AI agent reliability is a context problem
AI agent reliability fails because the same task assembles different context every run. Non-determinism is a context engineering problem, not a model flaw.
Agent Reliability
All posts
The Wire Blog
Page 2 of 9
Context pruning helps agents, until it doesn't
Context pruning helps AI agents in one regime and hurts in another. A 2026 study of models from 4B to 284B maps when to prune stale context and when not to.
Context Pruning
Knowledge graphs vs RAG: when graphs actually win
Knowledge graphs vs RAG: graph retrieval wins multi-hop reasoning and global summarization, but vector RAG matches it on simple lookups at far lower cost.
RAG (Retrieval-Augmented Generation)
Context bloat: why long-running agents break
Context bloat is when accumulated tool-call output crowds out an agent's task. Tool calls, not window size, break long-running agents. Here is the fix.
Context Bloat
Chunking Strategies Decide What Context AI Sees
Chunking strategies decide which context reaches your AI. How fixed-size, semantic, and late chunking change retrieval accuracy, with 2026 benchmark data.
Chunking
How agents manage their own context window
Agent context management is shifting from fixed harness rules to learned, runtime decisions an agent makes about its own window. What the 2026 research shows.
Context Engineering
Demand paging for the AI context window
A 2026 systems paper found 21.8% of tokens in agent context windows are wasted. Demand paging treats the AI context window as L1 cache, not full memory.
Context Window
MCP authorization decides what context agents see
MCP authorization became a context control plane in 2026. RFC 8707 token scoping decides which sources an agent can ever pull into its own context window.
MCP (Model Context Protocol)
Stateless MCP: the spec is dropping session state
The 2026 MCP release candidate goes stateless: no initialize handshake, no session ID, any server instance answers any request. What stateless MCP means.
MCP (Model Context Protocol)
MCP Tasks: long-running work as context offloading
MCP Tasks let a server return a durable handle instead of a blocking result, keeping a long-running tool call's interim state off the agent's context window.
MCP (Model Context Protocol)