Guides
Explainers with data, and claims graded against outcomes.
Claim checks
Markets Claim check Unsupported
Does "inverse Cramer" actually beat the market? Checked
We graded 41 on-air picks against six-month returns. The meme says fade him; the data says he tracks the index, noisily.
Investing Claim check Mostly true
"Your money doubles every 7 years", checked
The rule of 72 at historical index returns, tested decade by decade since 1957. True on average; wildly untrue in any given decade.
Investing Claim check Mostly true
"Credit card churning nets $2,000 a year", checked against real point values
Achievable for a disciplined household that redeems well. The average redemption falls well short of it.
Investing Claim check False
The $997 trading course that "pays for itself", checked
We ran the course's own published trade alerts for 90 days at stated position sizes. It did not pay for itself. The math inside.
Investing Claim check Unsupported
"The 60/40 portfolio is dead", checked against the last five years
One bad joint year for stocks and bonds does not settle a decades-long argument. We ran the longer window.
Investing Claim check Mostly true
"You need $1 million to retire", checked against real withdrawal math
True at the median assumed spend. Off by a wide margin once you plug in an actual budget.
Markets Claim check Unsupported
"The VIX always spikes before a crash", checked
It spikes during plenty of crashes. Whether it reliably spikes before one, in time to act on, is a different claim.
Markets Claim check False
Is a parlay ever +EV if every leg is individually +EV? The math, worked through
Combining favorable legs does not combine their edges the way the pitch implies. We show the arithmetic.
Markets Claim check Unsupported
"Sell in May and go away", checked against 40 years of returns
The seasonal pattern is real in the average. It is also smaller than transaction costs in most individual years.
Productivity Claim check Mostly true
"You should always negotiate your rate up 20%", checked against real outcomes
It worked more often than not in the cases we could verify. It also cost a few of them the offer entirely.
Productivity Claim check False
"Multitasking doesn't really cost you time", checked against the switching-cost research
The self-reported version and the measured version disagree by a wide margin. The measured version wins.
Productivity Claim check False
"The 4-hour workweek is achievable for most people", checked
Achievable for a narrow set of business models. We checked what it actually takes to get there first.
AI & tech Claim check True
Does more compute reliably make a better model? The scaling claim, tested
We plotted three labs' published benchmark scores against disclosed training compute. The line holds better than the memes suggest.
AI & tech Claim check Unsupported
"AI replaces 40% of jobs by 2030", checked against the actual studies
The number gets quoted as consensus. The paper it traces back to says something narrower, with wide error bars.
Automation Claim check True
"AI coding assistants cut build time in half", checked on our own repo
We timed the same class of task with and without an assistant across two sprints. The claim held up, roughly.
Automation Claim check False
"No-code tools replace a developer", checked against a real build
We rebuilt one working pipeline in a no-code tool. It got close, then hit a wall a developer would not have.
Explainers and briefs
Automation Explainer
Cron jobs, webhooks, and the difference that matters for a solo builder
When a schedule is enough and when you actually need to react to an event, with the cost of getting it wrong.
Automation Data brief
What a $0-budget pipeline actually costs once you count the free-tier limits
Every "free" tier we used, the ceiling before it charges, and where our pipeline actually sits against it.
Automation Analysis
What we learned running the same scraper for 90 days straight
Rate limits, silent schema changes, and the three failure modes that ate the most debugging time.
Markets Explainer
What actually moves a prediction market price
Order flow, resolution rules, and why a 62% contract is not a 62% probability. A tour of the mechanics with real settlement data.
Automation Guide
Automating a personal net worth tracker with a free spreadsheet API
No paid tools: a spreadsheet, a scheduled script, and a chart that updates itself every morning.
AI & tech Analysis
Fifty forecasts are better than one: reading an ensemble
ECMWF publishes 50 parallel model runs twice a day. The spread between them is the forecast most weather apps throw away.
Investing Guide
A first-timer's guide to reading a fund prospectus
The five numbers worth finding on page one, and the boilerplate you can skip.
Investing Guide
Compound growth: the arithmetic your gut gets wrong
Doubling times, the rule of 72, and why the last decade of a 30-year plan does most of the work.
Automation Guide
A zero-cost pipeline for your own market notes
GitHub Actions, a SQLite file, and 40 lines of Python: the setup we use to snapshot markets twice a day.
Markets Explainer
Brier scores, explained with weather bets
How a single number tells you whether a forecaster is honest, and how to compute yours in a spreadsheet.
Productivity Data brief
We logged 90 days of focus blocks. Here is the histogram.
The 25-minute pomodoro lost. The 52-minute block won. Data from a self-tracking experiment, with the raw CSV.
Investing Explainer
Expense ratios, explained with a 30-year compounding table
The difference between a 0.03% fund and a 0.75% fund, in dollars, at three different starting balances.
Investing Data brief
We tracked 12 months of dollar-cost-average vs. lump-sum outcomes
Same starting stake, two entry strategies, one real market. The results, month by month.
Investing Analysis
What a "buy the dip" rule actually returns, backtested since 2000
A fixed rule, applied mechanically, against a simple buy-and-hold baseline over five separate drawdowns.
Investing Guide
Index funds vs. picking stocks: the math on defaulting to boring
Fees, dispersion, and the base rate for beating a total-market fund over a full decade.
Markets Data brief
Every earnings-day move we scored this quarter, by sector
Which sectors moved more than the market expected, and which ones the market had already priced in.
Markets Explainer
Market makers, spreads, and why the quoted price is never quite the trade price
The mechanics of the bid-ask spread, and what it costs you on a thin contract versus a liquid one.
Productivity Data brief
We logged 60 days of freelance invoices. Here is what actually got paid on time.
Net-30 terms and 30-day payments are not the same thing. The gap, measured across real invoices.
Markets Data brief
We logged every Fed day move in the S&P for two years. Here is the pattern.
The initial move and the move that actually stuck three sessions later are usually not the same move.
Productivity Guide
Building a simple emergency fund plan in one sitting
Three numbers, one target, one automatic transfer. The whole plan fits on an index card.
Markets Analysis
What a VIX spike above 20 has actually preceded, historically
Twenty years of volatility spikes, and what the index did in the 30, 60, and 90 days after each one.
Markets Guide
Reading an options chain without the jargon
Strike, expiry, open interest, and implied volatility, walked through on one real chain.
AI & tech Data brief
Every "AI bubble" call from the last 18 months, graded against what happened next
We collected the public bubble calls and checked them against the following two quarters of results.
Productivity Explainer
Time blocking vs. task lists: what the research actually says
Neither wins outright. The difference comes down to one variable most comparisons skip.
AI & tech Guide
A plain-language glossary for AI earnings calls
Tokens served, inference margin, compute committed: what management actually means, translated.
Productivity Data brief
We tracked 90 days of freelance hourly rates against hours billed
Higher rates did not mean fewer hours worked. The actual relationship, from real invoices.
AI & tech Analysis
Reading a datacenter buildout announcement like an earnings call
Capacity commitments, power contracts, and the gap between a press release and a signed lease.
Productivity Analysis
What happens to output when a meeting gets cut from 60 to 25 minutes
Same agenda, less time. What got dropped, what got tighter, and what nobody missed.
Productivity Guide
A weekly review template that takes eleven minutes
Five questions, one page, timed across eight weeks until it stopped growing past eleven minutes.
AI & tech Explainer
Tokens, context windows, and why "unlimited" never means unlimited
The plain-language version of a spec sheet, with the math on what a stated context window actually costs to fill.
Automation Explainer
What "serverless" actually means for a side project's bill
Cold starts, per-invocation pricing, and the traffic level where serverless stops being the cheap option.
AI & tech Data brief
We tracked 30 days of AI model release announcements. Here is the base rate.
How often "coming soon" becomes a shipped model within the stated window, by lab, over one month of headlines.
Automation Data brief
We automated part of our own claim-check pipeline. Here is the error rate.
Where the automation matched a human grader, and where it still needs a second look before publishing.
AI & tech Analysis
What chip export controls actually change, priced into the names that matter
A rule change moves headlines fast and revenue slow. We separate the two for the stocks most exposed.
Automation Guide
Turning a Google Form into a lightweight expense tracker
No app, no subscription: a form, a sheet, and three formulas that do the categorizing for you.
AI & tech Guide
How to read an AI lab benchmark chart without getting fooled
Cherry-picked baselines, undisclosed prompts, and the three questions to ask before a benchmark chart changes your mind.