AI can't skip basics

India’s school system has a new AI story to tell. On April 1, Union education minister Dharmendra Pradhan launched a CBSE curriculum in computational thinking and artificial intelligence for Classes 3 through 8, to begin in the 2026–27 academic session. It is meant to build logical reasoning, pattern recognition, problem-solving and early AI awareness long before students reach high school. (thehindu.com) That sounds modern because it is modern. It also hides the harder question. What, exactly, is a child using to do all that thinking? The answer is not AI. It is language. The new curriculum for younger grades is threaded through existing subjects and delivered through worksheets, textbook-linked activities, puzzles, written exercises and teacher observation. In other words, children are expected to read instructions, interpret prompts, discuss ideas and write responses before the “AI” part means much of anything. (thehindu.com) This is the point the new reporting in *The Hindu* drives home. The curriculum treats computational thinking as something embedded across mathematics, science, language and social science. For Classes 3 to 5, it is not a separate technical subject. It is a literacy-heavy way of moving through ordinary lessons. A child who reads fluently can experience a puzzle as a puzzle. A child who does not will hit the text first and the thinking task second. (thehindu.com) That would be less worrying if India had already solved the basics. It has not. The National Education Policy called foundational literacy and numeracy the system’s “highest priority” and said the rest of the policy becomes relevant only after children can read, write and do arithmetic at the foundational level. The NIPUN Bharat mission set the target of universal foundational literacy and numeracy for primary-school children by 2025, later framed operationally as achievement by 2026–27. (nipunbharat.education.gov.in) The latest national snapshot shows why sequencing matters. ASER 2024, a large rural survey covering 649,491 children in 17,997 villages across 605 districts, found improvement in basic learning after the pandemic slump, but not mastery. In Grade 3, only 23.4 percent of children could read a Grade 2 level text. Enrollment is high. Basic reading is not. That gap is where flashy curriculum design goes to fail. (pib.gov.in) None of this means AI has no place in classrooms. The second piece, in *The Tribune*, points to a narrower and more believable use. Students often know material in theory but struggle to express it aloud. Practice-based language learning, presentations, discussions and contextual exercises can build familiarity. AI tools can help there by giving students more chances to rehearse, get feedback and keep working outside the thin slice of time a teacher can offer each one. (tribuneindia.com) That is a very different claim from saying AI can replace the slow work of learning to read, listen, speak and write. It cannot. Even UNESCO’s guidance on generative AI in education pushes a human-centered, age-appropriate approach and warns that schools need policy, validation and safeguards before treating these systems as normal learning infrastructure. The technology may be new. The dependency is old. Students still need the cognitive habits that let them follow directions, sustain attention, test an idea and notice when an answer makes no sense. (unesco.org) So the useful version of AI in school is the smaller one. Let it coach pronunciation. Let it extend speaking practice. Let it generate another round of exercises after class. But do not confuse that with the foundation itself. CBSE’s own rollout quietly reveals the truth: in Classes 3 to 5, the new program is allotted about 50 hours a year and works through textbook chapters, written prompts, puzzles and classroom activities. The machine may be the headline. The real gatekeeper is still the child who can read the page. (hindustantimes.com)