The Citation Problem Is Now a Content Strategy Problem
Every social media operator I know has quietly become an AI operator. We’re using Claude and ChatGPT to draft carousels, outline YouTube scripts, pressure-test hooks, and reverse-engineer why a competitor’s Reel popped. The output looks confident. That’s the problem. When I ask a model “how would Pieter Levels structure a launch tweet?” I get something plausible and fluent — and I have no idea whether it reflects what Levels actually said or what the model hallucinated from vibes. For creators, that distinction is the whole ballgame. Borrowed authority is only worth something if it’s real. That’s why The Brains, a launch from maker Victor Giurgiu, caught my attention — not because it’s a social media tool, but because it attacks the exact trust gap that poisons AI-assisted content work.
What The Brains Actually Is (and the Problem It Solves)
Strip away the launch-page language and here’s the mechanic: each founder, CEO, or investor is packaged as an MCP server built from their public talks, essays, interviews, and posts. Your AI client queries that server, retrieves what the person actually said, and — this is the load-bearing part — every answer links back to the source. The maker’s framing is blunt and, I think, correct: when he asked Claude or Cursor “how would Pieter Levels price this?”, the answer sounded confident, but he couldn’t separate what Levels said from what the model invented. The Brains exists to close that gap.
The current inventory is 37 brains, including Paul Graham, Pieter Levels, Naval, Lenny Rachitsky, Jason Cohen, and Ryan Hoover. You get one URL per brain or a single connector for all of them, and it works in Claude, Claude Code, Cursor, ChatGPT, Codex, and any MCP client. It’s free and MIT licensed, so you can self-host or build your own brain. The maker is explicit that it’s unofficial and built only from public material, with an email channel for anyone who wants something fixed or removed.
Why does this matter to someone who runs social accounts rather than a startup? Because the creator economy runs on borrowed frameworks. We quote Lenny on product-led growth in a LinkedIn post. We paraphrase Naval on leverage in a Threads thread. We cite Graham on doing things that don’t scale in a YouTube explainer. Right now that borrowing happens through memory, screenshots, and half-remembered podcast clips — and increasingly through an LLM that will happily fabricate a quote that sounds like Graham. A retrieval layer with citations changes the reliability of that entire workflow. My take: this is less a “founder tool” than a citation engine for anyone whose content leans on other people’s thinking.
Why this is an MCP story, not a database story
The interesting design decision isn’t the corpus — it’s the delivery format. MCP (Model Context Protocol) means the brains plug into whatever AI client you already use instead of forcing you into a new dashboard. In my own testing of retrieval-augmented setups, the tools that survive are the ones that live where the work happens. A separate web app for “founder quotes” would get opened twice and forgotten. An MCP connector that fires inside the Cursor or Claude session where I’m already drafting gets used daily. That’s the difference between a novelty and a habit.
How It Differs From the Tools You Already Pay For
The honest comparison set here isn’t Buffer or Hootsuite — those are scheduling and publishing layers, and The Brains doesn’t touch scheduling, publishing, analytics, or inbox. The closer competitors are the AI research and content tools: Perplexity for cited answers, NotebookLM for source-grounded document Q&A, and the general-purpose chat clients themselves. Against Perplexity, the difference is scope and depth: Perplexity searches the open web broadly; The Brains searches a curated, named-person corpus deeply, with the explicit promise that answers trace to what that person said. Against NotebookLM, the difference is packaging — you don’t upload sources, the sources are pre-assembled per person and shared across everyone who uses that brain.
There’s also a category of “creator knowledge” products — course libraries, swipe files, template vaults — that sell you someone’s frameworks. The Brains is a different animal: it doesn’t sell you the framework, it lets your AI retrieve the person’s actual words on demand, with receipts. For a social team, that’s the difference between a swipe file you skim once and a queryable second brain you actually interrogate before publishing a claim.
The incumbent comparison that matters most
The real incumbent is “just ask ChatGPT.” It’s free, it’s already open in a tab, and it will give you an answer in two seconds. The Brains only wins if the citation requirement is load-bearing for you — if you’re publishing under your own name and can’t afford to misattribute a quote, or if you’re building content that trades on a specific person’s credibility. If neither is true, the friction of installing an MCP connector probably isn’t worth it. That’s not a knock; it’s the honest segmentation.
What Creators and Social Teams Can Borrow From This
Even if you never install it, the launch is a masterclass in a content workflow shift I’ve been watching all year: retrieval over recall. When I scheduled 30 posts across five platforms last month, the bottleneck wasn’t writing — it was verification. Every post that quoted a founder, cited a stat, or referenced a framework needed a source check, and that check ate more time than the drafting. A citation-first retrieval layer collapses that step. Here’s how I’d actually use it in a content operation:
- Hook research with receipts. Before writing a LinkedIn post that opens with a Graham-adjacent claim, query the brain, pull the actual passage, and link it in your internal doc. Your editor stops asking “where’s this from?”
- Quote-accurate carousels. Instagram carousels that quote thinkers perform well, but misquotes get caught in comments. Retrieval plus citation turns a risk into a differentiator — you can put the source on the final slide.
- Script pressure-testing. For YouTube explainers, ask the brain “what did Naval actually say about leverage?” and compare it to your draft. The delta is usually where your script drifted into vibes.
- Repurposing with fidelity. When you turn a 40-minute podcast into ten clips and a newsletter section, the citation layer keeps the through-line honest across formats.
- Team onboarding. A new social hire can interrogate the same brains your senior strategist uses, which compresses the “learn the canon” phase.
The broader lesson: the creator-economy AI stack is bifurcating into generation tools (CapCut, Canva, and a hundred caption generators) and verification tools. Generation is commoditized. Verification is where the trust premium lives. The Brains sits firmly on the verification side, and I’d bet that’s the more durable half of the market.
Why TikTok and YouTube creators should care more than LinkedIn ones
LinkedIn rewards attributed thought-leadership — a named framework with a source reads as credible. TikTok and YouTube reward speed and personality, which is exactly where misattribution is most dangerous, because a wrong quote in a fast-cut video gets screenshotted and ratioed. A creator with 200K followers quoting a founder incorrectly in a Short has a real reputational exposure. The citation layer is cheap insurance, and for video creators who repurpose aggressively across TikTok, YouTube Shorts, and Instagram Reels, fidelity across formats is the whole game.
Where My Judgment Says It Falls Short
The launch thread itself surfaced the sharpest critique, and it’s the one I’d raise too: people change their minds. One commenter asked how the tool handles a founder whose 2008 position differs from their 2024 position — Graham being the obvious example. Another pointed out that the source date matters almost as much as the quote, because both answers can carry a valid citation while contradicting each other. The maker said he’d implement date handling, but as of the launch it’s a stated intention, not a shipped feature. That’s a real limitation for anyone using this to build content: a cited quote can still be a stale quote, and the citation alone won’t tell you which.
Other honest gaps:
- Corpus quality is unverified. The maker says brains are built from public talks, essays, interviews, and posts, but there’s no disclosed methodology for coverage, recency, or deduplication. “Not disclosed” is the right answer to “how complete is each brain?” — and that matters if you’re relying on it.
- Unofficial by design. The maker is upfront that it’s unofficial and built only from public material, with a removal channel. That’s the right posture, but it means the corpus can shrink or shift without notice.
- No social-native features. No scheduling, no analytics, no UTM tracking, no publishing. If you were hoping this replaces part of your Metricool or Later stack, it doesn’t. It’s an input to your content process, not the process.
- MCP client dependency. It works across Claude, Cursor, ChatGPT, Codex, and any MCP client, but you’re still dependent on those clients’ retrieval behavior and rate limits. API rate limits and context-window constraints are the client’s problem, not the brain’s — worth knowing before you build a workflow on it.
- The “37 brains” ceiling. The value scales with who’s in the corpus. If your content leans on thinkers who aren’t included, you’re building your own brain or waiting.
Who this is NOT for
If you’re a solo creator who never quotes anyone and writes purely from personal experience, skip it. If your team’s bottleneck is publishing cadence rather than content accuracy, skip it — go fix your scheduling stack first. And if you need verified, contractual-grade sourcing for regulated content (finance, health), an unofficial public-corpus tool is not your compliance layer. This is a research aid for people who publish opinions under their own name, not a legal citation system.
What I’d Watch / Test Next
Three concrete things an operator can do this week. First, install the connector in whichever AI client you actually draft in — for most social teams that’s Claude or ChatGPT — and run one real query against a brain relevant to your niche. Compare the cited answer to what the same model says without the connector. The delta will tell you instantly whether the citation layer earns its friction. Second, take your last five published posts that quoted a founder or framework and re-verify each against a brain. If even one was off, you’ve found your ROI case. Third, watch the date-handling feature the maker committed to in the thread — that’s the single upgrade that decides whether this is a research toy or a real editorial tool.
My take on the trajectory: the creator-economy AI conversation is moving from “can it generate?” to “can I trust it?” The Brains is an early, imperfect, open-source bet on the second question. It won’t replace your scheduling stack, and it won’t fix a bad content strategy. But if the next year of social is defined by AI-assisted publishing at scale, the operators who win will be the ones with retrieval and receipts — not the ones with the most confident-sounding chatbot.
