Griot is live. Heres why I built it.

I've been a creator, a ghostwriter, and I've worked at personal branding agencies.

At every level, the same problem kept showing up.

Not writer's block. Not lack of talent. Not bad AI tools.

The data was broken.

---

The manual loop

I wanted Claude to match my voice. Simple enough.

So I found one of my old podcasts. Downloaded it through Podchaser. Threw the file into Assembly AI. Copied the transcript. Pasted it into Claude.

That was one podcast. I had 40.

Then I remembered I also post on Instagram. So I went to my Reels, put the link into SnapInsta, downloaded the MP4, transcribed that too. Same thing with YouTube. Three out of five of those converter sites were down.

And then I had scattered context across my Notes app, my Notion, random voice memos. Ideas and memories that would have made my posts ten times richer. But I couldn't find any of it.

So much gold that never made it into a single post.

I just integrated Supermemory into a product and I was in awe by how many things I just forgot about. Years of ideas, sitting in folders I hadn't opened in months.

All of this could have been a system. But everything was static. Nothing was autonomous. Every time I wanted to write, I had to rebuild the context from scratch.

---

As a ghostwriter, it got worse

I'd reverse-engineer style guides from a client's existing posts. Google Docs. Maybe 3-4 pages. It worked for the first few posts.

Then everything started sounding the same.

When I was writing for Jesse, every post ended with "though, man!" and random all-caps words. Not because that was his actual voice. Because those patterns were in the style guide and the AI had nothing else to work with.

No variance. No learning. Just the same output, over and over.

And the research. My clients were founders. They'd been on Spotify podcasts, YouTube interviews, had personal websites, old LinkedIn posts, Twitter threads. I had to run the entire manual aggregation flow for every single client.

By the time I finished, the data was already stale. They'd just been on another podcast. Now I had to do it again.

No live database. No system. Just snapshots that expired the moment I took them.

---

At agencies, it's 20x worse

If a creator has fragmented data and a ghostwriter has it worse — an agency is that times every client on the roster.

The agencies I worked at had context scattered across Notion, Excel spreadsheets, Slack threads, Asana, text messages. Some didn't even store LinkedIn data or Instagram analytics.

The "system" was usually a long-running ChatGPT thread that someone pointed to and said, "that's our contextualized thing."

The posts were thin. So, so thin.

And here's the real constraint: agencies want to scale. If you're stuck at $50K a month, you want to get to $100K. But the only way most agencies scale is by hiring more writers. More headcount. More onboarding. More quality control.

Revenue stays capped by the number of people you can manage.

You can't build systems on a broken foundation.

---

The market is saying the same thing

This isn't just my experience. Every week on LinkedIn, agency founders are posting about this exact problem.

"I just can't close these deals right now, because that forces me to hire more copywriters — something which we've already done too often." — Founder of a LinkedIn agency, scaling from 26 to 60 accounts

"We blew the agency up to $70K/month. And it was a disaster. When the clients started pouring in, it broke everything. Our systems. Our minds. Our sanity." — Taylin John Simmonds

"Agencies scale with humans. If you don't know how to hire, train, motivate, and retain talent, this business becomes a nightmare." — Muhammed Rashid

"One writer maxes out at 4-5 clients. Not because they're slow. Because 80% of the time isn't writing. It's research." — Neo Lee, Co-founder at Imagine AI (YC F25)

"12+ hours spent weekly per client on content creation. 24 client profiles maxing out their 6-person team. 25% profit margins constantly under pressure." — Ruben Scholtz

The bottleneck isn't writing. It's context. And everyone is trying to solve it with more people instead of better systems.

---

Static context is why AI sounds like everyone else

That LinkedIn post with 368 likes and 1,240 comments that says "comment 'guide' for a link."

That's static context.

You download the guide, you inject it, and maybe the output is slightly better. More crisp, less obviously AI. But it's still thin. It has zero context about you — your voice, your stories, your evolving beliefs.

It can look good at first. But it looks good the same way it looks good for the 463 other people who downloaded the same guide.

I've seen so many of these AI writing frameworks going around. They provide a little bit of value. But they're also the reason there's so much thin, unpersonalized content flooding every feed right now.

And there are reverse network effects. The more people that use Claude with the same guide, the worse it gets. Everyone's writing converges.

Writing is wrapped around your life. Your beliefs, your voice, your experiences. All of those things change over time.

Writing is a dynamic problem. You can't solve it with a static solution.

---

What Griot actually is

Griot is the context layer.

Not a writing tool. Not a prompt library. The system that structures your brand data — LinkedIn posts, Instagram Reels, podcasts, YouTube, call transcripts, notes — into a living knowledge graph that your AI tools can actually use.

Connect your data sources. LinkedIn, Instagram, podcasts, transcripts, notes. Everything that makes each client's voice unique gets aggregated into one place.

It builds a knowledge graph. Not a static style guide. A graph that evolves with every new post, every new podcast, every new idea.

Your writers use it through the tools they already have. Claude, ChatGPT, Cursor — through MCP connectors that surface the right context dynamically. No copy-pasting. No stale docs.

Context compounds. The more a client uses Griot, the sharper and more personalized their content becomes.

Griot doesn't help you sound like everyone else. It helps you sound more like you.

---

This is my third attempt

I've tried to solve this problem three times.

The first time, I ran an agency on the side while working 80 hours a week. I was confident I could automate myself out of the loop. But I wasn't technical enough. I didn't understand the difference between static and dynamic context. It fizzled.

The second time, I built internal software for a personal branding agency — an early version of the MCP server and knowledge graph. They used it for a while. But the tool calls fetched too many posts, took too long, and the context wasn't scoped properly. I was spending an hour a week on it.

It fizzled again.

I keep coming back to this problem. Three times now. On top of everything else.

This time is different because I finally understand what went wrong. The previous versions failed because they treated context as a fire hose — dump everything into the LLM and hope. The real answer is structured, scoped, fast context retrieval. Give the AI exactly what it needs, when it needs it, and nothing more.

---

What I know to be true

Nobody else is building the infrastructure layer. Every competitor either does the content for you (expensive, doesn't scale), coaches you to write better (no data ingestion, no persistent context), or focuses on a completely different category.

Griot is the only system that says: bring your own AI. We'll structure the data it needs to actually sound like you.

The agencies that win won't be the ones with the most writers. They'll be the ones with the best systems.

Your agency runs on context. Right now it's scattered across Notion docs, stale style guides, and tribal knowledge.

Griot makes it a system.

---

We're live. Start your free trial or book a call if you want to see what structured context actually looks like :)