New to Venture

There are a handful of genuinely interesting ideas - the 'on accept vs. on demand' AI framing, Reply as a Trojan horse for OS real estate, and the mail→email→text advertising cycle analogy - but they are heavily diluted by dropout origin story, hacker house anecdotes, and platitudes like 'there's no playbook' and 'surround yourself with great people.' Insight rate per minute is low.

“AI is temporarily in this query response loop. Um, it's temporarily on demand, but really it's going to be on accept.”

“reply in, in my mind, has sort of always been this Trojan horse to buy the real estate on your computer, on the os”

The 'Invisible OS' memo concept and 'on accept' framing for ambient AI are genuinely fresh articulations, and the Trojan horse distribution logic via messaging is non-obvious. However, the episode is also laden with recycled startup tropes - dropout heroism, hacker houses, warm intros beating cold outreach - that dilute the original thinking.

“I wrote this memo, um, a while back, which I called the Invisible os.”

“it will be ambient and omnipresent, it will be woven into everyday apps and it will be almost indistinguishable from the os”

Molly is a genuine practitioner - actively building, shipping, and iterating - with a credible pre-seed led by OpenAI. However, the company is extremely early stage (sub-PMF, small team, consumer/prosumer product), and her operational experience is necessarily thin; she openly admits having no fundraising knowledge and has not done this at scale.

“I had literally no idea what I was doing. I, you know, funny story, I actually thought instead of it being valuation, I thought it was evaluation.”

“A large majority, I'd say maybe 60%, uh, have found us through either ChatGPT search or they've been trying to find Some of these solutions”

There are scattered concrete details - OpenAI-led pre-seed, Samsung watches as early hardware, 60% of users via ChatGPT search or Twitter, a specific user with a Notion spreadsheet doing manual copy-paste texts, named tools like Ramp and Brex in the OS vision example - but the episode lacks any revenue figures, user counts, retention data, or funding amount, leaving the most commercially useful numbers absent.

“A large majority, I'd say maybe 60%, uh, have found us through either ChatGPT search”

“I had one guy last week literally tell me he has a notion spreadsheet of um, all the text he wants to draft out per day and once again copies and paste”

The host lands a few legitimately sharp follow-ups - catching the iOS privacy contradiction and probing distribution theory - but defaults repeatedly to affirming filler ('Oh my gosh, what an awesome story') and never pushes on metrics, competitive differentiation, or the viability of the vision. The final three questions are purely soft and ceremonial.

“Don't you run up into the same problem with the iOS privacy and access to data?”

“Oh my gosh, what an awesome story. And it feels like there was momentum right out the gate”