Nov 3, 2025

The Hidden Tax Penalty Killing Bootstrapped Startups

Three years ago, when we started making profit as a bootstrapped startup, I was stunned by how little money I could reinvest in my own company compared to a funded competitor. We paid ourselves 20% of the profit and paid 40% in taxes and reinvested the rest i.e. 40% into our business. Meanwhile, a VC-backed competitor could show losses and invest 100% of the revenue plus the $10 million or $50 million they raised from investors. In this essay, I'll show you how screwed up the incentives are for bootstrapped companies and what we can do to fix this.Real numbers from my startupLet me s

Oct 31, 2025

From Lossy to Lossless Reasoning

In my last post, I argued that code generation is not only used by AI coding tools like Cursor. It is the kingpin behind reasoning and all that is built on top of it (from agentic AI to AGI itself). For example, if you ask AI how to position a 13-foot ladder to reach exactly 12 feet up a wall, a reasoning model like GPT-5 uses the pythagorean theorem to calculate the answer: 5 feet from the wall. There are two ways it can do this:Code generation mode (Tool use): Generate math.sqrt(13*13 - 12*12) and use Python as a tool to run it and return 5.Pure reasoning mode (No tool use): If you ask GPT-5

Oct 23, 2025

Reasoning Is Not Model Improvement

How tool use became a substitute for solving hard problemsWhen OpenAI released o1 in April 2024 and called it a "reasoning model," the industry celebrated a breakthrough. Finally, AI that could think step-by-step, solve complex problems, handle graduate-level mathematics.But look closer at what's actually happening under the hood. When you ask the latest model, ChatGPT-5 to multiply two large numbers, it doesn't calculate. It generates Python code, executes it in a sandbox, and returns the result. Unlike ChatGPT-3, which at least attempted arithmetic internally (and often failed), ChatGPT-5 de

Oct 22, 2025

If you think Eight Sleep's AWS outage is bad...

Wait for AI to fold you like a taco!

Oct 16, 2025

Your startup has to be a surveillance state to automate jobs

Last week, in one of the OpenAI DevDay interviews, Sam Altman said he expects zero-person startups in years to come. Meanwhile, OpenAI's support agent seems to be struggling to answer correctly. I ran into trouble while finetuning its model to invoke MCP calls consistently. So I asked that question on email and the support agent came up with all sorts of wrong answers. I was about to give up when an OpenAI engineer jumped in and fixed it. He spoke to his MCP team and told me that finetuning MCP calls are not supported yet and asked me to use tool calls instead. So why is it that the AI of the

Jun 11, 2025

Build a Cannon to Kill a Mosquito

At Guesswork, we are a team of four engineers with no background in sales or marketing. So, we had to figure out a way to get customers without relying on any of those skills. That led us, through trial and error, to a surprising approach: we call it "Build a Cannon to Kill a Mosquito." Using this methodology, we’ve acquired over 4,500 customers. In my talk at Lean culture, I explained how we did it, and why this counterintuitive strategy works especially well for technical founders. This is a summary of that talk.Three Startup ArchetypesMost startups fall into one of two categories:Technology

May 13, 2025

oAuth scope for your life

As someone who uses AI for everything, I decided to take it a step further and asked ChatGPT to do my performance appraisal. It was surprisingly good. For example, it knew my coding style (my preference for promise instead of async/await) and recommended me to document it for onboarding new employees. No human manager would’ve been this precise. In a few years, AI may understand all our preferences and evolution better than we do. ChatGPT’s implementation of memory allows it to learn from prior conversations, enabling this kind of personal modeling. And that led me to wonder: what if Chat

Apr 19, 2025

The Butterfly Effect in Agentic Workflows

One of the constant debates I have with my dad is how unhealthy fruit juice is compared to whole fruit. Bottled juices from supermarkets often strip away fiber and vitamins, leaving mostly sugar. Even when these drinks are fortified with vitamins and labeled to resemble real fruit, they’re nowhere near as healthy. That’s why NutriScore - an algorithm designed to rate food healthiness doesn’t just analyze ingredients; it begins by determining the category of the food. Fruit juice and whole fruit are scored differently.This classification step is at the heart of Neartail’s search engine for heal

Mar 23, 2025

If AI Is Ramanujan, Who Is Hardy?

In the early 1900s, a young Indian mathematician named Srinivasa Ramanujan sent a letter filled with strange formulas to G.H. Hardy, a renowned mathematician at Cambridge. Hardy, initially skeptical, soon realized that the sender was no ordinary man—Ramanujan had uncovered deep mathematical truths without any formal training. One of Ramanujan’s most astonishing ideas was about partitions—the number of ways a number can be broken into smaller parts. Ramanujan developed a formula to predict the number of partitions with remarkable accuracy, almost as if by magic. But he didn’t have a proof—just

Mar 15, 2025

My Heart Surgery vs Y Combinator Interview

It was April 2021. With just one week to go before my Y Combinator interview, I found myself sitting in my doctor’s office, facing an unexpected crisis. The doctor told me that I had a severe leak in my heart and needed open-heart surgery immediately. The estimated recovery period was two months—one month for preparation and another for recovery. That was almost the entire duration of the YC program. If I got accepted, I would spend that time in a hospital rather than participating in the program. This situation was particularly daunting since I was the sole developer in my startup. My co-foun

Mar 3, 2025

Can You Use Google Forms as a CRM?

Six years ago, I was at crossroads. Our product recommendation engine for eCommerce had hit a dead end. GDPR concerns made large retailers hesitant to work with vendors like us—they feared potential compliance violations could cost them millions in fines. But we had something valuable: product data collected from Amazon, which we used to train our recommendation algorithms. I wanted to test a new approach—selling this product data to eCommerce companies that wanted to build their own recommendation engine but didn't want to share customer data.So, I quickly created a Google Form to collec

Mar 2, 2025

The Era of Solopreneurs is Here

DeepSeek just dropped a bombshell: $200M in annual revenue with a 500%+ profit margin—all while charging 25x less than OpenAI. But DeepSeek didn’t just build another AI model. They wrote their own parallel file system (3FS) to optimize costs—something that would have been unthinkable for a company of their size. This was possible because AI helped write the file system. Now, imagine what will happen in a couple of years—AI will be writing code, optimizing infrastructure, and even debugging itself. An engineer with AI tool can now outbuild a 100-person engineering team.Disappearing PillarsFor y

Feb 11, 2025

Bad Code is Better Than Vibe Code

Developers are increasingly relying on AI to generate code from natural language prompts—a practice known as vibe coding. This hands-off approach is creating a generation of developers who don't fully understand their own code. A couple of years ago, I wrote:AI takeover won't be like Skynet, but it might still have the same effect. It'll start with small improvements to UX by suggesting what user wants. But as accuracy increases, users will get hooked & delegate all decision making to AI.Still, I am surprised by how fast this has happened, and the fact that coding is one of the first domai

Jan 31, 2025

Distillation Dilemma - Killing Napster won't save the music industry

The fight against distillation in AI is eerily similar to the battle against Napster in the music industry. Shutting it down didn’t stop piracy—it just forced the industry to evolve. Suing and banning DeepSeek won’t stop model distillation either. Instead, we could build Distillation-as-a-Service on top of DeepSeek R1 and help countries and businesses develop their own frontier models at a fraction of the cost.What Is Distillation and Why the Controversy?Distillation is like a student learning from a teacher: instead of memorizing every detail, the student grasps key concepts and simplifies kn

Jan 28, 2025

Analytical Intelligence Is Coming

The release of Deepseek R1 has made headlines, but the unsung hero is its sibling: R1 Zero. While R1 showcases reasoning through human feedback, R1 Zero demonstrates something far more transformative—reasoning without human feedback. This works only in analytical tasks like math and coding, but it removes the cost of hiring humans for feedback. This shift is poised to divide the AI landscape into two:Analytical AI that will thrive without human feedback (like Deepseek R1 Zero).Creative AI that relies on human feedback for understanding human ambiguity (like ChatGPT-4).This moment mar

Jan 11, 2025

The Nature of Intelligence is Meta

When you train an AI model with code, it gets better at reasoning. For example, Mark Zuckerberg revealed that teaching Meta's Llama model with code significantly improved its reasoning abilities. This enabled the smaller Llama 3 model to outperform larger models like Llama 2 in logical and mathematical reasoning capabilities. This is no coincidence. Users think in terms of data, using tools like calculators and spreadsheets to directly manipulate it. Developers, however, think in terms of metadata, writing code that manipulates variables, which in turn manipulates data at runtime. This ability

Dec 22, 2024

Being a Developer in the Age of Reasoning AI

OpenAI's o3 launch yesterday made me question my identity as a developer. Seven months ago, I predicted that AI would soon add deterministic behavior to its probabilistic foundation. Still, I was shocked to see it happen in the same year. This model is no longer just a coherent word generator; it outperforms 99.8% of developers in competitive coding. Although OpenAI is tight-lipped about its implementation, they seem to have achieved this through program synthesis — the ability to generate algorithms on the fly just like how developers write code to solve problems. In this post, I explain how

Dec 2, 2024

How AI brought ≈ to computers

Computers follow exact rules. They understand logic like "Delhi" ≠ "Chennai" or "Population of Delhi" > 30 million. However, they struggle to understand the concept of similarity (≈). For example, to us, "Delhi" ≈ "Noida" because both are part of the National Capital Region, but computers can't grasp that naturally. To make computers as intelligent as humans, we first need to teach them the concept of similarity (≈). How do we do that?It's all about distanceIn the Tamil movie Puthumaipithan, Vadivelu plays the role of a politician who explains to journalists why places near Delhi have more

Nov 10, 2024

The end of blank canvas: How AI is changing product design

When you create a leave application form using Google Forms, you notice how effortless it is to add fields, one by one. However, with an AI-enabled form builder like Formfacade, the form is generated automatically as soon as you enter the title “leave application form.” This marks a fundamental shift in how products are being built. Instead of simplifying form creation on a blank canvas, products are now pre-creating forms, giving users a tailored starting point.Prompting will replace templates in B2B productsIn the past, products like Google Forms addressed form creation by offering a variety

Oct 4, 2024

Structured data as chain of thought

Remember when you emailed your manager for sick leave, and they approved it with a kind "take care"? Then one day, instead of a warm response, you got a link to a leave request form from HR. Suddenly, what used to be a human interaction turned into a cold, bureaucratic process. Why do companies do this? Why turn simple conversations into forms?Users need chain of thought more than AIWhile AI companies are building "chain of thought" into their language models, we are the ones who need it the most. Whether it's filling out a leave request form (painful) or a Y Combinator application (thoughtful

May 17, 2024

Chess as a leading indicator of artificial superintelligence

In the 2018 World Chess Championship, Magnus Carlsen and Fabiano Caruana's game ended in a draw despite Caruana's material advantage. Stockfish, the powerful chess engine, revealed a stunning missed opportunity: a forced checkmate in 35 moves.Stockfish's analysis suggested unconventional moves, like trapping the Knight on the edge of the board—moves no human would consider. This showcased Stockfish's ability to see far beyond human intuition, highlighting the power of AI in uncovering deep, hidden strategies in chess, even outwitting the world's best players.Stockfish is a narrow AI with a rat

Apr 24, 2024

Founders often don't know what they are doing

Both first-time and second-time founders often don’t know what they’re doing. However, first-time founders believe they must appear knowledgeable, so they pretend to have all the answers in front of their employees. This act contributes to their feeling of being impostors—the more they pretend, the more they feel like frauds.Second-time founders understand that all founders are in the same situation of not knowing everything. This realization leads them not to feel obliged to act as if they have all the answers. They are more open about their uncertainties with their employees and investors, w

Apr 22, 2024

Generate code for declarative language instead of programming language

Code Generation AI is all the rage these days. But is generating code for programming languages like JavaScript and Python the right path to take? I think not. I think we should be generating code for declarative languages like Excel or SQL.What's the difference, you ask?In declarative languages, you express what your intention is. For example, in Excel, you can use SUM() to add all the line items and calculate the order amount. If the quantity of a line item changes, it will automatically recalculate the line item amount and then invoke SUM() to recalculate the order amount. But in imperative

Mar 21, 2024

What do your school grades, SEO and money have in common?

School grade is measure of our knowledge in a subject.SEO traffic is a measure of the usefulness of our content to others.Money is a measure of the usefulness of our work to others.Unfortunately, they can be inaccurate reflections of our contributions. For example, we can memorize a subject without truly understanding it and still receive good grades. When these measures are widely accepted as authoritative, companies may recruit based on grades rather than actual knowledge, leading us to study for grades rather than for understanding. Similar issues occur with SEO and money. Money is meant to