Intel · Heresies
Heresies.
Short-form opinion pieces challenging conventional investing wisdom. Written when the consensus has stopped fitting the regime — and when sticking to the textbook starts costing more than it saves.
-
Price Is the Only Truth — Closelook Heresy VII
Three regimes move prices: fundamentals, frenzy unwinds, deleveraging cascades. The fatal error is reading one while trading in another. Heresy VII.
-
The Selection Game — Closelook Heresy XIV
Hedge fund investors earned ~cash+40bp dollar-weighted over three decades. Advised accounts in Canada and Germany ran worse after fees. Picking the picker is the lottery one level up. Heresy XIV.
-
Diversified into the Losers — Closelook Heresy XIII
Bessembinder: 2.4% of firms created all $75.7T of global equity wealth. Every classic diversification fix sells that right tail. Concentration isn't the bug — it's the receipt. Heresy XIII.
-
The Haystack and the Needle — Closelook Heresy XII
Bessembinder proved 4% of stocks make all the market's wealth, then that only 5.5% of funds beat the S&P 500 over 30 years — $1.02T destroyed. Why thematic ETFs are the losing end of that math, by construction.
-
The Model Is Not the Moat — Closelook Heresy XI
Two of the biggest compute owners just became landlords. The moat was never the model — it is the harness bred around it and the loop that improves it.
-
The Numerator Regime — Closelook Heresy X
The buyback era inflated financial assets; the AI-capex era inflates bottlenecks — and pays whoever controls the constraint, not the spender.
-
The Constraint Relay — Closelook Heresy IX
Most AI picks-and-shovels are melting ice cubes; a few refreeze on NVIDIA's clock every generation. The Constraint Relay tells the hold from the rent.
-
Memory Is the New Compute — Closelook Heresy VIII
The market still prices memory through the old DRAM cycle. But HBM4 is an AI architecture layer, not faster PC RAM — and as HBM4E and HBM5 add accelerator-specific logic dies, memory crosses from component to co-processor. By Feynman it may rival the GPU. SK hynix is the cleanest expression of the shift.
-
Nvidia: Great Company, Bad Hold — Closelook Heresy VI
Nvidia's stock leadership is episodic, not permanent — front-loaded into each architecture, then restarted by the next. Hopper did it, Blackwell did it, and a successful Rubin — the likely case on product grounds — may open a fresh leadership leg. Own the cycle, not the marriage.
-
Build a Barbell — Closelook Heresy III
Markowitz, Sharpe and Value Investing are working on the wrong object. Bessembinder broke the orthodox toolkit. The barbell — QQQ+VEU long pole, Pattern 03 short pole, zero in the middle — is what survives. Live XIRR +109%.
-
Don't Buy Global Stocks — Closelook Heresy II
Bessembinder 2023: 64,738 stocks, 43 markets, 31 years. 1.41% of non-US firms create all $30.7T of non-US net wealth. The median non-US stock returned −7.6%. Build a barbell.
-
Don't Buy U.S. Stocks — Closelook Heresy
Four out of seven common stocks underperform Treasury bills. The best 4% explain the entire net gain since 1926. Why ‘stocks go up’ is true for the index, not the ticker.
-
Tech-Free — A Flag of Inconvenience
Two decades of value-investing underperformance is no longer noise — it is a signal about the methodology. Why 'I don't do tech' was never humility.
-
The First 50% Trap — Closelook Heresy
Why classical valuation metrics — and classical portfolio rules — break the moment a secular bull is real. Druckenmiller, memory chips, and the barbell.
14 heresies on file.
FAQ · from the current data · as of 2026-08-03
Heresies — Q&A
What is Heresies?
Contrarian, voice-driven investment opinion essays — the house positions that run against consensus, argued in the open. Hub: /heresies/.
How often does Heresies publish, and how many editions are there?
Irregular — when a heresy is ready; 14 editions are archived on this page, the latest "Price Is the Only Truth — Closelook Heresy VII" dated Aug 3, 2026.
Is this investment advice?
No. Closelook publishes an investment research diary — human-reviewed, AI-supported — not investment advice. Past performance is no indication of future performance.