Coding in the Age of AI: Rethinking Computer Science Education

For some, the journey to the center of Silicon Valley can follow a predictable map, but for Sanjay Srivastava, the path was paved with equal parts luck and a relentless drive to build. Long before he became the CEO of Vocareum, Srivastava was an undergraduate in India who eventually found himself in the computer vision lab at the University of Illinois Urbana-Champaign. It was there, surrounded by the hum of early workstations and a rare Pixar machine, that he discovered the sheer joy of creation. "I just realized I was having so much fun coding," he recalls, noting that having access to that hands-on, experiential space changed the trajectory of his life.

That spark led him to leave his PhD program early to "go code for a living". In the late 1980s, amidst a deep recession, he sent out fifty resumes to any company that mentioned coding. He eventually landed at a semiconductor design firm where nearly half of his colleagues went on to become CEOs. After decades of building and selling successful tech companies, Srivastava reached a point where he was too young to retire but hungry for a "real sense of purpose". That purpose manifested in 2014 with the founding of Vocareum, a platform designed to bring that same hands-on, experiential magic he felt in the lab to students everywhere.

A world of “euphoria and panic”

Today, Srivastava looks at the landscape of education and sees a mirror of his own early career: a moment defined by "extreme euphoria and extreme panic". The catalyst is, unsurprisingly, artificial intelligence. We have moved through the era of computing literacy, where the world demanded everyone learn to code, and through the era of data literacy, where business schools fought to make every graduate data-fluent. Now, we have arrived at the era of AI literacy.

For many educators, this shift feels like a threat. Department heads in computer science are left wondering what their role is if a machine can generate a hundred lines of Python in seconds. But Srivastava views this not as the end of the programmer, but as a fundamental promotion. He uses the analogy of an orchestra: programmers are no longer the individual players of the instruments – they are the conductors.

The skill is no longer just in the syntax, but in the orchestration. Srivastava recently put this to the test himself, spending a few hours building agents to scrape his emails for unaddressed leads and research customers in Slack. "It was an eye-popping moment," he says, realizing that he had touched eight or nine different technologies in a single afternoon. It immediately highlighted a new necessity: the need for "proper kind of controls and guardrails".

Defeating the "5% problem"

The panic in education often stems from a misunderstanding of how to use these tools. Many institutions have attempted to implement "AI tutors" or chatbots as a side-car to the traditional curriculum. Srivastava argues that this approach leads to the "5% problem". If you leave an AI tool on the periphery, only about 5% of students will engage with it, and typically, those are the high-performers who don't actually need the help.

To truly move the needle, AI must be "deeply integrated into the teaching and learning environment". It shouldn't be something students go to; it should be something that "pops up" exactly when they are struggling with a specific line of code or a complex math problem. This is the difference between an assistant and a "digital twin" that knows exactly what a student understands and where they are hitting a wall.

The results of this deep integration are not theoretical; Vocareum and UCSD have already worked in partnership on a project involving math literacy. By scaffolding an AI tutor directly into the pre-calculus coursework, the institution saw the failure rate drop by a staggering 70%. When AI is an "aid rather than something on the side," it democratizes access for the students who were previously afraid to ask questions in a lecture hall of 1,400 people.

Redefining mastery in an automated age

As the conductors of this technology, students must be measured differently. The old ways of testing are being circumvented by agents that can take quizzes or tools that can provide answers from a simple screenshot. Srivastava acknowledges that for high-stakes summative assessments, many schools are returning to "completely monitored and locked down" computer-based testing facilities.

However, the more exciting shift is in "formative assessment". Instead of just grading the final output of a coding project, educators are now using AI to probe a student's understanding. A student might submit a functioning app, but the AI then asks them to explain different portions of their logic. If they cannot explain why the code works, they haven't achieved mastery. This pushes education toward "authentic assessment," focusing on the deep understanding required to manage AI effectively.

Making scalable personalization in education possible

Srivastava remains an optimist despite the "doom and gloom" narratives. He does not believe we are in a "Model T moment" where the old ways are simply being rendered obsolete. Instead, he sees a "calculator moment", describing a shift that removes the drudgery and allows for higher-level thinking.

The ideal goal of education often comes back to personalization, but the cost has been a permanent barrier. Srivastava’s own grandson attends a school where every interaction is one-on-one. While that model is highly effective, it is traditionally impossible to scale to the masses. AI changes that equation.

"I'm hoping AI is the one that just gets them closer," he says, referring to educators' ability to deliver one-on-one levels of attention to hundreds of students simultaneously. By using AI to handle the rote explanations and the initial hurdles, teachers can move toward "flipped classrooms," where they spend their time walking through the room and helping small groups with real-world projects.

Ultimately, the goal is to produce graduates who have a sophisticated understanding of what AI is not capable of. As we outsource more of our productivity to machines, we must be careful not to outsource ethics or judgment. The future of work and education isn't about the machine replacing the human; it's about the human learning to conduct the machine with wisdom, purpose, and a bit of that same fun Sanjay Srivastava found in a computer lab decades ago.

Hear the full conversation with Sanjay Srivastava over at our Educast3000 podcast.