Success Is a Clean Cut.
A decade of gradual reform deepened Poland's crisis. A complete break, done all at once, ended it faster than every gradual alternative.
Blog posts by Julien Reszka
A decade of gradual reform deepened Poland's crisis. A complete break, done all at once, ended it faster than every gradual alternative.
People rationalize a failing decision for years, sometimes until nothing is left to save. Research shows what actually breaks the cycle before that.
The real reason your estimate is too low is not wishful thinking or a lack of experience. It is what you never named out loud.
Americans have fewer close friends than in 1990, but the fix is not simply adding more. It is adding the right kind.
Denver International Airport finished construction on schedule, then sat empty for 16 months because one system was never tested in pieces first.
Companies almost never get blindsided by disruption. Their own incentives stop them acting on threats they already see coming.
Hard work is grinding and only partly reliable. Luck is random by definition. Research on who you know is far more specific than most networking advice.
Americans spend more time in the kitchen than 20 years ago. Over the same 20 years, ultra processed food ate a bigger share of what actually lands on the plate.
Pension contributions are usually measured against gross pay, rent against take home pay. Put both on the same basis and together they eat past half of it.
Global life expectancy keeps rising. Healthy life expectancy is not keeping pace, and the gap between the two has widened in 203 of 204 countries since 1990.
Pay-as-you-go pensions need a growing base of future workers to fund a shrinking one of retirees. Not having kids is free, and someone else pays twice.
Italy's 1990s pension cuts raised the birth rate among the workers who lost benefits. Cutting pensions further is exactly the reform democracies survive worst.
Elections measure who performs best under a spotlight, not who can allocate resources or absorb hard tradeoffs. Confusing the two breaks governance.
Memory is reconstructive. When someone recalls events differently, it feels like gaslighting. It usually isn't. Document first, then negotiate.
Real pressure has a mechanism. Nonsense pressure is just social discomfort in disguise. Here is how to tell them apart and why it matters.
Power does not earn trust, it creates a debt of proof. The COVID lab leak consensus shows what happens when officials are not held to that standard.
Paste a raw file into an agent prompt and half your context window is gone before a fix is written. One CSV line per issue beats 500 raw lines every time.
Google Reader died in 2013 and everyone called it. But RSS never stopped powering podcasting, and now AI agents need exactly what RSS does.
The study everyone cites to dismiss interviews has been corrected. The update is not that interviews work. It is that structure is what works.
Most founders pick a market, then hope for the right exit. That is backwards. The exit you want determines which market you enter and how to compete.
Incumbents are required by law to list everything that could kill them. Read those lists. Then build one of those things.
Most founders ask 'is it worth building?' too late and with no framework. Three thresholds (comprehension, viability, growth) tell you when to stop.
Stop/start/continue gives you a board of post-its and no decisions. Three strategic questions written before the meeting changes that.
42% of dev time goes to technical debt. Most refactoring doesn't reduce that. It just moves it around while delaying real work.
Strong retention but zero referrals is a diagnosis, not a mystery. Your customers have the problem. They just won't admit it in public.
Owning a Rolex says more about you than it tells you the time. Products that make buyers look good get shared. That's not vanity, it's distribution.
Dropbox spent $388 acquiring customers worth $99, until referrals changed everything. Word-of-mouth doesn't just happen; it has to be designed.
Seven exposures before a message sticks. Most marketing wastes all seven by saying too much. The fix is compression, not more content.
83% of people act on recommendations from those they trust. The catch: they only recommend things they'd be comfortable defending.
Netflix's algorithm drives 80% of what users watch. The lesson: a product that feels like it knows you is one people don't leave.
Most software features are never used. A simple severity filter (does anything depend on it? is there a workaround?) cuts them before they ship.
Fear shrinks the prefrontal cortex and degrades the exact thinking you rely on for work. X is engineered to keep you in that state. The Stoics had the fix.
The Dutch sleep 8 hours and work 32. Japan sleeps 7.5 and works 50. That gap isn't discipline; it's legislation.
Automation reduces stress at the task level and raises it at the existential level. The worry doesn't disappear. It just moves to harder questions.
Automation doesn't make operators more careful. It makes them forget how to be. The more reliable the system, the less ready the human.
In theory, automation frees us for meaningful work. In practice, it filters out the easy tasks and leaves employees with a permanent queue of hard ones.
People cancel, get sick, and have their own agenda. Automated systems don't. The real case for automation is fewer humans in your critical path.
The future of work is not longer hours. It is learning when to step back, and designing systems so that when humans interfere, the damage is limited.
Machines destroying jobs is easy to debunk. Japan is among the most roboticised major economies and has one of the lowest unemployment rates in the G10.
After 25 years, Google still can't judge quality, so it defers to the crowd and makes you beg for backlinks instead of just writing something good.
The algorithm is a slot machine. Random rewards are more addictive than consistent ones. That's not a metaphor, it's behavioural science, and I was hooked.
Airbnb started with three air mattresses already in a closet. The business was not built from nothing. It was built from what was sitting unused.
Apollo had one customer and one deadline. Every planning question NASA actually faced was about which tasks blocked which, and what could run in parallel.
Frank Knight drew the line in 1921: risk can be measured, uncertainty can only be judged. Demand for anything new sits on the wrong side of that line.
Jeff Bezos rarely answered 'what will change?' His more useful obsession was the opposite: what won't? Amazon was built on three bets that held for decades.
Elon Musk's first rule: make requirements less dumb. Second: delete the step. Optimization is third, and most people start there.
Dating apps are built like casinos. Here is what worked instead: a foreign language, a dance floor, and showing up somewhere new.
WebMCP brings agentic AI into the browser the same way mobile-first rewrote the web in 2012. The gap it opens is wide and most people are not looking at it yet.
Strict liability with mandatory insurance self-enforces, prices harm, resists capture, is free to the state, and funds the science. Most policies fail all five.
Cutting spending when you already spend 57% of GDP gives you six times less growth than the same cut at 20%. France is on the wrong part of the curve.