Pandemonium

You know that feeling when you’re immobilized by decision paralysis because there’s a choice you need to make that almost feels like one of life’s major crossroads? That’s me right now. Any choice I make has what feels like an almost uncomputable amount of downstream effects on my life. This passage is going to be a brain dump of all the things I’ve been considering in making this decision.

Background

Roughly four years ago, I dropped out of UC Berkeley after my second year (May 2022) to work with some acquaintances on a startup (since then, we’ve become best friends). I was the first engineer there and the longest-tenured employee. We’ve gone through thick and thin together. From experiencing two major pivots, raising two rounds of funding, to building out an incredible team and working on fun problems, we’ve gone through the ups and downs of startup life and friendship together — practically coming of age in the process.

Now (more accurately, Dec 2025), a friend of mine sought me out to work on a project together. She mentioned that it had interesting applications in an early space. I’ve always been interested in new technology, so I happily agreed to spend a couple hours a week working together. Since then, the MVP has become usable, and it’s really promising.

So should I take the leap? And if so, when?

The World in 10 Years…

In Dec 2022/Jan 2023, we pivoted into the AI consumer/prosumer/SMB space with an AI website builder. This was when GPT-3 was available but before ChatGPT had launched. We were still called Relentlo at the time and rebranded to StyleAI to build out the AI-driven tool suite for SMBs who didn’t have dedicated marketing teams. Back then, AI was unreliable in latency, quality, and determinism. When we released the first variants of the text-to-website builder, we had to build in tons of redundancy and retries to make it work at all. There were definitely no production applications of the technology at the time.

Fast forward to today, and the use cases are emerging. Companies like Decagon, Sierra, Console, and Serval are automating human work with AI agents. Multi-agent systems are doing closed-loop reasoning with evals that help them converge on real-world ROI. The world’s largest SaaS companies are seeing a stock market “SaaS-pocalypse” (which I think is an overcorrection, because Salesforce is still likely to remain a business necessity for quite a while). What was unfeasible three years ago has slowly started becoming feasible in very specific use cases. The world’s scarcest commodity — human intelligence — is slowly starting to be commoditized. AI might not be good enough to replace all human use cases, but it’s starting to be good enough in some. And it can definitely augment almost every use case.

Now, how will the world look a decade from now? My take is that we will almost certainly reach the point where AI is good enough to handle most domain-specific use cases with evals. This is based on the current trajectory of improvement and the perhaps naive belief that even if AI stopped progressing today, there’s a plethora of new experiences that can ALREADY be created but haven’t been discovered yet. This, in and of itself, is a gold rush opportunity. Now, who knows whether we’ll reach a true general intelligence model, but that isn’t really a case we can prepare for anyway, as that’s the end of most white-collar work.

Personal North Star

Right now, it’s growth. My north star is to go where I can learn the most about building a generational company.

Why do I want to build a generational company? There are a few reasons.

  1. Asymmetric upside. Succeeding generates generational outcomes, and failure only materializes when you quit. At my age, the risk is minimal due to the privilege of not having too many extraneous responsibilities.
  2. Passion. It’s fun.
  3. Learning. You start broad and shallow, but as your company grows, your understanding of the field deepens and you naturally specialize in the technology and the industry. There’s nowhere else where you can learn nearly as much just by doing.
  4. Community. I want to work around the clock, and the startup community is the only one I’ve found filled with missionaries.

Opportunity Cost

Now, can Conversion provide these things? To an extent, I believe it can. It’s probably one of the best places in the world to be, sans my own company. Let’s measure against my north star metrics.

  1. The asymmetric upside is there. The team has proven to be resilient, operationally impressive, and battle-hardened through all of our pivots. We know how to gradient descent our way to success. However, after four years, I’m very invested in the business. Continuing at Conversion as an engineer silos my perspective in terms of engineering and people. The risk is that I’m already super overindexed on one company, so there’s value in diversifying. Additionally, I’ve captured most of my upside already, and I’m already certain Conversion will win. Let it ride.
  2. It’s soooo fun.
  3. I’ve learned a ton getting to where I am — the most I’ve ever learned up to this point in my life. However, I’m at a trough in my learning. The next big jump is when the business hyper-scales to a couple hundred employees. However, I don’t know when that’s going to happen, nor do I know what my role will look like in that world. I do know that I won’t be learning from a founder’s perspective but instead perhaps an engineering manager’s perspective. The question I’ve been agonizing over is this one. Not many companies ever hyper-scale like that, and Conversion is right on the cusp. That might be one of the most elite experiences one could have before embarking on their founder journey.
  4. The team is one of the best. Ever.

Other opportunity costs that I haven’t covered are conventional ones I don’t care too much about, like job stability, steady income, and a somewhat stable-ish trajectory (there’s still the startup risk). If I leave Conversion, I’m willingly giving up the above, which come pretty close to hitting my north star.

Relationships

Conversion has given me everything in terms of work, friends, and goals. The last thing I want to do is burn bridges here — I would lose what matters to me the most (the people). Right now, I’m working my hardest at Conversion to set them up for a potential future without me. I’m finishing what I perceive to be the most important engineering project for the business and training people to be capable of closing the gaps. As the first employee, I’ll always be useful to the engineering team, so there will never be an easy time to leave — but it will also only ever get harder. Not having the business to keep everything in our lives aligned will be tough, but I’ve met my best friends through this business and intend to keep those treasured relationships. My biggest fear is losing my best friends, either through my failure to gracefully handle this transition or through a long period of drifting apart.

It’s Now or Never

Now, the next question is: when? It’s my take that the AI agent era is analogous to the .com bubble or the mobile app era. It’s a once-in-a-decade opportunity where a new technology is capable of enabling new experiences that are 10x better than what existed before. Right now, there’s a bubble of arbitrage where what Silicon Valley knows about technology is a little bit ahead of what the rest of the world and other industries are aware of. That’s the perfect time to be first movers on the next set of generational businesses.

One thing I’ve also realized over the last four years at Conversion is that the team matters more than anything else. Having a wildly intelligent, resourceful co-founder is almost non-negotiable. I do feel like the stars have somewhat aligned in that I’ve met the right person, who has come with the right idea, at the perfect time.

Final Considerations

All in all though, I’m so unbelievably drawn to trying to start my own business that it’s almost unfathomable not to try. The idea quite literally keeps me up at night. Rationally, the risk-adjusted EV is close, and it’s arguable that it favors staying at Conversion. The odds that we hyper-scale and I learn how to build a team of hundreds are quite good. But I think the opportunity cost of not building a business during what I believe to be one of the most revolutionary technological shifts of my lifetime is even more tremendous. I’m a risk-forward guy generally, and there might never be another opportunity in my life where I’m this young, free, and irresponsible.