Host, Applied Intelligence Podcast 🎙️ | Co-founder Boosted Commerce & @ voitechnology | Uncovering applied AI investment edges in eCommerce & agents | From rea
Los Angeles, CA
Joined April 2008
- Tweets5.2K
- Following1.5K
- Followers3.2K
- Likes2.2K
The next moat in consumer AI might be your friends.
@a16z partner Josh Elman on agents that ask your friends' agents for advice
youtube.com/shorts/-LZp8ziBp…
Prompt engineering is essentially a dead art form.
Goal-oriented chat interactions are what produce outcomes now.
The fastest way to get a company using AI might be to make the executives build something first.
Sonos COO Frank Barbieri on an offsite from his Walmart days, where every table of VPs through the C-suite got a printed assignment to write an app:
"We all dove into that exercise with fright, fear and curiosity and built apps right there in that half day."
Meta is raising $97.5 billion in stock. The press release has a section titled "Investing with Financial Discipline."
For scale: this week's Meta raise, SpaceX GPU lease, and the $63 billion Broadcom chip financing for Anthropic add up to about $200 billion. Kenya's GDP is about $151 billion.
The Terminator came out 42 years ago today. Anthropic chose today to release its S-1 warning of "catastrophic or existential risks to society."
Go figure.
🤖 Made with AI
American Express beat on profit and is cutting about 5% of its workforce to "prepare for AI." The CEO also promised Amex is "not going to let assistants run rogue" through its systems. Good to know the bots are on a short leash. Membership has its privileges.
StubHub told Brewers fans whose World Series tickets never showed up: don't worry, you'll get your money back. AI can basically do everything for you nowadays, except figure out how to help StubHub live up to its "guarantee".
PHP runs roughly 70% of websites. This week AI found two critical flaws in it that had been sitting there for over a decade. Defenders got there first and patched them the same day.
That was the near miss. Here's why the next one worries me, from my July episode with @dcam_ai:
“A single person in their house is able to completely disrupt your network by themselves within a couple days what would've taken a nation state months.”
And they won't aim. They'll “spray it at everything.”
"One of the things I learned at Apple was how to package up the features and make those feel whole and complete."
@joshelman explains the product philosophy that your AI startup can benefit from.
That idea you shelved years ago because you couldn’t find an engineer, or because it seemed too hard to build?
That obstacle is gone.
@joshelman is watching people with no engineering background just go build it themselves. No venture capital. No hiring spree.
The only thing left between you and your idea is starting.
Every company is stuck between "go all in on AI" and "wait for it to settle."
@joshelman's answer:
"You can't put your head in the sand, but you also can't assume that everything is now static."
What that means in practice:
- Go deep. Pick a real workflow and move it to the best AI has today.
- Don't put down roots. The workflow you're replacing probably ran for 10 years. Be ready to flip the new one in a much more compressed timeline.
- Don't wait. Start two years from now and you'll be "so far behind everybody who's tried the steps along the way."
The hard part isn't the tools. It's getting your team comfortable changing how they work, and then changing it again.
You can watch the full conversation with Josh here:
youtu.be/mC4YzukA7Y8?si=qgSd…
People love to tout some of the flashy use cases for personal AI agents.
“It booked my hotel room!”
“It cancelled my subscription!”
“It bought this for me and just knew what I wanted!”
@joshelman presents a better framing for how personal AI agents can actually improve our lives: do the busywork, then leave the decision to us.
Are marketplace consumer businesses in trouble? Match Group CEO Spencer Rascoff (@spencerrascoff) explains the challenge and why unique inventory matters.
Rohan Nayak (Pocket FM) posted numbers entertainment will argue about all week.
$500M ARR. $250M added in 12 months. AI-only production after mid-2024. Output: 25,000 hours/year → 2.5 million. Ninety titles over $1M lifetime. Thirteen over $10M.
His frame: produce Naruto-like series for ~$1,000. Netflix spends $17B to find 100 blockbusters a year. What happens when you can make 1M high-quality shows for $1B?
He didn’t win because AI beats a human writer. He still hires writers and producers. He won because he scaled their talent — then made a lot more bets.
Six months ago at Stanford I told a room of writers/producers/directors: AI isn’t here to replace your judgment, relationships, or voice. It’s here to replace the blank page, the busywork, the first draft you hate, the research bottleneck, the fifteen versions you never had time to try.
Same taste. Faster iteration. That’s the whole thing.
Everyone has AI. Few have taste. That’s the edge.
If someone with your taste and relationships ran 10× the experiments you will next quarter — what happens to your career?
Source / context: Rohan Nayak on Pocket FM’s AI-scaled production (TechCrunch, Sep 10, 2026)
techcrunch.com/2026/09/10/in…
