Siddharth Pai: OpenAI?s India launch will steepen its learning curve

This is not just a pricing move—it’s a strategic shift that opens the gates to one of the world’s most dynamic internet user bases. With nearly 800 million users online, India ranks second only to China, but it differs in one critical respect: it is wide open to global platforms. There are no Great Firewalls or mandated domestic clones. Instead, international apps and services compete openly. 

For OpenAI, India represents both scale and complexity—two ingredients essential to push the limits of any AI model. India’s linguistic diversity is especially instructive. While Hindi commands large numbers, it is only one among many languages with tens of millions of speakers. Add hybrid forms and you get a digital environment that can stretch the adaptability of any language model. It is a natural test-bed, where users don’t just consume AI outputs, they interrogate, correct and argue with them.

Also Read: What OpenAI has up its sleeve: An AI gadget mightier than the sword?

This is exactly the kind of organic feedback loop that AI companies pay good money to simulate. Training and refining a language model is an expensive, compute-intensive process. But just as important is real-world human feedback—the stream of nudges, corrections and clarifications that help fine-tune the model. In the US, AI firms hire teams of annotators to provide this kind of oversight. In India, OpenAI is creating the conditions for natural access to such feedback at scale. Even better, the company will be paid to receive it.

Consider this: ChatGPT misuses a Tamil expression. A user points it out, perhaps with a bit of sarcasm. The model learns. Another user catches a factual error in its cricket statistics. Correction logged. Over time, such corrections help the model grow more local, culturally aware and accurate. From OpenAI’s perspective, it’s a win-win. 

The ₹399 subscription creates a paying customer base, but also a self-updating model improvement engine. What was once a cost centre becomes a revenue-generating stream of valuable training data. This is economies of scale meeting the psychology of participation. Brilliant move.

Also Read: Sovereign AI is best developed as an enabler of soft power rather than hard

The upside for Indian users is substantial too. At this price, ChatGPT undercuts many conventional learning and productivity tools. There’s also a cultural alignment. India places a premium on intellectual agility—the ability to absorb, process and apply information quickly. It wants not just answers, but conversations. Unlike static sources like textbooks or websites, the tool encourages iterative questioning, back-and-forth reasoning and curiosity. 

That said, the rollout of AI tools always comes with its fair share of hype, especially in software development. For months, tech media and startup founders have touted the promise of Generative AI replacing coders or dramatically enhancing their productivity. Many developers shared positive anecdotes: Copilot wrote 30% of their code; ChatGPT helped unblock a tough problem, etc. The impression grew that coders would soon be managers of AI rather than makers of software. I have also held that view.

Also Read: Reasoning AI: Cracking the IIT test is a giant leap for machine-kind

But then a reality check arrived. On 10 July 2025, METR (Model Evaluation and Transparency Reports), an independent research organization, published findings that question those claims (shorturl.at/CCA9R). According to its analysis, there is no consistent evidence that Generative AI tools significantly reduce the time that experienced developers spend on complex coding tasks. While these tools can offer useful scaffolding—boilerplate code, syntax suggestions, simple debugging—they fall short when it comes to deeper tasks like architecture, optimization or interpreting nuanced business logic.

Worse, over-relying on AI tools can sometimes introduce subtle errors that take time to detect. What starts as a quick shortcut can turn into a lengthy detour. METR’s report warns that while AI can assist with routine coding tasks, it cannot think like a programmer. It can mimic logic, but not create it.

So, where does that leave the story of the AI-assisted coder? Still valuable, but not as revolutionary as some expected. Developers like using these tools; many feel more confident and creative with them. Yet, it’s a big jump to say that these tools can replace real expertise. Like calculators and spreadsheets in earlier times, improvement in the ease of working is real, but limited. 

Also Read: SubwayTakes, Jimmy Fallon and the Jonas Brothers may not be enough for Google to attract AI sceptics

Productivity increases are there, but mostly for repetitive, obvious or well-understood tasks.

In India, this distinction will become clear quickly. Developers, engineers and students will test these tools not just in theory, but in daily tasks. They’ll expect results and identify flaws. If the tool proves useful, it will succeed. If not, it will be fixed, criticized or discarded—sometimes all three.

That is both a challenge and an opportunity. OpenAI has entered a country that is vast, demanding and deeply committed to technological progress. The “cheap" ₹399 subscription is actually a bet that Indian users will help shape AI. Given our history with mobile, fintech and digital services, it’s a solid bet. But even in an AI-driven future, it seems that some jobs like providing context, critiques and corrections remain uniquely human. And India has no shortage of smart human beings.

The author is co-founder of Siana Capital, a venture fund manager.