STUMP BEZOS
Adobe predicts online US holiday shopping to reach a record $275.1B this year. Cyber Week (Nov 26-30) is projected at $47.5B. How much of that is on Cyber Monday alone?
[ Answer at bottom of email ]
💰 YOUR NEW SELLER AI TEAM — EXPLAINED
Robert Prime from AI Intelligence breaks down the four “agents” everyone is talking about, what each one actually does, and which are worth paying for
Everybody is calling everything an “agent” right now. Claude Code, Hermes, Grok Bot and Jev all get the label, and they do four completely different jobs. The easiest way to think about them: they’re staff. Each has a role, a cost, and one thing you should never ask it to do.
Meet the team
The Architect: Claude Code (Anthropic)
You give it a job in plain English: “build a script that checks yesterday’s sales and posts a summary to Slack.”
It reads your files, writes the code, runs it, fixes what breaks, checks the result.
It builds your automations and checks the other agents’ work.
The catch: it works in sessions. On its own it’s not a 24/7 employee.
The Night Shift: Hermes (Nous Research)
Free, open source, installed on your own machine.
Runs scheduled jobs overnight, works through task queues, pings you on Telegram, WhatsApp or Slack when it needs you.
You only pay for the AI model it calls.
The catch: it’s only “always on” if the machine is. On a laptop, it sleeps when the laptop does.
The Remote Assistants: Grok Bot (SpaceXAI, formerly xAI)
Launched in August. You create named assistants, each on its own screen of a shared cloud computer with a browser, files and a terminal.
Hand one a task from your phone, close the app, it keeps working.
Big strength: websites with no API. The back-office screens nobody ever automated.
Comes with paid Cursor plans (from $20/mo) or SuperGrok.
The catch: all your Bots share one computer, so every Bot can use every login on it. Think hard before client or seller accounts go there.
The Triage Clerk: Jev (TypeSafe AI)
The odd one out. It cannot write a sentence.
It answers three kinds of question only: pick one of these options, score this on my scale, or true/false.
Answers in about a tenth of a second with an honest confidence percentage. Costs $0.042 per million input tokens, output free.
Built for sorting thousands of reviews, emails or leads before a pricier model (or a human) looks at the ones that matter.
Launched mid-September, early access only. Waitlist.
How they work together
The pattern: Claude Code builds it, the others run it.
Hermes + Claude Code have an official connection through MCP. Claude Code can read what Hermes did overnight and message you through it. Hermes can hand coding jobs back.
Grok Bot + Claude Code have no official link yet. Hand off through Slack or GitHub: one posts the task, a Grok Bot routine picks it up. A Bot can also install Claude Code on its own cloud machine.
Jev + everything: you don’t “talk” to Jev. Claude Code writes the code that calls it, and your automations use its answers to decide what happens next.
Two rules keep this from becoming chaos:
One owner per job. Two agents on the same task means double the cost and conflicting results.
Hand-offs happen somewhere you can see them (a Slack post, a GitHub issue) so there’s always a trail.
What this looks like for a seller
None of this replaces your team. It turns “a person repeating a process” into “a person approving the results.” Jobs that work today:
Overnight monitoring. Prices, stock and listings across every SKU, summary in Slack before 7am.
Sorting at volume. Score every review, support email or inbound lead by urgency. Only the important ones reach a human.
Drafting, not sending. Replies, reports and proposals get prepared. A human approves every one.
The admin nobody automated. Portals with no API, clicked through by an assistant.
House rules for any AI team
Logins stay human. No agent types passwords, 2FA codes or API keys.
Agents draft. You send. Nothing goes to a customer, a client or a bank without a human yes.
One owner per job.
A spend cap on everything. Every one of these can run up a bill while you sleep.
Where to start this week
Pick one repetitive job someone on your team does every week. Have Claude Code build it. Run it on Hermes for two weeks with a human approving every output. Only when that works do you add the next job or the next tool.
Start with one job. Prove it. Then add the next.
🛠️ BDSN SOFTWARE TOOL of the DAY 🛠️
Imagine your next shipment moving forward without taking over your working day. Fewer interruptions. Less chasing. More time to launch products, grow sales and build your brand.
SnipeRate is your AI shipping employee, saving you an average of 4.5 hours and handling 33 emails per shipment. You get time back, your team gets breathing room, and freight admin stops demanding so much of your attention.
Get more freight quotes competing for your business, without adding to your workload. Describe your shipment in one sentence and SnipeRate sends quote requests to your trusted forwarders, chases replies, compares quotes and negotiates on your behalf. You can also request quotes from freight forwarders beyond your usual contacts, opening up more opportunities to find a better rate without extra work.
Experience it on your next shipment. The first 100 brands get 125 free credits plus free guided setup. We’ll help you set up your account, add your forwarders and get your first shipment underway, so you can start handing over the work. No subscription. Sign up at sniperate.com and enter BDSN125 (in settings), then email [email protected] to arrange your free guided setup.
👁️ AMAZON LETS SHOPPERS SEE THEIR PROFILE
Amazon rolled out About You, a page where every shopper can see the AI-generated picture Amazon has built of them, then edit or delete it. Dietary habits, hobbies, the cat’s name, the sourdough starter’s feeding schedule. All of it, in one place.
It’s been live since May in the USA, but screenshots went viral this week and the internet is having a field day. More on that below. For sellers, this is the clearest look yet at what Amazon’s AI actually thinks it knows about the people typing in your keywords.
What About You is
A single dashboard of Amazon’s summaries, assumptions and characterizations of a shopper’s habits, interests and lifestyle.
Shoppers can confirm details, correct them, or remove anything they don’t want used for personalization.
Replaces the old scattered controls (recommendation preferences, browsing history edits) with one hub.
Also works by voice: ask Alexa for Shopping “What do you know about me?”
What feeds the profile
Purchase history
Searches
Items saved to Lists
Product reviews the shopper has written
Conversations with Alexa for Shopping
Amazon’s own examples: tell Alexa about your cat Benson’s favorite flavors and it recommends treats in those flavors. Add seeds to a List and it suggests soil and companion plants. Buy non-dairy milk consistently and your grocery recommendations shift to other non-dairy products.
How to find it
Mobile app: Me tab > tap your name > About You
Desktop: hover over your name above Account & Lists > click your name (or Who’s shopping? and pick a profile) > About You
Direct link: amazon.com/slc/hub
Appears to be US-only for now.
It’s not perfect. Sometimes it’s hilarious.
The Amazon AI guesses posted on a thread from BDSS Dream 100 member Jon Derkits and elsewhere range from spot-on to completely unhinged. Real entries people are posting:
“Owns a miniature blimp with an envelope, not high pressure”
“Has flat buttocks”
“Is too short to reach into a 40-inch deep box”
“Is not a big SpongeBob fan”
“Cannot keep real plants alive”
“Is interested in world domination”
“Likes cheeseburgers”
“Likely owns an iPhone”
One finance exec said his profile has him pegged with three kids (he has two) and a perfectly tidy home (“purely aspirational”). Another shopper: “It knows more about my kids than it does about me. But it knows my starting Wordle word.”
The sharpest comment in the thread: what gets you is how confident the wrong guesses sound. A clean record of what someone bought can still lead a model to the wrong conclusion. One gift purchase and you’re a blimp owner forever.
Why sellers should care
Search is personal now. Amazon says nearly every shopping experience is personalized. Two shoppers typing the same query are not seeing the same page. Your “rank” is increasingly a range, not a number.
Lists and reviews are profile inputs. A shopper saving your product to a List or reviewing it isn’t just social proof. It’s teaching Amazon who your buyer is.
Attributes drive cross-category recommendations. Amazon is connecting gardening tools to soil, cat treats to flavor preferences, non-dairy milk to the rest of the cart. If your listing doesn’t speak the same attribute language (diet, lifestyle, use case, what it pairs with), the AI has less to match on.
Wrong guesses are your problem too. If Amazon thinks a one-time gift buyer is a lifelong hobbyist, your product gets recommended to someone who will never buy again. Expect that noise to shrink as shoppers prune their profiles. The traffic that survives is higher intent.
This is just the beginning. Amazon is showing shoppers the profile because it’s about to lean on it harder, not softer.
Do this today
Open your own About You. Read how Amazon describes you. Note the categories of attributes it tracks (and laugh at the misses).
Ask: does my listing contain the words Amazon would need to connect my product to a profile like that? Flavor, diet, hobby, routine, who it’s for, what it pairs with.
Build Lists and reviews into your post-purchase flow. They’re no longer just vanity metrics.
Restock Planner: Your inventory sweet spot
Reach page 1 on Amazon simply by sending free products to Micro-Influencers
Use the platform Stack Influence to automate Micro-Influencer product seeding collaborations at scale (get thousands of collabs per month) and increase your Amazon ranking, generate UGC, and boost up your recurring revenue like never before.
Top Amazon brands like Magic Spoon, Unilever, and MaryRuth Organics have been able to get to #1 page positioning on Amazon and increase their monthly revenue as high as 13X in as little as 2 months.
Pay influencers only with products (stop negotiating fees)
Increase external traffic Amazon sales (get to top page rankings)
Get full rights image/video UGC (build your brand with authentic content)
100% automated management (don’t lift a finger to get influencer collabs at scale)
Don't believe it? Check out the results from the Blueland Micro Influencer campaign which generated a 13X ROI scaling up influencers on Amazon.
After successfully raising investment on Shark Tank, Blueland turned to Stack Influence to boost their Amazon sales and become a top selling listing using Micro Influencer marketing.
Increase your Amazon listings ranking for targeted keywords and multiply your organic recurring revenue in 2026!
Get 10% OFF by signing up this month
📺 YOU GOTTA SEE THIS → The SPILLOVER $ EFFECT
Nine of the 11 biggest telehealth brands (think the GLP-1, hair loss, skin and sleep players) don’t sell a single product on Amazon, even though thousands of people search their brand names there every month.
On this PPC Den episode, Sean from Spillover Commerce shows the receipts: search “Weight Watchers” and the shelf is filled by Olly gummies and Alli diet pills while WW is nowhere. “GLP-1 vitamin support” went from zero to 30,000 searches a month in under three years.
Protein snacks 4X’d. Fiber gummies jumped from 30,000 to over 200,000. Nature Made quietly moved 10,000+ units last month of a “GLP-1 companion” that’s basically magnesium, B12 and D3. That’s a $200 billion category growing 25% a year, and the brands creating the demand are letting everyone else collect it.
The second half is the playbook you can actually use, whether you’re in health or not: how to spot breakout keywords before the crowd, why the best sellers cut the bottom 5% of their catalog every quarter, how to turn a product insert into a Meta retargeting audience without breaking Amazon’s rules, and the one Amazon DSP use case Sean (a DSP skeptic) says actually makes sense. Watch it here
🌎 INTERESTING STATS
🔥 MORE HOT PICKS 🔥
🥃 PARTING SHOT
“Failure is not the outcome-failure is not trying. Don’t be afraid to fail.”
✌🏼 Here’s to your success …
The answer to today’s STUMP BEZOS is
Cyber Monday sales in the US are expected to hit $15.1B
