Andrej Karpathy, a founding member of OpenAI and former AI director at Tesla, has stirred debate in the tech community with a bold claim that programming has undergone a fundamental shift in just a few months, and the industry is “nowhere near business as usual.”
Karpathy, who earlier introduced the concept of “vibe coding,” recently shared a detailed post on X describing the rapid progress in AI-powered coding agents and how they are already transforming software development workflows.
‘Coding Agents Basically Didn’t Work Before December,’ Says Andrej Karpathy
In his post, Karpathy pointed to a sudden breakthrough in what are known as agentic coding systems, AI tools capable of carrying out complex development tasks autonomously.
“Coding agents basically didn’t work before December and basically work since,” Karpathy wrote.
He stressed that the change is not incremental but abrupt, describing these tools as “extremely disruptive to the default programming workflow.”
nanochat now trains GPT-2 capability model in just 2 hours on a single 8XH100 node (down from ~3 hours 1 month ago). Getting a lot closer to ~interactive! A bunch of tuning and features (fp8) went in but the biggest difference was a switch of the dataset from FineWeb-edu to… pic.twitter.com/qragh6hJL6
— Andrej Karpathy (@karpathy) March 5, 2026
‘This Is What Post-AGI Feels Like’
According to Karpathy, the impact of such systems means the traditional approach to writing code may soon look very different from what developers have been accustomed to for decades.
“Programming is becoming unrecognizable,” he wrote. “This is nowhere near ‘business as usual’ time in software.”
Karpathy also hinted at how far AI workflows may already be progressing.
In a post shared on March 6, 2026, he wrote, “this is what post-agi feels like… i didn’t touch anything.”
The comment suggested a fully autonomous AI workflow, where systems can execute tasks end-to-end without human intervention.
A Weekend Project Completed in 30 Minutes
To illustrate the shift, Karpathy shared an example from his own experience over the weekend. He used an AI agent to build a video analysis dashboard for his home cameras.
The agent carried out the entire process autonomously, including, logging into a server, configuring SSH keys
Downloading and benchmarking an AI model, building a web user interface, and writing a full report.
Karpathy said the entire workflow took about 30 minutes, with the agent encountering errors, researching solutions, and resolving them on its own.
“It was all hands-free,” he wrote.
From ‘Vibe Coding’ to Structured AI Workflows
Karpathy’s latest comments mark an evolution from the idea of “vibe coding,” a phrase he introduced in early 2025 to describe casually prompting AI systems to generate code with minimal oversight.
At the time, he described the approach as enjoyable but mostly suitable for quick or experimental projects.
Now, however, he is describing something much more structured: developers deploying AI agents, assigning tasks in plain English, and reviewing their work in parallel.
“You’re not typing computer code into an editor like the way things were since computers were invented, that era is over,” he wrote.
Debate Over the Future of Software Engineers
Karpathy’s remarks sparked debate among developers in the comments on his post.
One AI developer asked whether such advances could dramatically reduce the number of engineers needed in software teams.
“Are teams of hundreds of people going to be replaced by a few chosen prompters?” the developer asked.
Karpathy suggested that while vibe coders are becoming increasingly capable, deep technical expertise may become even more valuable.
“Vibe coders are now able to get somewhere, but at the top tiers, deep technical expertise may be even more of a multiplier than before because of the added leverage,” he responded.
Who Is Andrej Karpathy?
Karpathy is widely known in the artificial intelligence community for his work in deep learning and AI systems.
He is currently the founder of Eureka Labs, a startup focused on modernising education for the age of artificial intelligence.
Earlier in his career, he served as Director of AI at Tesla and was also a founding member of OpenAI.
During his PhD at Stanford University, Karpathy was the architect and lead instructor of the CS231n deep learning course, which later became one of Stanford’s most popular classes.
Also Read:
Zubair Amin is a Senior Journalist at NewsX with over seven years of experience in reporting and editorial work. He has written for leading national and international publications, including Foreign Policy Magazine, Al Jazeera, The Economic Times, The Indian Express, The Wire, Article 14, Mongabay, News9, among others. His primary focus is on international affairs, with a strong interest in US politics and policy. He also writes on West Asia, Indian polity, and constitutional issues. Zubair tweets at zubaiyr.amin
