Matthew Berman describes building a PDF reader and other applicationsCode generation uses AI or another automated system to create source code from instructions, examples, schemas, or higher-level specifications. to meet his own needs. He argues that improving coding models, lower costs and more autonomous agentsAn AI coding agent is a tool-using AI system that can inspect, modify, and validate software within a repository. could make customized software accessible to more people.
Berman describes experiments replicating existing software and compares his reported token spendingAI inference economics examines the revenue, cost, utilization, pricing, and margins involved in serving trained models. with commercial licensing. His claims of full feature parity and future cost reductions are his assessments rather than independently verified measurements.
He considers open-source sharingOpen-source software is software distributed under a license that permits users to inspect, use, modify and redistribute its source under stated conditions. and acknowledges maintenance and security as potential obstacles. His expectation that improving AI will lower those costs is a prediction, not proof that generated software is secure.
Watch on YouTube
