🕶️ Sneakerheads built a $5.1B startup

In partnership with

✦

Howdy, founders. Four days to Halloween, and nothing in tech is scarier than an AI bill.

Amazon just told its engineers that counting AI tokens is a "deeply flawed" way to measure their work. Whop is worth $5.1 billion with about 110 people, because it earns a cut of every sale its customers make, not because it hired fast. And Infisical raised $40 million it says it didn't need. Each story is about counting the wrong thing: tokens instead of launches, headcount instead of revenue per person, cash raised instead of cash earned. So which number is your team chasing, and does it actually pay the bills?

In today's newsletter, we'll get into:

How 6% of every sale built a $5.1B company
Why Amazon shut down its AI leaderboard
A $40M round nobody needed
Discord's servers are heading to Google

✦

The founder’s dashboard / Your quick roadmap

Rushin' Roulette
Find what’s slowing your GTM.

✦

FOUNDER BOARD

🗄️ Three humans run SaaStr's 20+ AI agents, and two wiped critical data. Now SaaStr names the backup copy before granting write access and starts new vendors read-only.
🧮 Five data steps at 80% accuracy leave your AI agent with 33%. Nooks fixed its account data and found 700 accounts it didn't know it had.
💳 Usage limits drove 51% of Evernote's 2025 paid conversions. If you raise prices like Evernote did, gate by usage, grandfather loyal users and warn them early.
🏷️ The FTC and 20 states sued Dollar General after 29,000+ complaints that registers charged more than shelf tags. Check that site, ad and checkout prices match, renewals included.
🥗 Sweetgreen will pay its best restaurant managers up to $200K, mixing sales bonuses and equity. Tie key-manager pay to unit results, and pay them for training their replacements.
🛡️ Arctic Wolf tracked 400+ attacks via fake "IT support" Teams chats. Fix: Teams admin center > Users > External access > Allow only specific external domains.

Cap Table Template for Startups

Slidebean’s Cap Table Template gives founders a clear and practical way to organize and visualize ownership from day one. Instead of juggling spreadsheets or guessing equity splits, this tool helps startups model current and future share distribution, understand dilution scenarios, and plan fundraising rounds with confidence.

Designed for simplicity and accuracy, it lets founders focus on building their business while maintaining clean, investor-ready capitalization data.

Check it out

✦

RUSHIN' ROULETTE

Five bullets of updates

💰 Infisical, which makes software for storing passwords and API keys, raised $40M at a $600M valuation in a round led by Elad Gil, while already cash-flow positive. It says it didn't need the money and has more in the bank than it ever raised, which is the strongest seat in any pitch meeting: raise when you don't have to, and investors start competing for you.
🏭 Re:Build, the Amazon-alumni manufacturing company that spun out EV startup Slate Auto, raised $2.1B in a round led by General Catalyst and JPMorganChase. Its companies already employ 2,100+ people across 2.4M square feet of US factory space, so hardware founders who can't afford their own plant now have a place to build in America instead of Shenzhen.
🔓 Cathedral, a $1.4B security startup run by former DOGE staffers, says it got into Anthropic's internal systems in under 12 hours through third-party business software, after days of failing to break in directly. Anthropic, which invites outside testers, confirmed limited access, and the weak spot was ordinary vendor apps, so audit the tools that hold your salaries and Slack.
🔎 Discord will give its 63,000+ Discoverable Servers, public communities with at least 1,000 members, their own web pages on Nov. 16, and their messages can start showing up in web search in mid-December. It's on by default, which means free search traffic if your community answers customer questions, or an opt-out under Server Settings > Web Presence if it's private.
⚖️ A New Mexico judge ordered Meta to pay $25B after a jury found it misled users about protecting their data, while most states settled similar claims for $459M. Meta plans to appeal, but the gap shows how fast a "we protect your data" line can become evidence against you, so make your privacy claims true as written.

Would you raise money you don't actually need?

Infisical just took $40M at a $600M valuation while already profitable — with more in the bank than it has ever raised. "Only raise when you need it" is the rule that guarantees you'll be raising when you're desperate.

✦

SPONSOR

Find what’s slowing your GTM.

Most GTM problems aren’t one big failure. They’re one or two weak spots quietly dragging down the rest.

The free GTM Audit scores your go-to-market across four dimensions, then sequences a 90-day roadmap for your stage.

Drop your URL. Get your score in about 10 minutes.

Get My GTM Score

✦

STARTUP NEWS

Whop is worth $5.1 billion with about 110 employees

Photo by Money Knack on Unsplash

What does a $5.1 billion company look like? In Whop's case, about 110 people. The Palo Alto startup just raised a $200 million Series D at a $5.1 billion valuation, co-led by Sequoia and Shopify Ventures, per Bloomberg. In February it was worth $1.6 billion, so it more than tripled in eight months.

Whop is a platform for people who run small online businesses, from startup product leads to 16-year-olds running their first company. They use it to set up shop, find customers and get paid. About 25,000 of these "operators" now sell $550 million a month through it.

Six cents on the dollar: Whop doesn't sell seats. It keeps a blended fee of about 6% of each sale, CEO Steven Schwartz told Bloomberg. Here's how that adds up:

$550 million a month is about $6.6 billion a year in sales.
6% of $6.6 billion is roughly $400 million.
That lines up with Whop's run-rate revenue (a year's revenue at its current pace): past $400 million, up from $60 million at the start of 2025.

Per-head count: $400 million across 110 people works out to more than $3.6 million per employee. That's possible because Whop's revenue grows when its customers sell more, not when Whop signs a new contract. If an operator doubles sales, Whop's cut doubles too, with no sales call. (The founders met as teens over software that tracked sneaker drops, so they've understood resale margins for a while.)

And Whop is stretching the model. Its newest tool, which suggests a business's best next move, is paid for with an extra cut of the revenue it helps bring in.

Skin in your customer's game: if your product sits where money changes hands, a small take rate can outgrow per-seat pricing, because you get paid when customers grow, not when they hire.

Looking ahead… Whop plans offices in Europe and Colombia. The test is whether 6% holds once its biggest operators get big enough to negotiate.

Read Full Story

✦

STARTUP TV

NEVER outsource your Minimum Viable Product

When launching a company, it's crucial to assemble the right team. Developers play a pivotal role in this process. While some may opt for outsourcing development work, it often falls short.

✦

BIG TECH NEWS

Amazon tells engineers that using less AI can be a good sign

Photo by Brett Jordan on Unsplash

Give engineers a scoreboard, and they'll find a way to top it. Just ask Amazon. A memo posted Tuesday tells its software engineers that token counts and AI adoption rates are "deeply flawed" measures of their work, and that skipping AI where it doesn't help is "a finding, not a failing."

Tokens are the chunks of text AI models read and write, and AI companies bill by the token. "Tokenmaxxing" means burning through as many as possible to look productive.

Leaderboard, meet loophole: in May, Amazon shut down an internal leaderboard after it found employees tokenmaxxing to climb the ranks. CEO Andy Jassy has said AI could shrink Amazon's corporate workforce in the next few years, and nobody wants to rank last on an AI usage chart after hearing that. Reward people for using a tool, and they'll use the tool, whether or not anything ships.

The memo, first circulated internally in September, resets the rules:

Token use and code-generation stats should be weighed "with care and judgment" next to other data.
Amazon wants to "measure, and be measured by, long-term outcomes."
An engineer who can explain where AI doesn't help "is demonstrating the judgement we are asking for."

Shipped, not spent: AWS vice president Marc Brooker told CNBC that Amazon looks at "feature velocity" instead: how many launches it does and how many changes succeed in production. It hasn't capped token use. It just stopped cheering for it.

Count launches, not tokens: if you've told your team to "use more AI," you may have built Amazon's leaderboard without the leaderboard. Track what ships and what breaks, and read your AI bill next to it. A rising bill with flat launches is tokenmaxxing by another name. If you sell usage-priced AI tools, customers will start asking what those tokens bought.

Looking ahead… Some companies already cap token use or shift tasks to cheaper open-source models, per CNBC. Expect "AI spend per shipped feature" in your next board deck.