38) Walter (pt. 2)

Behind-the-scenes building Vambrace AI, a company on a mission to figure out its mission. Please pardon the stream-of-consciousness style. Subscribe to follow along or visit the site here:

(typos are to make sure you’re paying attention)

Introductory Remarks

Dear Vambracers —

In last week’s post, What do I think?, I tried to provide a somewhat-structured set of beliefs around AI and its relationship with software, technology, and humanity. I’ve said this before, but I’m a hardcore techno-optimist, and I really do believe in a future where we’re brought to some post-scarcity reality where we can all thrive and grow and pursue self-actualization in a massive positive-sum environment. I’m also willfully naive in my dreaming—and I do acknowledge that we could also get wiped out by robots.

Walter In Action

In today’s post, I wanted to touch on a specific use case that Walter, my AI Chief of Staff (powered by #clawdbot), has executed in the past two weeks that has been extremely valuable. As brief background, I incubated and recently launched a tool that lets Substack writers upload their content and get a conversational archive in return (s/o AskMyStack). I’ll share more about the origin story and technical build when it’s ready, but I am pleased to report that we onboarded our first 2 users over the weekend. So, how did we get those first 2 users?

[Note: the following work was done almost entirely from my phone, texting with Walter.]

1) GTM strategy

I worked with Walter to develop an understanding of the Substack newsletter market—and then Walter also independently did research on what types of newsletters made the most sense to target. Walter discovered that Faith and Music newsletters have the highest paid engagement on Substack, and we decided to orient our outreach efforts around those groups. In total, we identified ~8 categories that we thought might be conducive to the product we were bringing to market (other categories include Parenting, Sports, Literature, Cooking, etc.). Walter built me a comprehensive GTM playbook that provided a market overview, discussed our positioning, and refined our approach.

2) Lead gen

From there, our initial plan was to use some sort of lead generation service to start scraping substack for leads and then use an enrichment service to get emails. But we realized that you can send emails to [publication-name]@substack.com, which many writers have routed to their personal emails. So then we realized that really all we needed for our lead list was the substack URL.

From there, Walter was able to scrape Substack directly and find publications listed on substack’s leaderboards. Walter then realized that most substacks had their own list of recommended substacks on their pages. So if we scraped ~50 publications from the leaderboards, then we could get additional hundreds of publications by scraping recommended newsletters, and then continuing that process until we had about 1,000 publications per core category of interest.

As this process currently stands, we built a lead list of over 6,000 publications across our categories of interest (and some other test categories), with publication name, subscriber amount, category, author name, etc., in a well-organized Airtable base that Walter managed. We’ve used this lead list for our initial test outreach and that’s how we landed our first 2 users, who each responded to cold outbound.

3) Personalized cold outreach

Finally, I crafted the actual outbound email with Walter, and we know that a personalized frontend is a critical component for any outbound campaign. To that end, Walter was / is able to actually look at the latest posts on a substack and create a specific 1-2 sentence intro paragraph that is 100% custom to the person we’re emailing. It also frames up the problem we’re trying to solve with AskMyStack.

The personalizations can honestly be a little silly, or over-the-top, but I think it’s a million times better than generic boilerplate outbound. I also acknowledge that most people would probably think it was written with the assistance of AI, but they wouldn’t be 100% certain.

Examples follow:

It definitely isn’t perfect, but it also isn’t all that bad. And the real mind-boggler is that I orchestrated all of this by texting from my phone—and probably have put in a cumulative ~2-3 hours of my own time. To get ~150+ personalized cold emails out there as part of a broader outbound campaign that was structured by Walter, 2-3 hours of my time is pretty insane.

The word that really resonates with me here is leverage. In the past, this would have felt like an insurmountable task, or something that I would have had to spend ~$50-$200 on to achieve anything of substance. Not to mention hours of time sunk into targeting and scraping and enrichment and so on and so forth.

But now, with $0, and over the course of a few hours (spread out over days), I built a comprehensive GTM playbook, built a high-quality lead list of 6K+ relevant prospective users, and delivered a personalized outbound email campaign for 150 test leads. That’s crazy to me!

And all I have to do moving forward is text Walter: “Hey let’s email 20 more leads for AskMyStack right now” and personalized emails are crafted and sent within minutes. This is a real, tangible use case that has big-time amplified my output and acquisition efforts—and really emboldened me to seriously pursue this initiative in the first place.

Looking Forward

The key to discovering impactful use cases for Walter (and AI more generally) is to struggle with AI. I’ve had so much fun testing and experimenting and have never felt more confident that we’re still so early in the development and dissemination of this technology. I also think we’re still long-term underestimating its importance and impact.

I hope you have a great week! Try something new!

Sincerely,

Luke