👋 Hey, I’m Jaryd. I share one stealable idea with you a week on product, growth, AI, marketing, or strategy. My goal is to help with the hard part—growing the thing you’re trying to build. For more: Explore the top ideas | Watch me yap

Good morning guys, happy Friday, and quick personal update

So, I’ve been at this video thing for a month and change now. I’ve had three questions from you guys, so let me give a quick answer:

So, before we start, I have one favor to ask you. If you enjoy what I say in this newsletter and want to get to know me better…pop me a follow—lots more to come there and I’d really appreciate it.

Ok, self-selling moment over 😮‍💨. Gotta do it.
Let’s get cracking—the last bit explain’s why your crypto is tanking today.

What’s going on: YC dropped Winter 2027 Requests for Startups. They open with this line → “AI’s biggest bottlenecks are moving beyond the models.” The 11 asks are space startups for a world of 10,000 Starship launches a year, robots made in America (fun fact, China made 94% of rare-earth magnets in 2024), data centre supply chain (transformers have a 3-year lead time), lenders for the physical AI buildout, Plaid for government benefits, AI for City Hall (19,000 US cities, $16B a year in extra tax revenue if permits took 10% as long), air-gapped AI, vertical foundation models, companies owning their own models, and “the $15-an-hour expert.”

Why it matters: The consumer apps are gone. Their last batch had three of them. Basically YC are adding to the message that intelligence is now cheap, but steel, permits, magnets, capital and trust are not. There’s one request in there that isn’t atoms—they want an LLM that writes well, because models have gotten worse at writing as they’ve gotten better at code and the labs won’t fix it.

The idea behind this: Look at where YC’s list points us. They’re asking for stuff where AI runs into a wall that it can’t code its way through. If your idea gets easier when a new model drops, so does everyone else’s. If it gets easier because you own the data, the permit, the relationship or the hardware, that’s the moat they’re funding. Funnily before this drop, I spoke about manual effort moats.

What’s going on: Thrive Holdings is buying ADT. $15.3 billion enterprise value, 12,000 employees, 6 million customers, and has been around since 1874. Thrive Holdings already owns 75+ accounting and IT firms it’s rewiring with AI. Now they are swinging at alarm systems.

Why it matters: The bet is buy a boring business with sticky customers and a big labour bill, put some engineers inside it and point AI very hard at the margin. General Catalyst is running the same play. ADT is full of monitoring centers, dispatch, billing and sales calls—all good work for agents. The margin improvement game.

The idea behind this: Every category has their own ADT. The one with the trusted name, good recurring revenue, bloated ops, no real engineering department. The AI opportunity is pumping it into an old company. The flip side though is that the same AI that makes a roll-up like this work is also sold to every independent firm by a dozen SaaS tools now, so the edge might get blunter. Anyway, I just think it’s interesting that the smartest money in AI is buying alarm systems.

What’s going on: Anthropic joined the the Muse/Dots/Instinct party with Duet last week. Duet’s a standalone personal assistant app that lets Claude use your connected tools. Same idea as Muse. I haven’t tried it yet but I like Dan Shipper from Every and he called it “the first personal agent I’d call a pro tool.” Lenny’s said “Claude Duet is beautiful. I still forgot to use it.”

Why it matters: All these personal assistant agents are built to be proactive but unobtrusive. We don’t want them in the way, their background-ness is what’s nice. But two things are in conflict…a product whose whole value is not bothering you has no habit loop, and “it works while you’re away” means there’s no moment where you see it working. A proactive agent that never nudges is a feature, but it’s also and equally a growth problem.

The idea behind this: If your product does it’s job in the background, you still have to manufacture a visible moment of value every single day. Duet’s morning brief is the hook whether they call it one or not. Hooks matter, even for trillion dollar companies.

What’s going on: Reddit beat earnings. Revenue +53% to $895M, yet their stock fell 18%—why? One reason according to Steve Huffman that everyone with a website has been feeling → Google’s AI Mode “does not drive traffic, not for us and not to anybody.” Search referrals for Reddit dropped by 1 million+ per week this month which took daily US users below their IPO mark, for the first time since their IPO in 2024. Reddit now plans to test a login wall on most content. Maybe 20 years of typing “reddit” at the end of a Google search is over.

Why it matters: Google crawls Reddit hard. It just stopped sending anyone back to it. Which is the ugly and sad version of platform risk: you’re still the supplier, you just stopped getting paid in traffic. Reddit’s answer is to force the relationship. Logged in, in the app, theirs.

The idea behind this: Go look at your acquisition mix…how much is rented? Then ask what Reddit did. If that source went to zero next month, what do you actually own? Reddit’s fix is a retention answer to a distribution problem.

What’s going on: Alphabet lost $482 billion on Thursday. Oof. That’s even hard to type. It’s their worst day since it IPO’d in 2004, and the second-biggest one-day loss for any US company ever. Amazon closed down 4% and then fell another ~7% after hours. But it’s strange—Google Cloud grew 103% and AWS grew 39%, its fastest in five years. Same day (yesterday), US GDP printed 3.4%, with AI equipment spending up 20.9%. That reads as the economy is being carried by the exact spending investors just punished.

Why it matters: Microsoft spent the same money, kept $18.6B of free cash flow, and ended the day flat. Alphabet (-$9.1B) and Amazon (-$42B trailing) went negative and got crushed. But Microsoft did one more thing. It reorganised its reporting into two buckets, “Agents and Infra” and “Devices and Consumer,” so the shrinking bit (Windows OEM down 22%) sits under one label while the cloud gets the headline. TLDR: how a company reports is how it wants to be judged.

The idea behind this: First, for companies burning cash to grow, the question is no longer just how fast revenue moves but when cash flow turns. Second, how companies organize their dashboard is the story. If people keep reading the wrong headline out of your numbers, you might not need better numbers, you might need to rename the rows. Yes wise one, that’s also how you hide a problem 😂 . Important to know which one you’re doing. But maybe Alphabet should split it’s search business from its cloud business.

That’s it for this week guys, have a fun Halloween!
—Jaryd