If we add the proclamations from the AI industry, by way beta versions and products themselves and the announced timelines thereof, it should have added to much more in AI than where we have reached so far. But that is the nature of any field that makes gigantic progress where one success feeds into another to create an ecosystem of the nature one finds in AI today globally, not getting restricted to one geography. It is not denying the fact that some areas and companies are exceedingly well, others also feel the urge, and AI moves on. Mastering coding is the next holy grail.
It is easier said than done. We have been reading about the transformative work AI is doing in coding and the impact thereof by the tech leaders who are leading giant IT behemoths. Undeniably all of them want to usher as much of AI enabled coding as they can and reach somewhere near self-sufficiency in the foreseeable future. The latest contender to the crown is the well known AI company OpenAI. On16th of May 2025 it announced the launch of Codex. This is described as the yet.company’s most capable artificial intelligence (AI) coding agent
OpenAI’s leading brand ambassador took to social media to announce the research preview of Codex. Codex is powered by the o3 reasoning model. Sam Altman wrote in X, “Today we are introducing Codex. It is a software engineering agent that runs in the cloud and does many tasks for you, like writing a new feature of fixing a bug. You can run many tasks in parallel.” The final frontier is translating natural language into code almost instantaneously. What can it do as of now? It can act as a “virtual coworker” for engineers, helping them to write codes, fix bugs – all at an exceptional speed.
Leaving the fanfare aside, OpenAI stated that Codex can complete approximately 37% of requests. The intention is very clear: it is meant to make human programming faster rather than to replace it. What does it excel in? Most is at mapping … simple problems to existing code, which is generally described as “probably the least fun part of programming.” The real aim has been to find a way of getting code written without having to write as much code. What does it deliver as of now? Research on Codex states that when it attempted each test case a 100 times, it “generated working solutions for 70.2% of the prompts.”
THE ARE STILL MILES AND MILES OF RESEARCH TO SUMOUNT BEFORE WE REACH REACH THE SOMEWHERE NEAR THE REAL AUTOMATED CODE WRITING AGE.
Sanjay Sahay
Have a nice evening.