07. What To Build (In A Graveyard)

If you want to build a startup in India today, the last place you’d look is the upskilling market. It is a graveyard. A few years ago, there were five hundred companies promising to turn anyone into a software engineer. Today, there are maybe sixty. What made them fail is instructive. They optimized for reach: taking any student who could pay. They optimized for breadth: teaching everything to everyone. They optimized for hope: selling the narrative of placements rather than building the capability to deliver them. When outcomes didn't materialize, neither did repeat customers. No word of mouth. No network. Just corpses.

I am personally against the business models these companies adopted. Some students from poor financial backgrounds are still paying EMIs to NBFCs because some sales rep was able to sell them a dream of a hot IT job when they were desperate for employment. I feel okay when someone charges premium to people of India 1; but I have seen these companies charging premium to people from India 3. I don't want to build any company that brings more pain in the society, than it does any good. Value creation still remains my core motto.

To build something that lasts here, I think you must not be another edtech. You must build something almost entirely different. What we should build instead is a highly skilled & talented community of technical builders: something closer to an academy, or even a startup accelerator for individuals rather than companies. And that distinction changes everything. The way to win in a graveyard is to do the opposite of what killed the previous inhabitants.

Start With People Who Don’t Need Motivation

We cannot build a sustainable talent transformation business by trying to transform everyone. Most edtech companies target the largest possible market: fresh graduates, tier-3 college students, people desperate for placements. This seems logical because the numbers are huge. But desperation is a poor foundation for learning. And they are the market that killed five hundred companies. These students arrive with low intent, high dependence on external motivation, and expectations built on promises of guaranteed placement. When you inevitably fail to deliver a miracle, when you cannot turn someone who hasn't built anything into a FAANG engineer in three months: they become your worst problem. They demand refunds, leave negative reviews on social media, put you in legal troubles, etc. The unit economics become terrible, and the feedback loop is poisonous.

We’re doing the opposite. We need diamonds from coal mines!

Somewhere in tier two colleges, in tier three cities, in basements and coffee shops, are engineers who are already teaching themselves. They stayed up at night building projects. They failed FAANG interviews and didn't just accept it, they learnt from it. They have specific pain. They have high intent. They don't need motivation; they need acceleration.

These people would eventually become very good at what they do. Not because of us. But because of who they are. They are the people who succeed regardless. What if we built something not for the masses, but for them? Our job is to compress that timeline, improve their odds, and give direction where there is currently noise. This is the fundamental reposition: we are not an edtech company. We are not even primarily an academy. We are building a professional community: a talent pool of technical builders, curated to an extreme degree of selectivity.

Selectivity is a feature, not a limitation. Y Combinator does not claim that everyone who applies should become a startup founder. But everyone who gets into Y Combinator is very likely to become one. This is the most powerful position a mentor or accelerator can occupy. It creates a virtuous cycle: selection drives quality, quality drives success, success validates selection, and our brand becomes synonymous with an outcome. We want to do the same thing for technical builders.

We need ruthless selectivity. Our admission process should be designed to identify and reject everyone except those who are already on their way to being great. We are looking for strong signals: our ideal customer is an engineer who has a portfolio, someone who has tinkered with side projects, candidates who can articulate what they've learned from failure. The one who has already failed once, maybe twice, and is hungry; not desperate, but hungry.

We Are Selling Transformations

We are not selling placements. We are selling transformation that inevitably leads to placements. The difference is crucial. Placements are outputs. Transformation is a capability. If we build real transformation in the right people, placements become a natural consequence. But if we promise placements directly, we have made a promise we cannot control. Instead, we should optimize for four fundamental learner needs.

1. Agency. Our students want to feel control over their career and capabilities. This means they are not passive consumers of content. They are building things. They are making decisions about what to learn and why. We create features and structures that expand their sense of agency: they choose projects, they direct their learning, they see themselves as architects of their own growth. When they leave, they feel like they drove their transformation, not that it was delivered to them.

2. Belonging. Humans are social creatures, and ambitious humans especially so. They want to be part of a community of people on the same journey, people as serious as they are. This is where the cohort model becomes essential. It is not just efficient: it is the entire point. Everyone knows everyone. Everyone sees everyone's growth. Everyone celebrates everyone's wins. This is what makes the experience stick. People leave with friendships, not just credentials.

3. Mastery. Real learning happens when we want to be genuinely good at something. Our curriculum should not be trying to teach "everything." It should be ruthlessly focused. What do technical builders need to know? What separates good engineers from great ones? Build that. Let them practice it obsessively until they feel competent and can see their own growth.

4. Proof of Work. In the attention economy, everyone claims to be an expert. The only thing that matters is demonstrated ability. Our students need to leave with a portfolio so credible, so undeniable, that hiring managers look at it and immediately understand what they can do. This is not a certificate. This is their body of work.

Product First, Lecture Later

Most edtech is lecture-first. It’s passive. You watch a video, you do a quiz, you get a certificate that no one believes in. For special talent like one we are seeking, we should not start with lectures; because they've been attending them in their college anyway!

We start with projects. Our students are building on day one. They choose a real problem. They start shipping. They fail, iterate, and ship again. This is how learning actually works in the real world. Alongside this, bring in workshops, mentorship, and whatever structure is needed; but everything serves the primary goal of building and shipping. A lecture is acceptable only if it unblocks the project. A theory session is valuable only if students see why they needed it.

Our obsession is outcome. What transformation do we want to create? What capabilities should someone have when they graduate? Now work backward. What projects develop those capabilities? What obstacles will they hit, and how do we help them through? This is the entire curriculum design process.

Most programs treat hiring as an end-stage activity. That’s backwards. We obsess about hiring signals from the first week: what companies actually want, what candidates can demonstrably do, where gaps exist, etc. Early cohorts may require manual job brokering: personal introductions, referrals, direct outreach. That doesn’t scale initially, but early-stage companies shouldn’t optimize for scale anyway. They should optimize for outcomes.

Experience Matters More Than We Think

Our students would eventually become very good on their own. But the timeline is what we are compressing. Instead of two years of self-directed learning, it is four months of intensive, guided, collaborative growth. We are not giving them something they could not get elsewhere. We are giving them something they could not get as fast elsewhere, in a peer group they would not otherwise access.

Likewise, we design the experience to be as experiential as possible. Weekly sprints where students present what they've built. Mid-cohort hackathons: 48-hour lock-in periods on weekends where teams build something real and present it. Competitions between teams. Regular fireside chats with people who have actually built things that matter. Everything is designed to create intensity, clarity, and visibility of growth.

They shouldn't feel any less engaged because it is virtual. They need to know each other. They need to feel the cohort as a real thing, not an abstraction. This is not a school, but it looks like a school: the same people, under the same roof (or connected to the same goals), moving through the same journey together. This is where belonging becomes real.

And yes, there should be ceremony. When someone graduates, it should feel like a real achievement. An oath-taking. A celebration. Something that marks the transformation that has occurred. This creates a ritual that bonds people together and signals to the outside world that completing this program means something.

Reputation > Revenue. Network > Net-worth.

This is where most edtech companies fail to think clearly. They think of a student as a unit: enroll a student, deliver a course, extract a fee, done. They optimize for linear growth. We are thinking about network effects, and this is the essential insight.

Our alumni are not our past customers. They are our future growth engine. A graduate becomes a leader in our community. They become a mentor to the next cohort. They refer their juniors: people in their network who are exactly like them, with the same high intent and hunger. They actively participate in our programs, in hiring panels, in mentorship. This creates a flywheel: selection + transformation \(\rightarrow\) outcomes \(\rightarrow\) credibility \(\rightarrow\) referrals \(\rightarrow\) better selection \(\rightarrow\) better transformation \(\rightarrow\) stronger outcomes \(\rightarrow\) stronger credibility. Each turn of the wheel makes the next turn easier.

This is how we will scale without marketing or without hiring sales reps. This is how we avoid the commoditization trap. We are not trying to reach everyone. We are trying to make our alumni so successful, so satisfied, so connected to our community that they are our primary growth channel. The quality of our alumni network is our entire competitive advantage.

Founder's Personal Touch

In the early days of an edtech, founder is the differentiator!

• Handwrite the first hundred welcome letters. Yes, this seems silly. Do it anyway. When a student opens an envelope with a handwritten letter from the founder, something shifts. They feel seen. They feel like they joined something run by real humans who care, not a corporate machine optimizing unit economics.
• Interview every applicant personally. We cannot scale this, and we should not want to. In the early days, this is how we'll calibrate our selection instinct. We will be wrong sometimes, but we will learn the signal. We will understand who succeeds and who doesn't.
• Manually broker the first jobs. When a student lands an offer, we helped make it happen. We introduced them to the right person. We advocated for them. This creates profound loyalty, and it ensures that hiring relationships are real, not transactional.
• Be present in every weekly war room where we review progress. Not every cohort, not every session, but in the early days, be there. Our presence matters. It signals that growth matters. It lets us catch problems early. It lets us understand our program deeply.

This is what Paul Graham and YC did, and it's non-scalable. But it's non-scalable in the way that builds something real. Once we have fifty cohorts running, we cannot personally interview every student. But we should have the earliest cohorts marked by this intensity. This creates the culture and the credibility that allows everything else to work.

The cult-like quality is not a bug. It is the thing that makes our community magnetic to the right people. We are not a professional service provider. We are a movement. We are a mission to take ambitious technical builders and accelerate their growth. This requires belief, intensity, and founder involvement that feels almost irrational from a business perspective.

Do it anyway.

Radical Transparency Is In Our Culture

We will publish our numbers, transparently! Cohort size, graduation rate, student dropouts, salary distribution, hiring challenges. Everything. This seems counterintuitive. In a market where everyone else is lying (or at least, carefully curating their statistics), we are going to publish the unvarnished truth. Some cohorts will not place well. Some students will not finish. Some will discover this program is not for them.

But here's why we do it: we are not trying to convince everyone that this works. We are trying to convince the right people that this works. The high-intent, intelligent student will look at our numbers and think, "Okay, this is real. This is not marketing. I can trust this." The wrong student will see those numbers and self-select out. Perfect.

And paradoxically, radical transparency about our failures builds more credibility than anyone else's claims of perfect outcomes. We become the only honest voice in a market full of lies. That honesty becomes our brand.

What We Are Not

• We're not a content platform. We are not trying to be the Netflix of education. We want the program to be experiential, but not directly gamified.
• We're not a software product that scales infinitely. Our primary scaling mechanism is network effects, not software efficiency.
• We're not a placement agency. We are building capabilities, not trading bodies.
• We're not a replacement for a degree or for self-directed learning. We are the accelerator for people who are already on their way. We are just making the journey faster and less lonely.

Our Hypothesis about Why This Should Work

Our moat is not our curriculum (anyone can copy that). Our moat is not our instructors (they can be hired away). Our moat is not our software (it's probably not even that sophisticated). Our moat is our community. It is the fact that after ten years, we have a thousand successful engineers who all came through our program, who all know each other, who all refer others, who all participate in mentoring and hiring. We are not a company then; we are an institution.

By the time we have built that, the market has changed. The dynamics have shifted. We are not competing on features; we are competing on trust, credibility, and network.

This is what the five hundred failed companies never understood. They thought they were selling education. They were actually trying to build institutions, but they lacked the patience and the founder intensity to do it right. In our case, we have to survive for as long as possible. Perhaps, one of the core motto for us should be: Don't die!

The First Cohort

Our first cohort is sacred. It will define everything that comes after. Do not take twenty students. Take eight if that's alright; but make them the best eight we can find. Be intensely involved. Make their transformation extraordinary. Make them our evangelists.

These eight people will tell twenty others. Those twenty will eventually represent fifty. This is how we want to grow without marketing. The network effects are the thing. Get that right from the beginning, and everything else follows. Get it wrong, and no amount of marketing or clever features will save us.

Risks/Blind Spots To Avoid

1. We are underestimating the difficulty in acquiring great talent. High-intent students exist. But they’re harder to find, skeptical, less impulsive buyers, and often already self-learning. So our funnel becomes small, slow and trust driven. YC didn't become YC overnight. Early years may look like manual hustle, slow cohorts, thin margins, and founder exhaustion. If you expect fast scale, this model will frustrate you. Biggest risk is that founder fatigue shouldn't kick in before network effects kick in. That's the real enemy, nothing else!

2. Selectivity shouldn't become elitism. Make sure admission process is not biased towards English fluency, urban exposure, or polished portfolios. You might miss raw talent from Tier ⅔ India. That’s fixable with good selection design.

3. Price premium, but avoid Income Share Agreements (ISA). This because of their legal enforceability, salary verification issues, and cashflow lag.

4. Employer partnerships are long-term gold standards to reach! Try for hiring partnerships, recruitment platforms, talent pipeline contracts, corporate sponsorships, etc. Not on day one, but this surely becomes huge later. Hiring trust compounds fastest.

5. If you are building community, position yourself by emphasizing identity, status, network access, reputation, etc., and not curriculum. This is closer to YC for engineers, Thiel Fellowship style, Research Lab culture, etc.

6. Go even narrower. Technical builders is too broad. Try ML Engineers, AI Engineers, MLOps Engineers, etc. Depth builds brand faster than breadth.

7. The 10x feature is our student's portfolios. They aren't just GitHub projects; our students build production tools, contribute in open source, publish research papers, and even deploy real customer-facing products.

Let's now formalize this as a problem statement that we can pitch to others when they ask, what are we building?