context rot

How We Hit 94% Accuracy at 20% of the Tokens Across 46 Repos

The claim that a verifiable IR beats vector search and GraphRAG only matters if the numbers hold up. So here is the run. 46 Kubernetes-ecosystem repositories, 150,000 files, 8 GB of code, measured end to end. What we tested, how we scored it, and why feeding the model 22x more context made it cheaper and more accurate at the same time.

Why Your AI Gets Dumber the Longer You Talk to It

You're not imagining it. Research tested 18 frontier models and every single one degrades as the conversation gets longer. Here's the science behind it.

Semantic Code Search Is a Retrieval Problem. Context Is a Representation Problem.

Semantic code search keeps getting better at finding the right code, and it keeps disappointing teams who expected better answers from their AI tools. The reason is that those are two different problems. Finding code is retrieval. Understanding code is representation. Better retrieval over a poor representation has a ceiling, and here is why a verifiable code IR raises it.

Softmax Attention and the Dilution Problem: The Math Behind Context Rot

As context grows, softmax normalizes attention weights so each relevant token gets less attention. This mathematical property is why AI accuracy drops with length.

Why Vector Search and GraphRAG Both End in the Same Context Rot

Vector search and GraphRAG look like opposites, but they share the same final move. Parse the code, retrieve some of it, rerank it, and stuff it into the context window. That last step is where both of them rot, because a model that is handed a pile of chunks degrades the same way no matter how the pile was chosen. Here is why the retrieval layer is not the cure for context rot, and what is.