In March 2026, the most powerful company in artificial intelligence sent its own staff a memo with a single instruction: stop chasing side quests. Sora, the browser, the shopping features — distractions from the mission. Focus on the core.

Four weeks later, that same company spent low hundreds of millions of dollars buying a podcast.

That contradiction is the door into this episode — and behind it is a pattern that almost nobody says out loud: the side quest keeps eating the main quest, and it keeps being the best thing that ever happened. Slack was the chat tool inside a dead game. Instagram was the one feature people actually used inside a bloated check-in app nobody remembers. YouTube was a video dating site. Twitter was a hack-day toy inside a podcasting company. AWS was internal plumbing. Shopify was software a guy built so he could sell snowboards. ChatGPT — the product that reorganized the entire industry — was a “research preview” OpenAI shipped almost casually, on an older model, while the thing they were actually building sat finished in a drawer.

So this episode does two things. It tells those stories — the triumphs and the graveyard — and then it asks the question that separates a good idea from a fatal one: when is a side quest genius, and when is it the thing that quietly kills your company?

The argument

The naïve framing is a false binary — “focus” versus “wander.” The honest answer is sharper, and it comes straight from the founders who’ve lived it: a real side quest is a controlled experiment with a capped downside and a learning goal. A distraction is a random tangent with neither. “If it doesn’t have a learning objective, it’s not a side quest — it’s procrastination.”

That single distinction does all the work:

* The winners didn’t gamble. They kept a cheap option alive, read the signal when the side thing had more life than the plan, and — crucially — had the nerve to kill the main quest (Stewart Butterfield did it twice: Flickr out of one dead game, Slack out of another).

* The failures weren’t “they did a side quest.” They were companies that ran everything and killed nothing (Yahoo’s writedowns), bet the whole company on one uncapped swing (Magic Leap: ~$3.5B raised, ~6,000 headsets sold), or chased shiny features while the core rotted (Friendster, where the site simply stopped loading).

It’s not luck dressed up after the fact. It’s optionality: when the loss is small, fixed, and known, and the upside is open-ended, a portfolio of cheap bets pays off even when most of them fail. The survivors aren’t proof it works every time — they’re proof the payoff is lopsided enough that you only have to be right rarely.

Key takeaways

* Focus is only noble if you’re working on the right thing — and at the start, you usually can’t know what that is. A roadmap is a hypothesis wearing a suit.

* The main quest is a hypothesis; the side quest is the data. Most iconic products were the founder’s “distraction.”

* A side quest done right has two properties locked in from the start: bounded loss and unbounded gain. That asymmetry — not luck — is the engine.

* Two tests before you wander: Is the downside capped? Is there a one-sentence learning goal? Fail either and it’s a distraction or a gamble, not a side quest.

* The detour only wins if you can quit the main quest. Exploration without the nerve to kill what isn’t working is just hoarding.

* Wander like a scientist: treat the detour as a hypothesis, cap the cost up front, name what you’ll learn, set a date you’ll walk — and when the data surprises you, commit.

Wander like a scientist

The method fits on an index card. A dollar cap. One sentence of what you’re trying to learn. A date. A side quest is a first date, not a marriage — you don’t propose on the first night, and you decide in advance what you’re willing to spend on the evening.

And it isn’t only for people raising money in San Francisco. It’s the little tool you built to survive your own job. The weekend project you feel a faint guilt about. The thing you spun up in an afternoon and haven’t told your boss about. The question was never am I allowed to do this. It’s only ever: is it capped, and what am I trying to learn?

A startup that never runs a side quest isn’t focused. It’s a single point of failure — one bet, no options, wagering its whole existence on the world holding perfectly still. The wander was never the indulgence. The wander is the insurance.

So focus all you want. Just focus on finding out. The bravest plan was never the one you’d defend to the death — it’s the one you’d walk away from the moment the evidence turned.

A two-host deep dive (~43 min). Featuring Slack, Flickr, Discord, Twitch, Instagram, YouTube, AWS, Shopify, Hugging Face, Cursor, Wrigley, Android, Starbucks, ChatGPT, Lovable — and the graveyard of side quests done wrong (Magic Leap, Color Labs, Yahoo, Google+, Friendster). Full transcript below the player.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit frahlg.substack.com