"Quit your job and start a company." "Drop out." "Bet on yourself."
I heard this for years. I believed it. I did it.
And now I'm telling you: I no longer think that advice is a safe default.
Not because risk-taking is wrong. Because AI changed some of the costs and uncertainties in my own work.
The Old Game
From 2009 to 2022, "take the leap" sounded persuasive to me.
Execution was the bottleneck. Building something was genuinely hard. You needed a team. You needed time. You needed money. The ability to actually ship - that was the filter.
Low rates and available capital helped make the upside look attractive. That did not establish decent odds for every founder or make equity a safe lottery.
And the leap itself was the hard part. Most people wouldn't do it. If you could just start, you were already ahead.
So I accepted the advice: Quit. Jump. Figure it out.
I no longer want to give that advice as a default.
What Changed
Some implementation became much cheaper for me. Execution is not free: review, security, support, compute and time still have costs.
Some prototypes that once felt like team projects became things I could attempt over a weekend. That is not a measured equivalence in scope, quality or maintenance, and it does not show everyone can do the same.
The thing that used to separate you from everyone else - your ability to build and ship - is table stakes now.
This changes everything.
Some ideas can be tested while keeping a job and limiting spending. Others require capital, a team or dedicated time. Employment obligations, health and household commitments constrain what is practical.
Going all in can be necessary or unwise depending on the opportunity and constraints. The drama of announcing it is not evidence either way.
I worry about founders spending months on a wrapper whose advantage a later model release could erode. I cannot know that each product will be worthless in twelve months.
Fast, careful delivery can still find demand. My concern is spending that effort on an advantage I have not tested.
Judgment became a bigger concern for me, while execution and delivery still mattered.
When more people can build prototypes, choosing a useful problem matters more to me. Delivery quality and earning demand still matter too.
I do not have a measurement dividing success into 90% execution or 90% judgment. The change in emphasis is my interpretation.
The risk used to be not starting. Now the risk is spending years on the wrong thing.
What High-Agency Means Now
If you're high-agency, this is one of the best times in history to start something.
But high-agency in 2026 means something different than it did in 2015.
It means you can kill your own babies.
I decided to sunset our old SaaS direction in December 2025 and close the old consulting direction. That was a decision about the work I wanted to pursue, not proof every service had already ceased or would inevitably fail.
Editorial update, October 11, 2026: The later author-confirmed chronology is a December 2025 sunset decision, a June 2026 official pivot and a two-year customer transition, with the app still online. This January essay preserves the early change of mind. The renewal decision and the July building postmortem give the dated follow-through; they are not a guarantee that a reader can reproduce my result.
A few years ago, people who love me would have told me to get a job. This time, no one said that. They know I'll figure it out.
But here is the thing: I chose that transition. No one made me do it. I thought the direction no longer fit, even though customers were still being served.
That is the judgment I am trying to develop: being willing to reconsider something that works when the risks change, without pretending I know its value eighteen months ahead.
Tolerance for uncertainty matters. A new model can threaten an advantage, but it does not automatically destroy every product built before it.
Ready to pivot fast. Founders who fall in love with their idea are playing the old game. Wrap it up today, start something new tomorrow.
Protecting your agency. Think twice before hiring. Think twice before raising. Think twice before any decision that makes you answer to other people.
The founders who defined themselves by outworking everyone? Most exposed right now. The game doesn't reward hours anymore. It rewards decisions.
If You're Not There Yet
I was here too. Had to ask permission for everything. Afraid to take steps. Followed playbooks instead of designing my own.
So I am not going to tell you to "just quit and figure it out." It ignores too much about the person and the work.
You don't need to make a dramatic leap.
You don't need to announce anything. You don't need to quit - at least not yet.
I cannot infer how much of your workday AI can save from my own experience. The task, quality bar, review and constraints need to be checked.
Time, context switching and other obligations are real costs. A small experiment is only useful if those costs are manageable for you.
What you actually need is reps on judgment.
The gap between you and high-agency people isn't courage. It's pattern recognition. They've made enough decisions to know what matters and what doesn't.
When three things come at you at once, can you identify which one actually matters?
I spend minimum time on things that aren't important - use AI, get it done, move on.
For important things, I take days. Sometimes weeks. Train yourself to know the difference.
There's an old story about a kid who digs eighteen holes in the backyard to show his father how productive he is. The father asks: productive at what? You dug eighteen holes.
That's the risk now. Being productive at the wrong thing.
Start small. Make decisions. See what happens. Ship something. Learn whether your judgment was right. Repeat.
The reps compound into pattern recognition. Pattern recognition lets you make bigger bets later.
The Point
The old risk was not starting. The leap was hard. Execution was the filter. If you could just begin, you had a shot.
For me, the risk shifted toward building the wrong thing more efficiently. Implementation costs fell on some tasks, but they did not disappear. Judgment still needs evidence from delivery and demand.
"Quit your job and start a company" isn't wrong. It's just incomplete. It comes from a world where starting was the hard part.
Some starts are easier now. Knowing what is worth continuing remains hard.
I do not want to spend the next five years executing well on the wrong problem. This is a question for my own decisions, not a forecast of what most people will do.
That is the mistake I am trying to avoid.
Train your judgment. Make decisions. Kill things that aren't working.