Porting Graph:Easy to TypeScript with GPT-5.2 and Azad (tomisin.space)

A recent experiment detailed on Hacker News showcases the successful porting of the Graph::Easy library from Perl to TypeScript using the advanced capabilities of GPT-5.2 combined with the coding tool Azad. The original Graph::Easy, a significant library for rendering graphs, had not seen updates in 15 years and consists of 23,000 lines of code. The task, which was anticipated to be resource-intensive and costly, was completed autonomously by leveraging GPT-5.2's strong long-context understanding and autonomy. Over 24 hours, approximately 250 million tokens were processed, resulting in a feature-complete TypeScript version that precisely matched the output of its Perl predecessor, all for under $150. This achievement highlights the evolving capabilities of large language models (LLMs) in software development, emphasizing their potential for complex tasks such as code translation and refactoring. The experiment revealed that GPT-5.2 can effectively manage intricate tasks without direct human intervention, provided it is given clear parameters and evaluative criteria. The results illustrate a significant leap in AI-assisted coding, suggesting that models like GPT-5.2, along with tools like Azad, could redefine how developers approach legacy code. As the AI/ML community explores more efficient ways to specify tasks for LLMs, it holds the promise of drastically reducing the costs and time involved in software modernization efforts.