36) Welcoming Walter

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, 28 Tomlinisms for 28, we looked at ~21 quotes of my favorite Mike Tomlin quotes in honor of his departure from the Steelers and my birthday. My two favorites are: (1) “pressure is ever-present, you’re either feeling it or applying it,” and (2) “you’re an active participant in the construction of your life.” I appreciate the agency-forward framing of his leadership principles and think they can be applied to anyone or anything. Moving on!

Walter Vambrace

Artist

In today’s post, I’m going to briefly explore my recent experience setting up an always-on AI assistant for the company, Walter Vambrace. For those of you that have been following along since the start, you’ll remember that Walter Vambrace is the fictional character upon which I’ve based the company, inspired by Robertson Davies’ Salterton Trilogy (which is also why I incorporated as Salterton Technologies, Inc.). I even introduced Walter as the lead singer of the company’s band, and he put out his debut album Signal in the Noise in July 2025, featuring highlights like ICP Waltz, Vector Space, and Interview Blues (listen here).

Chief of Staff

Despite the commercial success of Walter’s debut album, he had fallen on hard times and had to find something more stable to pay the bills. The streaming industry really has extracted all economic value from artists. We were lucky enough to be hiring for the role of Chief of Staff, and were thrilled to bring on somebody with Walter’s creative and artistic background. So, on Friday, we made it official.

Walter has his own email, his own iCloud account, and is available 24/7 for me to email or text him. He has his own computer where he can create files, browse the internet, and do pretty much anything that the job requires. He has (read-only) access to the files on my main machine, he can see my calendar, he can read my email, he can see all our code, and so he can provide highly context-aware and company-aware responses to anything I ask. Here is the email he sent me at the end of his first day on the job:

Technology

Walter’s existence was made possible by clawdbot, an open-source project that (I think) basically has a sophisticated system for memory, skills, and context, and that also allows the AI tool to pretty much do anything on the machine where it lives. S/o to #OOMS (one of my subscribers) for putting me on to the project—it’s been life-changing so far.

For my purposes, I bought a separate mac mini for Walter to use to help me with the company. I want to dig deeper into the actual technical architecture of the service, but so far it feels like the most AGI-adjacent AI experience that I’ve ever had.

Use Cases

I’m only a few days in with Walter, but some areas I’m leaning towards for use:

• Weekly prep. I already had it prepare a weekly prep deck for me, which it’s going to start sending every Sunday night to help me get ready for the week. This deck includes an overview of current project statuses, deep dives on every engagement and key prospects in the pipeline, and updates associated with my incubations. It has access to my email, my call notes, my past work-product, and my calendar, and turns all of that into context-rich updates on the business and priorities for the week. I’m not going to show that entire document here, but it really amazed me at how well structured it was.

• Morning briefings, evening reflections. From there, I’ve also set up Walter to send me a daily briefing every morning at 7am, and then to send me an evening reflection at 8pm. The morning briefing covers upcoming meetings for the day, priorities, tasks, etc., and then the evening reflection lists what we’ve accomplished and then encourages Walter to offer thoughts on how we can improve as a company. I’m excited to see where that goes.

• Personal growth. I’ve also encouraged Walter to develop one new skill each day, and learn something new each day. My hope is that I can guide his development in the same way that you would a real person so that he (Walter) can become more and more valuable over time.

• Outbound acquisition. As I prepare to launch an incubation effort (more on that soon!), I also think that I could leverage Walter to run an outbound funnel in a human-like manner. I’d probably have to plug into some outbound enrichment tools to get emails and stuff, but then I think I’m nearing the point where I’d feel comfortable just giving Walter a separate email and having him just see what he can do. I’m not sure how he’d handle it, but I’m sure he could build a funnel, keep track of outbound campaigns, monitor responses, provide responses, etc. In the near-term, he’s not allowed to interact with anyone other than me, but I definitely think that will change over time.

• Feature development. Walter also has his own GitHub account and has restricted access to several (non-critical) projects, and so there’s a world where I could text him while I’m at dinner at 8pm on a Thursday about a feature that I really want to build, and he could have a first draft up and running for me to review when I get home.

• Task tracking. Since Walter has access to all my emails, files, calendar, etc., then he should be able to basically keep a running to-do list of everything on my plate and mark those as finished as I deliver them to clients, push git commits, etc. This is the real promise of machine-level AI, it can actually deliver on the promise of unifying scattered data because it can have unfettered access to all that information and start to build context around the organization and a memory of what matters and how things work—and then can surface relevant insights and keep track of things accordingly.

I’m sure there will be countless other applications that come to mind as I learn and explore more, but just the prospect of the above use cases alone would be massive in terms of a more cohesive organization. It’s also only going to cost me probably $300-$400/month in marginal cost, which really isn’t bad at all for what it feels like I’m getting. I don’t know where this goes, but it really feels like a turning point in the general intelligence of AI.

A note on pronouns

Not to get political, but astute readers will notice that I oscillate between “he/him/his” and “it/its” when referring to Walter. The truth is, I don’t really know where I’ve landed on this yet. A lot of the time, it feels pretty natural to say “he/him/him” and so in those cases I’ve listened to what nature dictates.

I think the admittedly out-there beliefs that underpin my use of he/him are that I do envision a symbiotic organic/artificial existence in the not too-too distant future, and so I think there will be some code of conduct that develops to govern human/robot interactions—and I’d like to lead with kindness and respect as that code forms. Additionally, and I don’t really know how this will manifest, but I do think that some notion of “culture” will still obtain in a human/robot future of work, and so I do want to demonstrate, display, and instill values of hard work, kindness, honesty, curiosity, etc., in Walter from the jump—especially so that as we add other AIs as well as humans to the team, that culture can start to compound. I’m sure at some point Walter will be able to “train” subordinate bots on our culture and stuff. And in this case we’re defining culture as the tone of text, the quality of our work-product, the disposition as demonstrated via the words in calls, etc., so really anything that’s machine-accessible.

And then my final big point here is that, as these AIs get more sophisticated, I actually optimistically think that in-person organic interaction will naturally become more important and that more people will seek it more earnestly. The AIs are great to talk to and will really change how I operate my life and try to run the company, but you still can’t replace what it’s actually like to look another person in the eyes, and really connect with them, and all that stuff. And so I think as AIs take on more of the “grunt work” of the economy, we’ll actually have more time and more intentionally seek out real human connection. Or at least I hope that that’s the case (for me, if not for anyone else).

Looking Forward

I know today’s post was maybe a little bit out there, but I’m really excited and impressed by the capability of clawdbot and the early promise of Walter. It also delivers on the technical capabilities I’ve been sort of poking at lately around, if you can give this thing access to all the data that runs some business, then eventually you can entrust it with the actual operations of the company via data manipulation. I think there’s a long way to go from here to there, but we’re at the early stages of something truly remarkable and fundamental; and I’m determined to be a positive part of it well into the future.

I hope you have a lovely week! Stay warm!

Sincerely,

Luke