Seth Rosenbauer
CEO & Co-Founder
When we started Joggr, we thought we were solving the age old "our docs suck" problem. Then as AI coding assistants exploded onto the scene, something unexpected happened: developers started feeding our documentation directly to their AI tools.
We couldn't believe the results: dramatically more accurate outputs, tasks finishing twice as fast, and way fewer errors. The AI had the exact context it needed, so it didn't waste time (and tokens) running through the ls, grep, read, mcp cycle to continuously rebuild the context that is missing.
That's when we realized: we weren't building a documentation tool. We were building context infrastructure.
Anthropic views context engineering as the natural progression of prompt engineering. Prompt engineering is about writing and organizing LLM instructions. Context engineering is about curating and maintaining the optimal set of information during LLM inference, including everything outside of the prompts themselves.
Prompt engineering is how you write instructions. Context engineering is curating what information the AI has access to.
The shift matters because bigger context windows make the problem worse, not better.
Large language models suffer from Context Rot. As you add more tokens to the context window, the model's ability to recall information degrades.
The math is brutal: at 10,000 tokens, the transformer creates 100 million pairwise relationships. At 100,000 tokens, that's 10 billion relationships. Your 200k token context window isn't a feature. It's a liability if you're dumping everything into it.
What happens in practice:
1. Too Much Irrelevant Information - Feeding your entire codebase to an AI tool doesn't help, it hurts. The model wastes its attention budget on files it doesn't need.
2. Missing Critical Context - The AI doesn't know why architectural decisions were made, what was tried and failed, or the business context behind technical choices.
3. Disconnected Information - Code lives in GitHub, decisions in Jira, discussions in Slack. The AI can't connect the dots because the dots aren't connected.
We built an AI context engineer that automatically:
Results:
This is where we came full circle, everyone needs context: Developers need it to understand the codebase, AI tools need it to give accurate suggestions, new team members need it to onboard, and tech writers need it to keep docs current.
Context engineering isn't just a feature, it's foundational infrastructure you leverage everywhere. Joggr is that infrastructure. We build and maintain your context and the relationships between it: AI tools query it for structured context, developers browse it as documentation, and teams build custom tools on top of it.
As Anthropic's research shows, the key is "minimal, high-signal context at each step." That's exactly what Joggr delivers automatically, whether you're a human or an AI.
Welcome to the context era.
Book a demo to see how context engineering transforms AI-assisted development.
Want to dive deeper? Check out Anthropic's research on context engineering and Chroma's work on context rot.
