Why I stopped pasting decks into ChatGPT
Investors live in their inboxes. ChatGPT lives in a tab. The friction between the two adds up to hours per week. Here's the case for an AI that sits where your work already is.
What we're building, why we're building it, and what changes for you. Mostly use cases. Some engineering. No fluff.
Investors live in their inboxes. ChatGPT lives in a tab. The friction between the two adds up to hours per week. Here's the case for an AI that sits where your work already is.
We just shipped a homepage that calls Marvin your AI Chief of Staff. Fair to ask: what does that actually look like? Here's a real Tuesday morning, from the perspective of an investor who's been running Marvin for three months.
By week three, most recurring briefs recycle the same stories with new dates. Marvin remembers what it has sent you and only surfaces what's genuinely new.
Pick a topic. Pick a cadence. Send Marvin one email. A brief tuned to exactly what you want shows up on schedule. No subscriptions, no algorithm, no editor.
Every AI assistant claims to read the web. Then you send them an x.com URL and they hallucinate something close enough to fool you for a minute. Marvin reads X posts accurately. Here's why that's harder than it sounds.
An AI assistant that invents plausible-looking citations is worse than one that admits it doesn't know. Every URL in a Marvin newsletter is one a real search this morning actually returned.
How to create Marvin newsletters for investors, founders, sales, research, parents, travelers, operators. Six real prompts and what comes back.
Forward Marvin your watchlist. The next morning, a brief shows up with only the material news on your holdings, grouped by company, ready to read in two minutes.
What formats Marvin accepts, what he does with public versus private companies, how to add and remove holdings, and how to keep the brief tuned to what you actually want to read.
Your scheduled newsletter is yours. A CC'd colleague can't reschedule it. A forwarded thread can't accidentally duplicate it. Here's how we made task ownership behave the way you'd expect.
Deep research takes minutes, not seconds. The point of an asynchronous AI assistant is that you don't have to wait around. Here's what that actually looks like in practice.
You get a research report on Monday. On Wednesday you want to dig deeper into one section. Marvin now builds on the prior report instead of starting from scratch, and it's faster, cheaper, and more coherent.
Forward a YouTube URL to Marvin. Get back a structured summary with timestamps and key points, backed by Gemini's native video understanding, not a transcript-and-pray approach.