AI Coding Hits the Context Window Wall
DevOps.com, Thursday, November 5th, 2026
Blitzy Field CTO Neeraj Deshmukh tells Mike Vizard why AI coding tools need structured codebase context, not just bigger prompts.
In this DevOps.com interview, Blitzy Field CTO Neeraj Deshmukh talks with Mike Vizard about the limits context windows place on AI coding tools working in large codebases.
He argues that vague instructions let small model deviations accumulate, and that context must link a requested change to the relevant modules, dependencies and intended behavior.
He describes Blitzy's approach: a knowledge layer built from the codebase, coordinated agents that produce an action plan for human review, and validation that runs alongside code generation.
He also warns that faster code generation can overwhelm review and QA, so teams need to change the software development lifecycle.