How can my lifestyle business survive AI?
Surrendering alpha versus consciously defining alpha
I briefly contemplated starting a lifestyle business after I quit my job at DeepMind in 2024. I’d spent years on the core Gemini team, and wanted to break free from The Man. Yet because I’d spent years teaching LLMs to generate code, I couldn’t unsee their broader impact on the viability of entrepreneurship without VC funding. I started tugging at the question of what it’d take for me to start a lifestyle business. I sat in this question as I started talking to and helping entrepreneurs create value with AI.
A distinctive pattern started to emerge. Many solopreneurs started feeling overwhelmed and under pressure to continuously ship as coding agents improved, leading to challenges within their businesses. If you’re a solopreneur with a lifestyle business, this essay offers a tool called an Alpha Ticket. It’s an ongoing practice that’ll help you better manage this overwhelm, by helping you clarify and define your business’s alpha compared to the market.
This essay follows Rohan, a fictional composite of real patterns I’ve seen in the wild. If anything here resonates, could you please help me out by giving me some quick feedback via comments or email at varun@doubleascent.com? It puts the wind beneath my sails, and gives me more data points of what does/doesn’t work for people.
Rohan quit his Google job to start SpeedLaunch. It was a SaaS company that made ~$300k/year and sold opinionated services to help entrepreneurs rapidly zero-to-one new businesses.
His initial excitement around AI coding agents turned to fear as customers started using them to generate code that his company used to sell them. As the models started improving, it became harder to sell to new customers and retain existing ones. Fortunately, he wasn’t screwed yet. His customers often got stuck navigating rare, but important edge cases that often befuddled the LLMs. Edge cases that his libraries often covered due to his years of experience at Google. However, it begged the question of how could his company survive AI?
He started spending more time on X when Claude Code took off, since that was where the latest news was. Eventually, his regular five minute breaks on X started becoming forty minutes of doomscrolling. It’d end with a frenzied desire to “do something”. Claude Code made it trivial to build new features, so he’d rapidly fire off lots of parallel Claude Code tabs. Often as a reaction to something he saw online, or feedback from churned customers.
Using Claude Code this way increased his velocity, but also rapidly bloated his codebase, his product and introduced lots of bugs. Spending more tokens automating routine tasks like code reviews, etc. didn’t work as expected. It indeed increased his PR throughput but didn’t change the overall trajectory of his bug tracker. Customers increasingly noticed and started churning due to all the bugs. The bloat made it harder for new and existing customers to understand what the product did and how it’d help them. Specifically, why they should pay for SpeedLaunch instead of putting the money in a Claude Max subscription so they could do everything themselves.
This became untenable, so he clamped down with a sprint planning process he learned at Google. He created two-week cycles, triage buckets, etc. It didn’t totally turn around customer perception, but it stopped the bleeding.
Eventually, he saw a post on X that triggered him into abandoning his sprint planning process, and falling back into frantically shipping PRs with Claude Code. A competitor had seemingly vibe-coded a sloppier SpeedLaunch clone, but it got more eyeballs than SpeedLaunch ever did. Falling back into frantic shipping started reintroducing the bugs and bloat, which again hurt his bottom line. He realized he was stuck, and getting increasingly stressed.
He’d left Google to give himself and other founders more sovereignty in their lives. He wanted a lifestyle business that was aligned with his deeper values.
Yet he kept finding himself torn between panic shipping and bureaucratic self-control. His panic fueled a frenzied productivity allowing him to ship faster, but made his offerings less coherent. His bureaucratic self-control created a planning cage that indeed slowed him down enough to reduce the chaos, but was too sluggish to meet the moment. Any planning under pressure started to feel like letting competitors pass him by, and any attempt to ship quickly started to feel like restarting the chaos that had already damaged customer trust.
His frenzied productivity was driven by the legitimate fear that being too slow would leave him behind. He valued his sovereignty from The Man too much to simply ignore it. The bureaucratic cage was driven by the legitimate fear that an absence of sufficient structure would undo years of hard-earned trust. He valued his responsibility to his customers too much to look past it.
All this was driven by the rapidly increasing code generation capabilities of AI. They suddenly started reducing the gap between the value SpeedLaunch could create, and the baseline that someone could get for “free” from within a Claude Max plan. The size of this gap of value creation, or alpha, dictated how much Rohan could charge for SpeedLaunch, or how many customers would stick with him. It was scary how fast SpeedLaunch’s alpha was shrinking, and he was stuck between a rock and a hard place.
Getting unstuck involved shifting his perspective from unconsciously surrendering SpeedLaunch’s alpha to his circumstances, to consciously defining it instead.
Merging frenzied PRs triggered by FOMO on X, random customer feedback or competitors copying him surrendered his capacity for creating alpha to his reactivity. Fear-based sprint planning rituals slowed down damage to his product from his AI-fueled frenzy. But it surrendered his capacity for creating alpha to Google’s bureaucratic processes.
Consciously defining alpha involved an ongoing practice of honestly asking himself what unique value he could create using AI, above the baseline capabilities of AI, as those capabilities increased across time. It invited him to embrace the discomfort of admitting when SpeedLaunch no longer had alpha in places where it used to, simply because the AI baseline had grown. It invited him to shift away from ongoing reactivity towards ongoing creativity.
This led to the practice of creating Alpha Tickets before making non-trivial changes to the codebase. Especially if those changes were triggered by the X feed. This process allowed him to simultaneously unlock greater aligned velocity, without losing control.
Each Alpha Ticket asked him to answer five questions:
What customer problem is this idea intended to alleviate?
What deeper customer motivation makes this problem worth solving?
What value can coding agents already create out of the box aligned with this deeper motivation?
What alpha should SpeedLaunch take responsibility for, above what can be provided by that AI baseline?
How will this specific alpha be measured?
This shift of consciously defining alpha started turning things around. One week, he almost shipped a knee-jerk feature after seeing a competitor’s hype-ey launch thread. But the Alpha Ticket successfully made him kill it because he couldn’t consciously define SpeedLaunch’s alpha above what he thought LLMs already provided customers. Alpha Tickets successfully filtered out various misaligned features before they bloated the product, and forced each non-trivial change to justify whether/how it’d actually create alpha.
He shipped fewer PRs than he had during his frantic peak, but more than in the pre-AI world. Most importantly, they were consistently the right PRs. Churn slowed down and sales calls got easier because he could easily define SpeedLaunch’s alpha to customers over what a Claude Max subscription provided out of the box.
The next time you hear yourself say something like, “I need to ship something or I’ll get left behind”, I’d invite you to pause and write an Alpha Ticket first.
If this essay resonated with you, or if you have feedback, I’d love to hear from you at varun@doubleascent.com. Like I said, I’m absolutely fascinated with how solopreneurs and lifestyle businesses can survive AI, and eventually thrive. I’d love to help out however I can.
Acknowledgements - Thank you to my executive coach Brian Whetten for giving me the leadership skills to write this essay.