The Problem

Every PM has access to AI tools that can generate output in terms of list of features a product should build, feature PRDs, user stories, market research, product strategy, prioritization of features, etc.

This enhances the ability of a PM to generate output by 10-100x. You can write 10 PRDs in a day, whereas a single PRD used to take 10 days earlier. But LLMs aren’t reliable. They can generate very plausible outputs, but fail to judge the quality of these outputs reliably. 

And so humans (PMs) have to exercise their judgement more and more on the output generated by LLMs. This fundamentally shifts a PM job from creating the entire output like a PRD or user stories or market research documents, to judge a generated output through LLMs. 

To build a feature PRD in the previous world, a PM had to

• Go through hundreds of reviews to understand the pain point of users
• Synthesise and prioritise the pain points
• Brainstorm solutions for the pain point
• Do research on how others have solved similar problems
• Write a product requirement document
• Get feedback from others on the document 

In the AI-native era, the PM can 

• Upload all reviews to LLMs and ask them to summarise the pain points
• Take inputs from LLMs around prioritization
• Brainstorm solutions with LLMs around a specific pain point
• Ask LLM to create a report on how others have solved similar problems
• Ask LLM to write a PRD
• Ask LLM to critique the PRD
• Get feedback from others on the document 

Rather than going through the mundane, cognitively demanding tasks, a PM can choose to get everything done via LLMs. 

So what skill becomes essential in this AI-Native Era? The PMs who thrive aren't those who use AI most—they're those who can judge when AI's output is brilliant versus mediocre, when to override it, and how to find solutions AI would never suggest. 

Even when you build a strong judgement, there is another, often overlooked problem, with using AI: AI makes judgement harder to master.

AI short-circuits the cognitive work that develops judgement. When you get instant answers, you stop holding complexity in your head. MIT research (June 2025) found that 80% of LLM users couldn't recall what they'd written. The research concludes

"Convenience came at a cognitive cost when using AI, it diminished users' inclination to critically evaluate the LLM's output."

So the tool that makes you faster is eroding the judgement you need to use it well. This means that we have to deliberately use LLMs, rather than blindly using them. 

Fast, Reliable Product Judgement

Fast, reliable product judgement requires three capabilities. 

Advanced Product Skills: Product sense (taste to build superior products), product-led growth (ability to move metrics like activation, retention, monetization), strategic thinking (positioning, competitive analysis, long-term decisions). Start by mastering these. Fill the gaps where you are the weakest. 

Deep Domain Knowledge: Industry, competition, users—the context that makes your judgement specific rather than generic. Build this while you cover topics in advanced product skills. For example, while learning about industry structure, get a personalised industry report based on your company, role and industry. The adaptive learning platform of pmcurve delivers this report to you seamlessly.

Using AI for Productivity: Don’t rely on pre-built templates that outsource your judgement. But rather master advanced prompt engineering, agentic workflows, and AI prototyping and then build a personalised OS of prompts and tools designed for your context.

The program aims at helping you all of these.

Each of the modules in the program (shared with syllabus) conclude with building your own domain thesis + your own OS. You will read 300+ pages on your domain covering all the modules in this program, and build around 15 AI agents that will collectively build a highly productive OS for you.