Actionable insights — 5 Stocks Superinvestors Are Buying

1. Clone deliberately: take the idea, re-derive the reason

1. Treat a superinvestor's disclosed position as a source of candidates, never as a conclusion. The filing tells you what was bought, not why, at what price, at what weight, or against what alternative.
2. Before acting, re-derive the case in your own terms: how does the business make money, and what specifically would have to be true for the price to be wrong?
3. Check the lag. Filings are backward-looking by a quarter; verify the price today is still near the price then.
4. Only copy where you can carry the position through a drawdown on your own reasoning — you will not receive the manager's next update.

• Cloning position sizes as well as names — a 2% position in a $10bn fund and a 20% position in your own account are different bets on the same idea.
• Managers whose mandate forces them to be invested; their "buy" may be a relative preference, not an endorsement.

2. Read the gradation of a sale, not just the direction

1. For any name being sold, record how much each holder sold: complete exit, partial trim, or a reduction that keeps the position meaningful.
2. Full exits by more than one independent manager in the same quarter suggest a thesis break; partial trims usually mean a valuation adjustment with the thesis intact.
3. Ask which of the two the price already reflects.
4. Look for the same split among buyers — a new core position is a stronger signal than an addition to an existing one.

• Sales driven by fund flows or position limits rather than by view; a full exit is only informative if the manager was free to stay.
• The valuation half of the case being quietly dropped: at 20x rather than 50-70x, the same AI risk supports a very different conclusion.

3. When two good investors take opposite sides, find the single fact they price differently

1. Identify the pair of opposite trades and write down the one variable both are forecasting.
2. State each side's implied answer in a sentence you could check in two years.
3. Decide whether you have any edge on that variable. If not, the honest position is neither name, or both at half size.
4. Record the disagreement with dates, so the outcome can be scored later against the reasoning rather than the noise.

• A third possibility neither side is trading: that both incumbents pay for the capacity and neither earns the return.
• Your own tendency to resolve the disagreement in favour of whichever manager you admire more.

4. Test whether your index is still diversification

1. Look through the index to the underlying economic exposure: what share of value sits in companies whose fortunes depend on the same thing?
2. Ask whether the top holdings are spending on the same project — shared capital expenditure is shared risk regardless of different sector labels.
3. Apply a market-level valuation gauge (total market capitalisation to GDP) so the concentration question and the price question are answered separately.
4. If the answer is "this is one bet," either accept it deliberately or replace it with exposure you chose.
5. Exit in stages if you act — you are correcting a composition drift, not calling a top.

• Replacing a concentrated index with an equal-weight or quality index that shares the same top names in smaller size — check the look-through, not the label.
• The Buffett Indicator's known weakness: US-listed revenue is increasingly foreign, so the ratio drifts upward structurally. Use it for direction, not as a threshold.

5. Check whether the attribution survives to the recommendation

1. In any "what the pros are buying" piece, mark each name with the investor cited for it.
2. Where the attribution disappears — a name presented with a business explainer and no filing behind it — reclassify it as the author's own idea.
3. That is not a reason to reject it; it is a reason to apply the author's own valuation standard to it rather than the borrowed authority of the headline.
4. Cross-check the unattributed names against the author's published list: overlap tells you which shortlist you are really reading.

• Rankings ("#1 to #5") that imply a measurement no methodology is given for.
• The useful residue: an unattributed name that is not already on the author's list — that is the one worth asking about.