HubSpot's big conference is called UNBOUND now. They renamed it from INBOUND, and they built the entire thing around one phrase: "the context layer."
One session is literally titled "HubSpot's Context Layer: Your GTM Brain." There's another about "closing the context gap," and a third arguing that agents should run inside the same platform as your customer data so they "don't have to start cold." Their CPTO already wrote the thesis in a blog post: "The Real AI Race Isn't About Models or Data. It's About Context."
I found this out the way I now find out most things about my industry — an AI agent read the conference schedule while I was asleep and filed a note in my second brain.
That part isn't a metaphor. I run a thing called GTBrain. It's a folder structure of markdown files that an agent maintains, and it's also the honest answer to "what would happen if you automated your own note-taking." (Yes. The name. It's on the nose. I'm aware.)
So here's the setup: the vendor whose product I live in at my day job just renamed its flagship conference around a phrase that is, word for word, the thing I named my side project after and the thing my job is supposed to be building.
And here's the punchline, which I'm going to spend the rest of this post earning:
They're not wrong. They just can't build it. And I already own a version of it — and it's a flat tag that says "experimental."
Let me unpack that.
The sharpest line in the entire conference came from a CEO presenting on stage, and it's the whole thesis in one sentence: every sales and marketing team can use the same tools and the same LLM models, so intelligence alone isn't an edge. What actually differentiates a GTM team is the context layer — account history, content consumption patterns, ideal customer profiles... and the exceptions that skilled reps and marketers know.
Read that last clause twice. The moat isn't more data. Anyone can buy more data — enrich a contact with Clay, bolt on firmographics, whatever. The moat is the exceptions. The tacit, hard-won, non-obvious layer that separates a senior rep from a fresh one, and that a model doesn't have unless somebody captured it.
That's the real thing. And HubSpot can only say it, because HubSpot doesn't own its customers' context — the exceptions live in a senior rep's head, and HubSpot's product is a bet that it can coax them into a form.
I have the opposite problem, which is also the better one. I own a context layer. It's my side project.
IndianCoffeeBeans.com. 92 roasters, 1,400 SKUs of Indian specialty coffee, scraped from roaster sites, normalized by a language model, run through a review pipeline, and published into a structured catalog. I've been building the context layer for a product domain for two years — region, estate, elevation, variety, roast level, tasting notes, all canonicalized.
And the single most valuable dimension of that context — the one carrying the entire price premium — is a catch-all tag that says "experimental."
Here's the number: of the 1,238 coffees I was tracking in July, 241 carry an "experimental" process tag. That's 19.5%. Anaerobic alone is 111 coffees — the second-largest process category after washed. The catalog is growing into the experimental frontier faster than my taxonomy can name it.
And "experimental" is doing a lot of work. Look at any roaster's page and you'll see the flat vocabulary in situ: "Carbonic Maceration 1 · Experimental 1 · Aerobic 54." That one word is standing in for lactic fermentation, yeast inoculation, co-fermentation, and the newest one — cryomaceration, where a producer chills sorted cherries to 2–10°C so fermentation stretches from hours into days, borrowed from Burgundy Pinot Noir winemaking.
A buyer looking at two "experimental" coffees cannot tell that one is a cryomacerated lot — bright, floral, and must be rested two to four weeks or the volatiles die — from a carbonic maceration that's darker and fruitier and needs different storage. The knowledge exists. It just isn't attributed.
And the roasters will tell you exactly what that costs. Dhiraj Agarwal of Ground Zero Coffee Roasters, two years into experimenting with cryomaceration, said the quiet part out loud: "documenting the process clearly for buyers becomes essential to justifying the premium."
Read that again. The premium is the process. And the process lives in prose — a roaster's private notebook, a WhatsApp broadcast announcing a lot. The one thing that distinguishes a $5 cup from a $25 microlot is buried inside a catch-all tag and somebody's data log.
This is the same problem I already solved once, on a different axis.
When I was adding area figures to my region pages, I pulled from two government sources. The Coffee Board (2014) says the Coorg coffee belt is ~82,000 hectares. ISRO's 2024 satellite atlas says 135,796. Both official. Both defensible. 65% apart.
I didn't resolve the disagreement — I made it impossible to hide. Every area figure now carries a source and an as-of date, enforced by a database constraint, so an unattributed number literally cannot be written. The frontend can't forget to attribute, because the schema won't let it forget.
The process axis is the next instance of the same proof, aimed at the one dimension of the catalog that's growing fastest and carrying the most value. Context without provenance is just knowledge. And knowledge is free.
That's the whole thing. The difference between HubSpot's context layer and mine is provenance. HubSpot can say "context is the moat." I can enforce it. A schema you can't enforce is just a slogan with extra steps.
So while the GTM industry spends three days in Boston discovering "context is the moat" as a marketing word, I'm going to go make my most valuable tag stop lying. Turn "experimental" from a terminal label into a flag for "needs process extraction." Give the fermentation method, temperature, duration, and culture a schema. Make the one thing that justifies a $25 premium provable instead of prose.
Not because I'm trying to impress the RevOps crowd or the coffee people. Because I hold the reconciliation layer in both places — the CRM by day, the catalog by night — and I've spent a month learning the same lesson from every direction: the moat was never the data. The moat is the verified, first-hand, attributable layer around the data.
HubSpot named it. I'm going to build it.
If you're staring at your own stack and wondering where your context layer is, here's the question I've been asking myself all month: what's the "experimental" tag in your data — the flat label that's quietly hiding the most valuable thing you own?