BlazeCobol (presentation - english version)

BlazeCOBOL — The AI-Powered Way to Learn COBOL Over 220 billion lines of COBOL power the world's banking, insurance, and government systems, processing $3 trillion in daily transactions. Yet the workforce that maintains this code is retiring, and universities stopped teaching COBOL decades ago. BlazeCOBOL bridges this critical skills gap by providing the first AI-native, browser-based COBOL learning platform — no mainframe access required. Target Users BlazeCOBOL serves three audiences: developers transitioning into mainframe modernization roles, computer science students exploring legacy systems, and enterprise teams upskilling employees for COBOL maintenance. Users need only a browser — no z/OS access, no JCL setup, no ISPF configuration. How Users Interact Users write COBOL directly in a browser-based code editor with syntax highlighting and compile against a real GnuCOBOL compiler running on AWS. The AI Learning Companion provides progressive hints, explains COBOL concepts in modern terms, and debugs compilation errors in real time using AWS Bedrock (Claude) and Watson AI. When a user gets a "syntax error, unexpected end of file," the AI doesn't just explain the error — it identifies the missing period, shows the corrected code, and teaches the underlying COBOL rule. We started the project manually, then we used Kiro to finally moved to IBM Bob.