Hey y'all, let’s open the books.
Back in August I told you two things happened in the first month of setting up a company brain and starting to run an agentic system on top of it:
I cut $425,724 in annual expenses
I increased my company’s shipping velocity by 10x
This week I’m breaking down how I did that (including what got cut), how things have gone since then, and how the system underneath it works.
I’m also running a free workshop on Oct 29th at 1pm EST where I’ll walk through step-by-step how to build your own company brain (the exact system I lay out below).
Forwarded this email? Join 60,000 founders and sign up now.
Can you afford your next hire?
Founders learn the answer to this the hard way. Hiring changes your expenses, burn, and the month your cash runs out.
Parallel connects to your accounting and gives you a live forecast built from your real numbers. Add a hire, see the fully loaded cost, and the impact on your runway & cash. So you can hire (and raise) knowing exactly where you stand.
Founders backed by Y Combinator, Khosla Ventures, and Techstars use it.
Start your 15-day free trial (no credit card)
How AI Saved Me $425K
Why I Built a Brain
After 3.5 years of building, I figured there had to be bloat inside the company.
I’ve done sporadic cuts in the past but, even before Opus 5.5 (which IMO was a big step forward that somehow sort of went under the radar) and Jev launched, I knew I was likely leaving margin on the table by not making Rye Valley fully AI-native.
In the spring I’d had my team set up a team of OpenClaw agents for various tasks, but that was overkill.
Tbh most startups don’t need teams of agents.
They just need a company brain that keeps every piece of context that normally gets forgotten, lost when key people leave, or becomes “tribal knowledge” instead easily accessible by anyone in an instant.
And one agentic system layered on top that can automate existing workflows and create new ones. I use Hermes for this.
So after launching JOLT! a few months ago, I turned my attention to auditing everything we did and building out this system for us.
What Got Cut
Rye Valley isn’t a huge company. We’d had ~$1.12M in operating expenses for the 12 months prior to running the audit.
I determined what could be cut by doing two things:
I thought about what jobs truly make sense for humans to still be doing. These are things I personally wouldn’t want AI to automate today
I made a list of everything we do across the company including how much time it takes, how often it happens, and whether it’s rule-based or requires judgment. Then I had Hermes go through our books to see how much we were spending on all of them
When I ran through that it was clear that almost 40% of everything we pay for could, in theory, be cut by implementing a well-structured agentic company brain system.
That was pretty crazy to see. It broke down like this:
Area | Annualized cuts | Share |
|---|---|---|
Contractors | $250,572 | 59% |
Services | $131,160 | 31% |
Software | $43,992 | 10% |
Obviously, this was a net positive for the business due to the flexibility it gives us but there’s a real human element here too.
It’s never fun to say goodbye to folks on a team, especially when it’s for reasons that have nothing to do with the quality of their work or their dedication to the business.
But smaller teams are a new reality. You simply can (and should) do more with less now.
What I Built Since
Cutting $425K is one thing. What you do (and have to do) next is another.
At a minimum, we’d add some AI costs to cover the various tasks, tools, and human hours that had been cut.
Beyond that, I wanted to use some of the savings on:
Ongoing monthly costs to replace what was cut
One-time builds to replace what was cut
One-time builds on net-new systems
I kind of went a little crazy here, with nearly 6,000 GitHub commits in the last 2.5 months. Compared to the same time period before I started this project, that’s a 17x increase across the company.
(Yes, average commit size has gone down, but lines of code are up 7x so a 10x overall velocity improvement is a fair read.)
And those net-new projects have been high leverage:
A new client portal
Internal tooling to help our team move faster
Even a whole new company, and a ton more
Anyway, those costs break down like this:
Type | Cost |
|---|---|
Recurring | $5,482/mo |
One-time replacement | $11,108 |
One-time net-new | $37,467 |
If you just look at the recurring costs it’s a savings of $29,995 every month, which means it’ll take less than 2 months to make back in savings what I invested in all the one-time build costs.
That’s the type of math that makes any founder think about being even more aggressive.
The System
The system itself is straightforward and has four pieces:
The feeds → The new information about what happens at your company each day. Every meeting transcript, Slack message, email, decision, and more. Fully automated.
The brain → The home of all context across the entire company. It has a log of all decisions made, a graph of all your teammates’ extended networks, and is the destination of every feed so it stays up to date constantly (with cron jobs making sure the right info gets routed to the right place quickly). It’s the chronicle of your company.
The agent → A self-improving system that can make use of the context in the brain both proactively and reactively. This is what I use Hermes for.
The harness → Your company’s specific library of .md files that the agent can use to perform the right tasks, the right way. Hermes builds this itself as it learns, but you can also tell it to build specific skills (for example, I wasn’t happy with the quality of design Hermes was doing for me so I had it spend a few hours crawling Airbnb and Linear’s sites to learn everything it could about good design and now it’s solid) or import ones you find online.
I run mine on a Mac Mini, but it’s actually easier to host yours in the cloud and both options let it run 24/7.
That’s what I’ll be demoing in a free workshop I’m running on October 29th at 1pm EST.
Whether you’re bootstrapped and just trying to trim the bloat or a venture-backed founder whose investors are asking what you’re doing to become AI-native, a robust company brain with a proactive agentic layer on top can help you do more with less.
How We Can Help
Become a member to get full access to our case study library, private founder community, and more.
We can also help your startup in a few other ways:
TikTok + IG
Build your founder personal brand with shortform video and be impossible to forget.
LinkedIn + X
Share your unique POV with the startup world and make investors + candidates chase you.
SEO + GEO
Teach answer engines the truth about your startup.
Advising
I’ll help solve a specific challenge you’re facing with your startup. (New link).
Advertise in my newsletter to get in front of 60,000+ founders.
Our Partners
Founders: treat yourself for once.
A few of our partners are offering $$$ here for taking a quick call with them (that’s it).
It’s companies like Rho, Deel, Brex, and Thera, so it’s quality services that could genuinely help your startup.
Honestly, block off a couple hours, knock them all out, and walk away with some cash to get off the ramen diet for the weekend.
Founder Interviews
Fundraising
Growth
Product & PMF
AI
Founder Mode
