Professor Aaron Quigley, Dean of the ANU College of Systems and Society.
For aspiring scientists and established software developers alike, coding in the age of generative AI has changed radically. From Programming for Scientists, an introductory course for coding beginners, to Agentic Coding Studio, a more advanced offering for experienced programmers, the ANU School of Computing is teaching students at every level to harness generative AI — part of a broader effort supported by Google and Anthropic.
“The grants from Google and Anthropic really accelerate what we’re trying to do,” said Aaron Quigley, Dean of the College of Systems and Society, which hosts the School of Computing. “They give us the space to explore and experiment with the use of generative AI in education and research.”
Professor Antony Hosking, Director of the ANU School of Computing.
“Industry support gives us the capability to deploy these technologies into our curriculum quickly,” added Antony Hosking, Director of the School of Computing. “We are also building our own long-term AI infrastructure and local capabilities at the ANU.”
Generative AI refers to the modern variants of artificial intelligence, including LLMs like ChatGPT, that respond to prompts to generate content like text or images. These forms of AI have found widespread usage in generating code.
In Programming for Scientists, students learn about the power of data science applications for their work in areas like physics and biology. By leveraging generative AI programming to design custom programs for their fields, students turbocharge their use of data science.
AI-assisted programming lets students focus on creative decision-making
Professor Nick Barnes and Associate Professor Minh Bui, co-creators of the AI-enhanced Programming for Scientists curriculum.
Computing Professor Nick Barnes and Associate Professor Minh Bui developed the AI-enhanced course as a way for students without a computing background to reach further by using AI. They introduced students to software testing, proper syntax, and program lifecycles and encouraged them to be in the driver’s seat with enough working knowledge of coding to get the most out of the AI models.
Previous iterations of the course spent a lot of time learning the programming language Python, and less time on the applications to students’ scientific disciplines. AI-assisted coding is transforming how scientists approach large datasets and freeing up time to focus on the underlying research questions.
Students in the class got further than previous cohorts when the course did not include AI tools. They were more efficient with AI and could automate traditionally time-intensive tasks, freeing them up to focus on creative decision-making. Because of the lessons in rigorous testing and focused design, students could leverage AI’s strength with coding while mitigating its weaknesses.
The course was made possible in part thanks to a donation from Google to help support innovation in teaching AI within computing. The course embraced the idea that AI won’t replace programming but will replace some parts of it.
Google Senior Program Manager Marie Efstathiou visits the ANU School of Computing with Dan Andrews, Nick Barnes, Peter Höfner, and Antony Hosking.
“It’s still key to have people who understand how to drive the abstraction for their own work,” Barnes said. “The role of the programmer will shift in time. We’re trying to embrace the technology with students and find the best way to teach students to work with AI.”
Abstraction refers to an objective-grounded way to model the key components of a research problem and their relationships. Choosing how to specify this may not be straightforward. If you come up with a wrong way to model the problem, any way you try to solve it will be impacted. This is an area where human intuition and understanding continue to play a critical role.
Rapid app development in Agentic Coding Studio
Dr Ben Swift, Senior Lecturer at the ANU School of Cybernetics.
Programming for Scientists is not the only course at the ANU that is leveraging the power of generative AI to expand students’ use and knowledge of programming. School of Cybernetics Senior Lecturer Ben Swift teaches Agentic Coding Studio, which is designed from the ground up for the age of agentic AI tools like Claude Code as part of the human-centric computing major.
Agentic AI represents the latest advance in generative AI models. Instead of a single AI generating content in response to a user’s prompt, agentic AI provides the AI with access to external software tools, enabling it to use a search engine, run code and check the result, or send an email. Large teams of agents can split up a larger problem, prompting and coordinating with each other. This shift in the paradigm of using AI has led to leaps in their power and capabilities.
As part of the course, each week students receive a brief and specified program they need to write for and submit a new prototype app or website. They demonstrate the project in front of their classmates, and Swift leads a constructive critique. Swift compares the approach to that of art schools, where each week students bring in work for review by their peers and instructors. An example project might be to design a university course the student wishes to exist and build the website that would run it.
Senior Lecturer, ANU School of Cybernetics
“Software development is still about humans making things for other humans,” Swift said. “But now that we have little robots that will make all the tests go green, coding by itself no longer makes sense as a pedagogical task. Through the process of doing and explaining to colleagues and peers, we teach students how to use these tools to ship software that matters.”
The course relies on a generous $500,000 USD donation of API credits from Anthropic.
“Anthropic has been a perfect partner, giving us resources but giving me free rein to figure out how to use them in a meaningful fashion,” Swift said.
The credits are managed by School of Computing Associate Professor Alex Potanin. He sees computer science education at a pivotal point where instructors, industry, and government are all figuring out how to best use AI tools for tasks. He emphasizes workforce-integrated learning and industry support as more important than ever since companies increasingly ask for more AI in curriculums. By directly supporting computing programs that leverage AI tools, like that of the ANU, companies can better prepare graduates for AI-first workplaces.
Computer science fundamentals become even more important
Associate Professor Alex Potanin, coordinator of the Anthropic initiative at the ANU School of Computing.
Computer science fundamentals remain a critical part of education, as even if it’s easier than ever to create code, people still need to review the code, run rigorous tests, and confirm the program works correctly. Professional settings with large codebases have considerations like security, compliance, and long-term maintenance. While AI can vastly accelerate coding processes, programmers still need to decide what the problem is and define the way they want the program to work.
“AI is very good at writing the code for students, speeding up the code development, and accelerating the learning curve,” said Bui. “But students are still required to fully grasp the fundamental concepts to be able to ‘decode’ the AI-generated code, and to rigorously test the code, an area at which AI is still not that great. I believe AI should be used in a supervised manner: do not just rely on AI to make the decision for us.”
“Without the fundamentals, you can’t understand all the code being generated,” Potanin said. “Even if the way we teach is changing, computer science education will not be replaced. There will be an increase in demand for people who know these fundamentals.”
Anthropic CEO Dario Amodei in a fireside chat with Dr Aruna Sathanapally at the Anthropic Futures Forum.
Potanin recently participated in two international working groups on the use of generative AI within CS curriculums and the future of CS education: ACM ITiCSE Exploring the Impact of Gen-AI on Team-Based Computing Capstone Projects and NUS-Google Workshop: Future of CS Education and Curriculum in the AI Era.
The ITiCSE workshop focused on the use of generative AI tools in team-based computing capstones, using interviews with students, graduates, and employers. Their report will be released on 13 September 2026. It shows policy problems at universities without consistent AI use policies, as students risk the erosion of technical skills from reliance on AI. There is a need to rethink traditional university expectations of how we evaluate and credit students for what they do in the age of AI.
The NUS-Google Workshop focused on how to change CS curriculum in the case of generative AI. It looked at both staff and student perspectives and found that CS fundamentals are still needed, something that the School of Computing is focused on with new and ongoing initiatives in AI and education.
“In our report, AI-native competencies became important,” Potanin said. “Abstraction and computer science principles need to be taught, including making formal verification part of the core computing curriculum. The ANU is good at both fundamental computer science and AI.”
“AI is everywhere, in administration, marketing, research, infrastructure, and education,” Quigley said. “With Anthropic and Google’s support, all the experimentation in the School of Computing flows out to the college and the wider university. People on committees throughout the ANU understand what’s going on in computer science thanks to the School’s work.”
Hosking sees an opportunity for the ANU to position itself as a thought leader for Australia in the age of AI. The School of Computing has strengths in areas that current generative AI tools struggle with, such as symbolic AI, planning and optimisation, formal verification, mathematics of machine learning, and embodied AI where the system must operate in the real world and can’t bluff its way through challenges.
“These are all areas that will be the key to safeguarding generative AI systems as they are deployed in the real world,” Hosking said. “We have a role to play in the national conversation about the future of AI in Australia and the world.”
