Most people don’t need a second brain. They need something a lot less sexy and a lot more useful: a context layer.
I was on the treadmill, not thinking about content, not thinking about AI — just wrapping another session of getting my reps in — when I was hit with what’s been bothering me about a specific corner of the AI-creator space. You know the corner: the usual suspects — the automation enthusiasts, the serial software entrepreneurs, the agency bros who just made the shift to AI, and the constantly running pipeline of any-old-fucking-guy who has quite literally built an entire channel around riding whatever wave crested that week. To their credit, that’s a genuinely smart content strategy. Chase the hype, get the views, move to the next hype before the algorithm notices you’ve said the same thing four times. But it leaves wreckage behind it, and the biggest piece of wreckage right now is the concept of the “AI second brain.”
You’ve seen the video. Some guy opens Obsidian, zooms into a web of interconnected notes that looks like a conspiracy board, and says some version of “this is my second brain now.” And the audience walks away thinking two things simultaneously: that’s cool as hell, and I have absolutely no idea what I’d do with that. Nobody’s lying to you on purpose here. But somebody’s using the wrong word, and the wrong word is doing real damage.
The Second Brain Fallacy
Here’s the uncomfortable question nobody in that content genre wants to sit with: what the fuck is the average person supposed to use a second brain for?
Not “is a second brain cool?” It is. Not “is it technically possible?” It is, for the right person with the right setup. The question is whether the thing being described in ninety percent of these videos is actually a second brain at all, or whether we’ve all quietly agreed to label this thing with a much cooler name because the much cooler name gets clicks.
So let’s actually talk definitions. A second brain — a real one, not marketing hype — has to clear two bars simultaneously: First, it needs enough data about you, fed into it over enough time, to meaningfully mirror and expand on the way you actually think. Not just what you’re interested in. How you reason. Second, and this is the bar almost nobody clears, it has to be able to make confederated decisions without meaningful human intervention. It has to know what needs to be done, go do it, and then tell you afterward that it’s done — not run a checklist past you first. Not “does this look good?” Not “what about this? what about this? what about this?” It just knows, and it acts, and it reports.
That second bar is the whole ballgame. A system that checks in with you constantly, no matter how sophisticated the checking-in is, isn’t a second brain. It’s a really, really attentive junior-level employee. And there’s nothing wrong with junior employees — we all start somewhere — but nobody markets a junior hire as an extension of your own cognition, they market them as help. Somewhere along the way, the AI-content world decided “extremely competent assistant” wasn’t a good enough pitch, so they upgraded the label without upgrading the product.
Most people don’t have a second brain. Most people have a personal assistant. And a personal assistant, properly used, is genuinely valuable. It’s just not what’s being sold.
Why Most People Will Never Clear the Bar
Here’s where I’ll actually give the space some credit, because I don’t think this is purely a “we’re early” problem that resolves itself in eighteen months. I think it’s a little bit of both — early, and also structurally unlikely for the average person, for two separate reasons that don’t fix themselves with better models.
The first is the data problem, and it’s a bigger problem than anyone wants to admit. Building something that genuinely mirrors your judgment requires an amount of referential data that even a technologically capable person doesn’t have lying around. I’d put myself in that category — I’m using AI-native workflows, I’m comfortable building automations, I’m not scared of the tools — and even I don’t have the writing samples, the invoice history, the years of social data, or the accumulated texture of a decade of decisions necessary to build something that reasons the way I reason. The average person has even less of that. The realistic ceiling for most people isn’t a second brain: it’s a continuously evolving, increasingly tailored store of preferences and patterns that gets a little smarter over time. That’s genuinely useful. It is also, categorically, not the thing being advertised.
The second problem is one almost nobody brings up: even if the data existed, most people’s lives aren’t configured to need fully confederated decision making from an AI system. You can have automations out the wazoo and skills stacked six deep and still not be running an operation complex enough to justify handing over the wheel entirely. Having the infrastructure and needing the infrastructure are two separate conditions, and both have to be true before “second brain” stops being aspirational marketing and starts being an accurate description of what you’re using.
And underneath both of those is a trust question that I don’t think gets asked honestly enough. Set the data problem aside entirely: Would the average person today actually feel comfortable with a system that acts without checking in first, then just tells you what it did afterward? I don’t think so. That’s not a smarter-assistant kind of trust. That’s a different category of psychological commitment — the same leap between “I trust my accountant to flag something weird” and “I trust my accountant to move money without asking me first.” Most people aren’t there. Most people shouldn’t be there yet. And that’s fine. It’s just not what a “second brain” YouTube thumbnail is selling you.
The Gym Membership Problem
There’s a version of this that should sound familiar if you’ve ever bought a gym membership in January. You don’t buy the membership because you did the math on marginal cost per workout. You buy it because the version of you that exists in the sales pitch — disciplined, showing up five days a week, transformed by March — is a more appealing person to imagine yourself being, than the version of you that is sitting on the couch right now. The second-brain pitch works the same way. Nobody’s actually evaluating whether they have the referential data, the operational complexity, or the trust threshold required to hand decisions off to a system with zero checkpoints. They’re buying the version of themselves that the pitch implies: someone important enough, complex enough, busy enough to need a second brain. That’s an aspirational purchase, not a functional one, and aspirational purchases are exactly how you end up with an Obsidian vault that looks like a conspiracy board and gets opened twice a month.
This isn’t a knock on wanting to be more capable. It’s a knock on selling capability people don’t have a use for yet, dressed up as a tool they urgently need. The gym membership isn’t the problem. The membership marketed as a personality transplant is the problem. Same thing here — the tool stack isn’t the problem, the label slapped on it is.
The Fix Isn’t Ambition. It’s a Word.
None of this means people should stop trying to build sophisticated AI systems. If anything, the ambition in this space is one of the better things about it. The problem is narrower and more fixable than “the whole industry is lying to you.” The problem is that we’ve picked the wrong word for the thing people are actually building, and the wrong word is quietly sabotaging everyone’s ability to understand what it’s for.
What most people are actually building when they wire Obsidian or Notion into an LLM isn’t a second brain. It’s a context layer — an external, structured store of information that makes whatever model you point at it meaningfully more useful. A markdown vault, on its own, is just a pile of markdown files. Layer an LLM on top of it with the ability to read and reference it, and it becomes something functionally different: context, on tap, for whatever the model needs to do next.
I’ll be the first to admit “context layer” is a markedly less sexy phrase than “second brain.” Nobody’s opening a YouTube video with “check out my context layer” and expecting it to trend. But it’s accurate, and accuracy here isn’t just a pedantic win — it’s the thing that actually tells you what the tool is for. “Second brain” describes a feeling. “Context layer” describes a function. And function is what tells a professional or a business whether they should care, and why.
Where the Actual Money Is
This is the part that gets buried every single time someone reaches for the cooler, vaguer word instead of the accurate one. “My Claude remembers my ideas now” is not a business case. Who fucking cares. It’s a neat parlor trick until you can point to where it hits the bottom line, and there are three places where it very clearly does.
Enterprise onboarding compression. The standard benchmark for bringing on a new employee is roughly ninety days. The first thirty are burned on paperwork, IT setup, and basic orientation — none of which requires human judgment, all of which requires human time. The next sixty shift toward actual task ownership and collaboration, which is where a new hire finally starts being worth what you’re paying them. A corporate context layer — SOPs, contact structures, institutional knowledge, operational history, all pre-loaded and structured for an AI system to draw on — could plausibly cut that window by half or more. Not because you’re replacing the human relationships that onboarding also builds, but because you’re stripping out the enormous amount of dead time currently spent re-explaining things that were already written down somewhere, just never in a form anything could actually use.
Think about what actually eats those first thirty days. It’s not deep, judgment-heavy work. It’s a new hire getting handed a binder — or worse, a Slack channel full of scattered pins — and being told to “just ask around” whenever something isn’t covered. Every question they ask is a question someone else already answered for the last new hire, and the one before that, and the one before that. A context layer isn’t magic. It’s just the first time that an accumulated institutional answer key gets put somewhere an AI system can actually surface it on demand, instead of somewhere it quietly rots until the next onboarding cycle rediscovers the same gaps from scratch.
A portable professional context layer. This one’s more personal, and I think it’s the more interesting of the two. Imagine a context layer that belongs to you, not your employer — how you like to work, what kind of outputs you actually want, how you prefer to communicate, built up over years and carried with you regardless of which company or which model you’re currently plugged into. It doesn’t matter if you’re a Claude person walking into a ChatGPT shop, or running Kimi while everyone around you is on something else entirely. Your working context travels. You could even take it a step further and design that personal layer to reconcile itself with whatever corporate context layer a given employer hands you on day one — letting your habits and their operational requirements sync instead of fight each other. That’s not a second brain either. It’s something a lot more useful: a version of you that shows up prepared, everywhere, without having to re-explain yourself from scratch every time the org chart changes.
Project- and case-specific continuity, siloed on purpose. The onboarding case is about compressing time. This one’s about not losing the thread at all. Think about a doctor handing a patient off from general practice to a specialist — right now that handoff runs through whatever fraction of the chart gets forwarded, and a specialist ends up rebuilding context that already existed somewhere, just not anywhere they could get to. Or a lawyer getting pulled onto a case mid-stream, who now has to get spun up on every document, every argument already tried, every angle already exhausted, from scratch, usually by sitting in a room while someone more senior recaps months of work from memory. Or any heavily project-based creative work, where a production’s entire history — briefs, revisions, rejected directions, the reasoning behind why — lives in someone’s head or scattered across a dozen threads instead of anywhere a new collaborator could actually pick it up.
A context layer built around a specific case, patient, or project solves two problems at once. The new person on the team gets dropped into the same context pool instead of a secondhand summary of it, and the data stays siloed to what it’s actually for — you’re not dragging an entire firm’s or practice’s worth of unrelated information into a system just to hand one new person what they need for one matter. That’s not the same shape as the enterprise-onboarding case, where the context layer is general and organizational. This one’s deliberately narrow, scoped to a specific piece of ongoing work, and that narrowness is the whole point.
Say the Accurate Thing
Here’s the thing about hype cycles: they don’t just oversell a product, they actively bury the more interesting question underneath it. “Do I need a second brain” is a question almost nobody actually needs to answer, because almost nobody is building something that clears that bar, and honestly, almost nobody should be trying to yet. “Do I need a context layer” is a completely different question, and it’s one that actually applies to you whether you’re a solo operator, a professional navigating three different AI tools across two jobs, or a company burning sixty days of productivity re-explaining the same institutional bullshit to every new hire.
The average person doesn’t want a second brain. They want less friction and more time back in their day, and they’ll take it in whatever form actually delivers it. That’s not a smaller ambition than the second-brain pitch. It’s a more honest one — and honestly, it’s the one worth building toward, because it’s the one people will actually use.