Introducing KnowCore

This is The Startup Breakdown, the newsletter where we learn, laugh, and love startups. By joining this growing community of hundreds of future startup aficionados (think i spelled that right?), you're getting a beachside view of the ocean that is the startup and VC scene. This ain’t your grandpa’s newsletter, so prepare yourself for an inbox full of 4/20 jokes and Succession references.

Howdy, folks.

And happy 2025, happy Thursday, happy almost March, etc that I missed since the last edition of The Startup Breakdown from July 2024.

I decided to stop writing my newsletter to go all-in on PostOnce. Things were going well, scaling to nearly 100 users, many of whom were paying $14.99/mo for a subscription, but eventually, things started to break down. Churn was high, technical debt was catching up to us, and user feedback dwindled to the point where the only responses I would get were “can you cancel my subscription and refund me?”

Simply put, I wasn’t solving a problem.

So I made the decision to stop development on it last month, reached out to one of the smartest people I know about building something together, and we got to work on KnowCore.

I apologize for the radio silence. And I’m sorry to disappoint (okay at least pretend for me here), but this isn’t my return to regularly writing this newsletter. However, I missed you guys and felt you deserved to know where I’ve been and what I’ve been doing.

With that, below is the first ever KnowCore investor update. It outlines our traction so far, our product, and most importantly, our vision for the future that we hope to build.

Enjoy, and as always, never hesitate to reach out :)

3, 2, 1, Action

February was a big month for us. But I hope it’s the smallest month we’ll ever have again.

Recap:

• Incorporated

• Signed our first pilot (thanks Wefunder (YC W13) <3)

• Agreed to a partnership with our first agent (thanks Luthor (YC F24) <3)

• Got suspended on LinkedIn for sending too many DMs

• Launched our website

• Added integrations for Slack, Notion, HubSpot, and Stack Overflow

• Built v1 of our API and SDK

But it was also a pretty big month for our biggest competitor, Glean.

First, hats off. $100M ARR in 3 years is astonishing. To the point that I question whether an intern accidentally misplaced a decimal or something.

But Glean has truly become the standard in enterprise search.

And now, they want to become the standard of enterprise agents, too.

99% of their demo event was them highlighting the hundreds of internal agents that their customers had built for tasks like meeting preparation, daily news recaps, customer research, and more.

They flaunted massive customers like Zillow and their partnership with Databricks.

And despite all of the apparent momentum, as well as their decision to pivot directly into our focus, I couldn’t help but smile watching the video, thinking they were idiots for choosing to focus on internal agents and cast such a broad net.

I admit that I’ve never used Glean. However, I also haven’t met anyone who has and had good things to say about the product.

Everyone seems to point to a poor UX, and even poorer accuracy, as reasons for not using the licenses their companies purchased.

Don’t get me wrong, if your Notion document has a clear “The sky is blue,” your Glean assistant can find it and tell you the color of the sky if you asked it.

However, if you asked it whether a certain company was eligible for a discounted rate, and an intern from 3 years ago said that they were giving discounts to everyone they spoke to (even though the discount changed from 20% to 15% since then), and your official documents outlined that companies must have no fewer than 100 employees and spend no less than $100,000 every year with your company, the odds that you get the correct answer are probably 1/2.

Context is at the root of agent performance. Shitty context → shitty output.

This understanding is baked into our pipeline architecture. Everything that we’ve built was carefully constructed to reflect that real data is more complex than x = y. There’s nuance, change over time, implicit relationships, and so much more confounding AI’s ability to simply read and regurgitate information.

KnowCore’s knowledge pipeline preserves this nuance. It maps relationships across time, across people, and across platforms, allowing for a richer “context” than any existing AI systems on the market today.

And not only does this allow for better performance for agents relying on this data, but we’ve constructed it in a more sustainable, self-reinforcing way where over time, it costs less for third-party agents than building these integrations and storage systems out themselves would.

As strongly as I feel about Glean’s performance shortcomings, however, these opinions pale in comparison to how vehemently I disagree with their hypothesis that the future of enterprise AI is internal agents.

What percent of 9-5 employees, even at big tech companies, do you think are going to put a pause on their daily responsibilities to try to automate them via agents?

My bet would be very few, even if they can understand the long-term benefit of doing so, and even if they take a look around the room and acknowledge that none of their coworkers are going to do it for them, either.

Plus, even if they were directed to do so from their managers, how many would take the time to build the first version (which can conveniently be built using natural language explanation), experiment, and improve this first version’s shortcomings?

Again, I’d bet my nonexistent money on no.

The future of enterprise AI is third-party agents.

There are going to be hundreds of vertical copilots becoming absolutely fucking brilliant at very specific, defined tasks within a job function because there are going to be hundreds of founders spending 80 hours per week for 2 years building them to be so. And these infinitely-iterated upon solutions are going to laugh at the half-assed “agents” that some PM with weekend hiking plans at Zillow built to save them from doing a Google search.

KnowCore is built for these agents.

We’re building the thousands of integrations that enterprises use to operate and putting a bow on them for project management agents. We’re carefully ironing out permissions protocols, and handling these secured connections to copilots for filing corporate taxes. And most importantly, we’re building the most intelligent context layer in the world, providing the comprehensive overview necessary to truly unlock the potential of enterprise AI.

Looking Forward

So… this sounds great. But how the hell do you do it?

Internally, we have debated this pretty fiercely.

Initially, we were all in on:

1. Build an AI enterprise knowledge base for BD/Sales teams specifically in compliance-heavy industries

2. Use this as our wedge to develop our tech and prep for scaling

3. Start selling API access to agentic startups looking for someone to handle the “plumbing” for them and helping them to sell to the customers we were already working with

However, Glean’s announcement shortens our leash.

Do we continue to go enterprise, selling an “integration software” that allows them to seamlessly and securely integrate new agents into their ecosystem?

• This approach would mean bigger contracts and would be attractive as a single integration, where they only had to worry about and trust one company, KnowCore, to be able to access their internal data. It would pretty much force prospective agents to adopt our API if they wanted to get their product into the hands of our customers.

• However, the ecosystem is too young. How many agents can you actually name that are in the market? It would have been like trying to sell the idea of gas to someone when everyone was still riding around via carriage. Plus, it’s really hard to get a big company, even a startup, to agree to pay us when we have no skins on the wall.

So do we take the opposite approach, selling based on usage through the agents themselves?

• It would be amazing to practically outsource sales to each individual company, allowing us to rapidly squeeze our way into companies across the tech landscape. Then, once we have enough credibility, we could transition to enterprise sales and help usher in commonplace enterprise agents.

• But what incentive does any agentic startup founder with $3 million in seed funding from YC and a16z have to work with us? If they are pre-product and haven’t built any of their own integrations yet, then sure, we could help speed them to market. But how many agents are you hearing about pre-product? We would have to go around to every CS building in the Bay Area asking students if they were thinking about building agents or something, and then try to get them to build them and just use our API for their integrations. Not likely, so we crossed that out. And for the agents that do have a product, they’ve already built the integrations they need themselves… Sure, we could show them demo after demo about how our context (and hence their performance) is better, or how difficult it is to add and maintain new integrations and security measures, or how we’ll be cheaper over time, but when these founders have 40 hours of demos with prospective customers scheduled every single week, their integrations are the last things on their mind.

So… what?

We’re whoring ourselves out. We’re getting on our hands and knees, and we’re crawling to every single agentic startup founder with a simple ask: will you let me sell your product for you… for free?

This might seem simple, but given the number of college students, “agencies,” and YC-recommendation seekers founders get (I know, because I’ve been all three…), this seems like the scam of all scams.

However, what if we can manage to convince 2-3 to let us be their unofficial, affiliate BDRs?

• For the founders, we expand their sales arm, potentially getting them customers for literally no time or money. The only conditions are that we get to build their integrations and the agent has to use our API for that customer, and we take a small percentage of the revenue from that user to cover costs. Plus, they get the performance/cost/plumbing benefits mentioned above 😉 

• For the enterprises, imagine we can come to companies and say “hey, I’m working with these three agents that can handle payroll, customer discovery, and document queries for you better than any other tools on the market, and you only have to integrate and authorize a single entity to access them” and they’ve completely revamped their corporate operations.

If you’ve see the latest podcast, then you’ll know that the econ minor in me is obsessed with incentives. Will this GTM strategy work? ¯\_(ツ)_/¯

But after hours and hours of careful thought and deliberation, we agree that this approach maximizes our chances of becoming the $10B company we want to be.

And what can you do to help? Intros, intros, intros…

If you know a founder building enterprise agents, lemme know.

If you know an enterprise that is curious about how they can leverage AI safely, hit me up.

And if you know an investor looking to bet on the picks and shovels of the agent gold rush, venmo is @trey-layton

How am I doing?

I love hearing from readers, and I’m always looking for feedback. How am I doing with The Startup Breakdown? Is there anything you’d like to see more or less of? Which aspects of the newsletter do you like most?

Hit reply and say hey - I’d love to hear from you!

Cheers to another day,

Trey