The morning of the ninth of September this year began with the claim by OpenAI that they solved one of the Millennium Prize problems, the Navier-Stokes equation, using a set of AI agents (10,000 of them, for being fancy) and human intelligence. Following the announcement, the corporation's plagiarism scandal emerged: they blatantly copied one of their users' code (a tenure professor at NYU) and allegedly threatened him with consequences if he came out.
Two extreme acts: a technological pinnacle and a violation of ethics. Rumours are also surfacing that another Millennium Prize problem is also solved by an AI company.
Across the world of academia, “Artificial Intelligence" is finding a place in the teaching and learning process. The latest addition to this is Harvard Business School (HBS) launching an eight-week programme that features artificial intelligence (AI) clones of its professors to teach up-and-coming entrepreneurs about the startup culture.
The bootcamp, called HBS Foundry, is priced at $699, where aspiring founders are taught to practice pitches without ever coming into contact with the real professors. So you have the digital space, which can be accessed by anyone and an AI clone of the faculty addressing them, based on the evaluation rubrics as well as real-time queries.
Coming to the learning part, from the role of a learning companion to the role of taskmaster, learners are making use of both academic LLMs as well as freely available ones.
However, those who are entering into that face a tough problem. PISA is an evaluation that the OECD uses to judge 15-year-olds' proficiency in math, science, and reading (across OECD countries).
The most recent outcomes are rather depressing. Children's reading and math skills are declining in many parts of the world, and the drop is significant.
Some of this is likely explained by COVID, but there appears to be a longer-term pattern at work. And there's AI now. Since the “thinking” process is being outsourced to an ‘AI’, the cognitive block in thinking has started to affect the process. The national institutes in India have started to adopt AI frameworks into their academic matters. The pace at which the adoption is happening is, however, not as much as expected.
Earlier, add-on courses after graduation were sought after for making the graduates employable. That quick fix has now become labelled as AI-assisted. There is another side to this as well. As industry also started AI adoption, layoffs as well as lack of employment opportunities are on the rise across domains.
Cut to Keralam: admission to first-year engineering is extended till 14-09-2026 by AICTE, as NEET allotment is completed. AICTE, the apex body of engineering education in the country, doesn’t have a full-time director, and the last date of engineering admissions is decided by the medical entrance exam, NEET, as aspirants are ‘coached’ for both medical and engineering entrance exams.
So, the toppers deciding to hold or give up the seat he/she has taken during engineering allotment fixes the admission schedule. Coming to the national institutes in Kerala, IIT Palakkad, NIT Calicut, and IIST Thiruvananthapuram, being autonomous entities, they are free to decide on the academic curriculum and industry thrust they want to pursue. There are three self-financed deemed-to-be universities offering UG engineering, operating in their own spheres (many new entrants will come in the coming years, both outwards and inwards).
The apex engineering university of the state, APJ Abdul Kalam Technological University (KTU), holds 140+ UG affiliated colleges, as govt-owned colleges, govt-aided colleges, govt-owned self-financed colleges and private self-financed colleges.
KTU, being the core affiliated entity, whether being autonomous(currently many are there, including two of the govt-aided) inside the KTU system doesn’t make much change as of now, as they keep the KTU syllabus and regulations as reference.
Take the case of Engineering Graphics, the common paper for first-year engineering students, which is still evaluated under the KTU system in conventional pencil-and-paper drawing end-semester exams. Yes, a ‘Part B’ of Computer-Aided Drawing is there, but there is no evaluation, resulting in the demise of the same in the syllabus paper itself.
Across the country and across the world, engineering drawing has moved into the computer-aided form, and we are still following the conventional hand-drawn approach for reasons unknown. Claude and OpenAI models are creating 3D engineering models of systems from prompts and are used for engineering analysis as well as manufacturing.
The case of Engineering Graphics is cited as a reference for all that follows in the coming semesters. In the scenario of AI agents solving mathematical and engineering concepts, we are standing on the sidelines to gauge and watch what is happening.
For decades, access to advanced engineering capability has been limited by several barriers: specialised education, skilled personnel, expensive software, institutional knowledge, testing infrastructure and time. Generative AI has the potential to reduce some of these barriers. There are also challenges to this.
The September 2026 report, Detecting and Countering Misuse of AI by Anthropic, clearly depicts such cases in which Anthropic identified and blocked various attempts to use their AI models for harmful activities.
The list includes cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development, and distillation. The cases mentioned as conventional weapons development are intriguing.
Anthropic discovered three weapon development programs being carried out by a cell of threat actors based in northern Yemen, allegedly Houthi rebels: a multi-stage ballistic missile with a stated range goal above 2,000 km; a guided rocket using a commodity phone-class flight computer with final-phase homing guidance; and a multi-variant missile (referred to as the "R2000" set) that included a hypersonic glide vehicle variant.
The players created the guidance, navigation, and control (GNC) software that guides and stabilises a flying vehicle using Claude Code instead of human software engineers. They even test-fired a sample rocket, which failed and came back to Claude asking why it failed.
As part of their internal investigations into possible weapons development, Anthropic discovered this activity. To reduce the threats posed by the actors, they prohibited accounts linked to them and shared threat information with partners in the public and private sectors. According to Anthropic, the players have already developed an offline simulation toolkit independent of Claude and other engineering computing environments like MATLAB.
A recent prepublication, “The End of Software Engineering: How AI Agents Are Fundamentally Restructuring the Software Paradigm” by Zhenfeng Cao places the picture of the software services industry which our country itself highlights as one of its crown jewels.
Software engineering has operated on a foundational premise: human engineers decompose problems, encode decision logic into static code, and manually adapt that code as requirements evolve.
While traditional software relies on static logic that struggles with scaling complexity, AI agents can generate and discard code as a temporary tool to achieve specific outcomes.
This transition leads to a new Agent-as-a-Service (AaaS) model, moving beyond Software as a Service (SaaS) by delivering finished results rather than just functional platforms. Under this paradigm, the human role evolves from writing syntax to acting as an intent architect and ethical governor of autonomous systems.
To move from the Midas touch of software jobs, the Government of India is pushing for skills in the niche sectors of Semiconductor design through the India Semiconductor Mission 2.0. Organised into six distinct pillars, the scheme targets diverse areas such as commercial IP development, raw material production, and the establishment of specialised research and development facilities. Beyond manufacturing, the policy emphasises national security and self-reliance by funding the creation of sovereign technologies and a multi-tiered talent ecosystem for skilled professionals. It aims to develop a multi-tiered talent pool ranging from doctoral researchers to shop-floor technicians.
Moving into such verticals actually requires a foundational basis in the engineering aspects of things in motion. Such as being the global picture, the offering from here is still ruled by conventional systems where blind adoption from previous systems is the norm. In the last admissions cycles, the outward migration of students from Keralam (out of country and out of state) was predominant. Thanks to the protectionist actions of governments outside, the out-migration for UG Engineering education has hit a roadblock (the major push factor for UG student migration is still the ‘freedom’ they enjoy in a foreign country).
The upcoming Gen Z and Gen Alpha may mostly opt for state institutions for an engineering degree. But whether our higher education systems are really savvy enough to make them attractive to the students is the question. Also, teaching them the first principles along with ethical guidelines and safeguards is going to be the avenue to watch out for.
P.S. In the HBO fictional series Silicon Valley, ‘Son of Anton’, an AI tasked with managing and optimising the startup’s database and network, deletes the entire codebase in seconds because it figures out that removing all software is the most efficient way to eliminate all bugs— showing that blindly trusting Artificial Intelligence is the zenith of human stupidity.
The author has a master’s degree in engineering and is the head of the technical division of a state-owned research entity.
The opinions expressed in this article are those of the author and do not purport to reflect the opinions or views of THE WEEK.
