lawyers taking about $22.7M and device buyers getting roughly $18 to $56 each.
- Meta was ordered to identify every employee and executive who approved downloading pirated books from shadow libraries to train its AI, after unsealed internal messages showed the decision was escalated to “MZ” for final approval — initials that likely refer to “Mark Zuckerberg.”
- Anthropic got hit with two more music lawsuits from Reservoir and from Elvis Presley’s rights holders, who say Claude still reproduces lyrics despite its guardrails.
- Amazon was sued by an employee who alleges it failed to properly investigate her report that a coworker raped her on a 2023 business trip, and that HR told her to keep the allegations confidential in violation of Washington law.
- Google won dismissal of a lawsuit claiming its Contact Center AI illegally listened to customers’ calls to Home Depot, though the judge is letting the plaintiffs refile.
In layoffs this week…
- eBay is cutting about 9% of its workforce, its second round of layoffs this year after eliminating roughly 800 jobs in February.
- Automattic began laying off staff at WordPress.com and its VIP enterprise unit, a month after a failed board attempt to oust CEO Matt Mullenweg ended with the entire board replaced.
- Loop, the Shopify returns platform, laid off an undisclosed number of employees in a reorganization that cuts management layers and moves more engineers onto its largest merchants.
- American Express is cutting about 5% of its workforce to automate more work with AI, and said it will only let AI agents into its systems through approved API or MCP channels.
In corporate shakeups this week…
- Shopify CTO Mikhail Parakhin is leaving the role after about two years and will stay on as an advisor, with no successor named yet.
- Citi hired former Google engineering director Sebastian Gnagnarella as head of applied AI and agent factory to build the bank’s AI agents, its second Google executive hire this year.
- Meta chief global affairs officer Joel Kaplan is stepping down after 15 years, months after former Trump administration official Dina Powell McCormick became Meta’s president.
- Amazon plans to hire more than 300,000 holiday workers, about 20% more than last year, even as overall seasonal retail hiring is expected to hit its lowest level since 2008.
- Amazon also pledged to fill 30,000 entry-level UK warehouse jobs for young people by 2030, after cutting about 30,000 corporate roles over the past year.
- Zalando hired Tom Christenson, who launched Amazon’s Seller Fulfilled Prime program, as SVP of its B2B business, combining its Zeos, Tradebyte, and Scayle units under one team.
- Flipkart promoted Renuka Singaram to SVP overseeing Minutes, its quick-commerce business, along with its supply chain for smaller items.
- Anthropic is offering up to $345K for a human executive communications writer, even as its CFO says AI writes 90% of the company’s code.
- Anthropic named Harvard astrophysicist Christopher Stubbs to its Long-Term Benefit Trust, which elects most of its board, ahead of an IPO that could come in November.
- Several Google DeepMind employees resigned over Google’s AI contracts with the US and Israeli militaries, saying Google has no say over how the Pentagon uses its models.
🏆 This week’s most ridiculous story… The refund checks from Amazon’s $2.5B Prime settlement credit “President Trump’s Federal Trade Commission,” making them the first FTC refunds ever to name a president, according to Politico. The phrase appears not just once, but three times in the notice Amazon sent with each check, which former FTC general counsel Stephen Calkins called “unseemly and unfortunate,” and former FTC chairman Bill Kovacic said could hurt the agency’s professional reputation in later cases. The FTC said the settlement belongs to “Trump’s FTC” because he heads the executive branch, but the Biden administration was the one that actually brought the original lawsuit in 2023. You know, I’d be fine crediting the President with “Trump’s Universal Healthcare” or “Trump’s Consumer Protection Laws,” if they were to exist, but a $51 Amazon settlement check? Really? What’s next, Trump’s 5-Cent Bottle Deposit Refund?
10. Seed rounds, IPOs, & acquisitions
Whop, a marketplace and payments platform for selling digital memberships, courses, and software, raised $200M in a Series D round co-led by Sequoia and Shopify Ventures at a $5.1B valuation, just eight months after Tether invested $200M in the company at a $1.6B valuation. The fresh capital will fund Whop’s expansion abroad and research into payments and advertising. Its recent launches include Business Blueprints, a tool that sets up a ready-to-sell business in minutes, a CLI that lets AI agents run a business through Whop’s API, and a recommendation engine called Economic Intelligence that suggests each business’s highest-ROI next move in exchange for an extra cut of the revenue it generates.
Salesforce agreed to acquire Braze, a customer engagement platform that sends personalized messages across apps, email, SMS, and WhatsApp, for about $5.66B in enterprise value. Salesforce is paying $47.50 per share in cash, a 58% premium over last Monday’s close of $30.04, but still well below the company’s 2021 IPO price of $65. Salesforce plans to add Braze’s tools to Marketing Cloud once the deal closes in its fiscal 2028, pending shareholder and regulatory approval, however, co-founder Bill Magnuson will stay on as CEO, and Braze will keep running under its own name.
Cuttable, a platform that turns a brand’s existing images and video into ads for Meta and TikTok, raised $20M in a Series A round led by Airtree Ventures at a valuation above $100M. Cuttable’s software pulls a brand’s Meta posts and Shopify order data, builds new ads from them for a human to approve, and records the message each ad was testing, such as a discount, so future ads build on what promotions and verbiage worked. The company will put the funds towards hiring engineers, adding Google to its supported channels, and expanding in the US, a market that already brings in 40% of its revenue.
Redo, a commerce platform for returns, order tracking, fulfillment, and AI marketing, acquired Relay Goods, a certified pre-owned online marketplace that sells discounted, brand-name running shoes, for undisclosed terms. Redo started out handling returns and exchanges for online stores and now also sells order tracking, package protection, customer service, and marketing tools to more than 4,100 brands. Relay buys shoes that customers sent back because they didn’t fit, along with excess stock, from brands and authorized retailers, then cleans and inspects the shoes and resells them with a warranty. Redo says putting Relay’s resale operation in front of its merchants will help more brands make money from returns and leftover inventory.
Azoma, an AI search optimization platform that helps brands get recommended by AI shopping agents, raised $11.2M in a Series A round led by Puma Growth Partners. Azoma’s software tracks how products show up across ChatGPT, Gemini, Alexa for Shopping, Walmart’s Sparky, and Instacart’s Clementine, as well as how brands perform across agentic commerce protocols like Universal Commerce Protocol and Agentic Commerce Protocol. The company plans to use the funds to build attribution tools for large consumer brands, including its existing clients L’Oréal, Unilever, Mars, Beiersdorf, and Reckitt, as well as expand to AI personal assistants such as ChatGPT Dots and Meta Muse.
Stord, an omnichannel fulfillment provider that helps independent merchants compete with Amazon on delivery speed, acquired AMS Fulfillment, a third-party logistics company that warehouses and ships orders for beauty, wellness, and DTC brands, for an undisclosed nine-figure sum. The deal, which is the largest of Stord’s 11 acquisitions, adds 10 warehouses in the Los Angeles, Salt Lake City, and Philadelphia areas and the more than 150 brands whose orders AMS ships, pushing Stord’s network past 100 owned and partner sites. Stord also picks up the automated sorting systems already running in some AMS warehouses, which it plans to build on through Stord Labs, its robotics and AI research unit.
CapitalOS, a platform that lets software companies offer their own branded corporate cards, digital wallets, and bill pay, raised $21.9M in a Series A round led by LGVP. CapitalOS runs the underwriting, compliance, fraud monitoring, and support behind those products, and it also funds the card spending itself, so a platform doesn’t have to raise or set aside capital to offer credit. The company says its transaction volume grew 1,000% over the past 18 months, with more than 25,000 cardholders across dozens of platforms.
DeepSeek, a Chinese AI lab that develops low-cost, open-weight LLMs, raised 100B yuan ($15B) in a round that included Tencent and CATL among the biggest investors, according to Bloomberg. The round is one of the biggest ever for a Chinese company and comes just four months after a roughly $7.4B round in June that valued DeepSeek between $52B and $59B. DeepSeek plans to spend the funds on building out data centers and developing new open-source models as it prepares for an IPO.
Elon Musk posted on X that next year he plans to hand each American child one SpaceX share from his own stake, which, with roughly 72M Americans under 18, would come to about $10.5B in stock based on today’s value. Musk had replied a day earlier to Anthropic’s plan to set aside 5% of its shares for a public benefit fund, saying the shares should go straight to the public. He hasn’t said which children qualify, how they would claim the shares, or whether they’d have to hold them for a while before selling. However, in a later post, Musk hinted the shares could go through Trump Accounts, which under rules Treasury published in September can receive donated stock that generally can’t be sold for five years, a move that would also help stabilize SpaceX’s share price during that period.
OpenAI is trying to lock in commitments for a roughly $30B funding round at a more than $1.4T valuation, with existing backers SoftBank and MGX in talks to join, according to Reuters. OpenAI wants the money committed before Anthropic’s IPO in November lets investors compare the two companies’ numbers, which feels to me like they’re not very confident in their stated valuation. OpenAI is telling investors its annualized revenue reached $50B after it launched GPT-6 Astra and should hit $70B by year-end, while Anthropic was at about $65B by July. However, OpenAI reports revenue after partners take their cut, and it says Anthropic inflates its numbers by booking the gross value of deals made through cloud providers, which Anthropic says is standard accounting.
In the meantime, as OpenAI waits for more money, it continues to spend what it’s got in the bank. Last week the company acquired Orchestra Bio, a four-year-old San Francisco startup whose AI agents help biopharma companies manage drug development programs, for undisclosed terms. OpenAI’s venture arm also co-led a $45M seed round for Coherence Neuro, which is building a brain implant that uses AI to detect and treat brain tumors. Quick question: How badly does Sam Altman want to be Elon Musk? What’s next for OpenAI… electric cars? OpenAI has absolutely no financial discipline and can’t figure out if it wants to be a search engine, coding tool, advertising platform, business productivity suite, or biomedical company, among other ventures.
Nvidia is in talks to invest several hundred million dollars in Firmus, an Australian GPU cloud provider, according to The Information. Firmus is looking to raise $2B or more from private investors after withdrawing its roughly $5B IPO this month. Nvidia has already backed Firmus in each of its last three rounds, and in June it agreed to provide credit support for Firmus’ roughly $15B purchase of racks holding 170,000-plus Nvidia GPUs, in return for a share of its cloud revenue. Nvidia keeps backing debt-ridden GPU clouds like Firmus, CoreWeave, and Nscale to protect its turf as Amazon, Google and Microsoft begin heavily pushing their own chips.
CRB Group, the parent of fintech partner bank Cross River, agreed to acquire Accrue, a startup that began as a ‘save now, buy later’ service and evolved into a payments and loyalty platform, for an undisclosed price. Accrue sells merchants a branded wallet that ties payments, stored value, rewards, and store-credit refunds together, so shoppers keep spending with the same brand instead of moving their money elsewhere. Cross River already holds the deposits and issues the linked Visa debit card behind Accrue’s product, so the deal gives it the shopper-facing front end it was missing. Cross River plans to offer the wallet to all of its clients, letting any brand on its banking platform build one.
Meta plans to pay a January stock dividend of one non-voting Class C share for every Class A and Class B share, giving it stock to sell for cash, employee pay, or acquisitions without diluting anyone’s voting power. The move comes after its AI buildout pushed free cash flow to negative $2.12B in the third quarter for the first time since 2012. The dividend also lets CEO Mark Zuckerberg sell some of his new shares to pay California’s proposed billionaire tax without losing voting control. A special board committee approved the share sales after Zuckerberg threatened to leave the state over the tax, and in return he promised to keep living in California for eight years, since the committee says it’s critical that he keep running Meta from its headquarters amid the AI race.
FundPark, a Hong Kong startup that lends working capital to online sellers, lined up as much as $250M in debt financing from Bain Capital to fund more loans. Bain committed $100M now, and the remaining $150M depends on conditions FundPark didn’t specify. FundPark lends sellers money for inventory, marketing, fulfillment, and expansion, and decides how much to lend based on a merchant’s live sales data instead of relying only on a traditional credit check. The company said it has now lent more than $10B in total to over 40,000 online shops in Asia and beyond since it was founded in 2016.
Daytona, a startup that provides AI agents with their own dedicated cloud computers, raised $60M in a Series B round led by FirstMark Growth at a $1B valuation, with Shopify CEO Tobi Lütke among its individual investors. Unlike standard cloud servers, Daytona’s environments start in milliseconds, keep their state while paused, can be cloned for parallel work, and disappear once a task is done. They can run Linux, Windows, macOS, or GPU workloads, so agents can do things like test Mac apps or operate CAD software. Daytona says its revenue has grown 25-fold since January and its customers now include Meta, Cohere, Postman, and n8n.
Mynt, the company behind the Philippine mobile wallet GCash, went public on Tuesday after raising about P60.9B ($967M), in the largest IPO in Philippine history. GCash had 41.5M monthly active users in the second quarter, who use the app to pay merchants, cover bills, send money abroad, borrow, save and invest. The company plans to put most of the roughly P10.6B ($168M) it raised for itself into expanding its lending business, with the rest going to product development and general corporate purposes.
Thanks for being a Shopifreak!
If you found this newsletter valuable, please leave a review on Google and share the newsletter with your friends and colleagues to help us grow.
See you next Monday,
PAUL
Paul E. Drecksler
🌐 Shopifreaks.com
🧑💼 Add me on LinkedIn
📧 [email protected]
📱 +1-828-273-3031
⭐ Leave A Review
PS: What does a panda ghost eat? Bam-boo! Happy Halloween!