The Abominable No-Man

Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own. We spent the last 6 months making sure Flash is great for real world use cases. It's available everywhere now!

Still wild to me that Warren Buffett’s Berkshire Hathaway no longer owns $V or $MA. Top 5 positions: 1. $AAPL ~$58B (22% of portfolio) 2. $AXP ~$46B (17% of portfolio) 3. $KO ~$30B (12% of portfolio) 4. $BAC ~$25B (10% of portfolio) 5. $CVX ~$18B (9% of portfolio) $BRK.B kept the classic consumer, financial and energy compounders but exited two of the cleanest global payment toll roads in the market.

To be clear, our sale of $GOOG was not a bet against the company. We are very bullish long term on Alphabet. But at current valuations and in light of our finite capital base, we used $GOOG as a source of funds for $MSFT.

BREAKING: Berkshire Hathaway discloses new purchases of Alphabet, $GOOGL, Macy's, $M, and Delta Airlines, $DAL. Berkshire Hathaway has also exited its entire position in Amazon, $AMZN, UnitedHealth, $UNH, and Dominos Pizza, $DPZ.

BREAKING: Berkshire Hathaway discloses new purchases of Alphabet, $GOOGL, Macy's, $M, and Delta Airlines, $DAL. Berkshire Hathaway has also exited its entire position in Amazon, $AMZN, UnitedHealth, $UNH, and Dominos Pizza, $DPZ.

While I’m here tonight, here’s my informal 13F for Q1: Q4 2025 -> Q1 2026 $MKL 23% -> 20% $GHC 21% -> 22% $GOOG 19% -> 20% $JPM 14% -> 5% $BRK.B 12% -> 11% $CVS 7% -> 6% $GOOGL 3% -> 3% $HHH 1% -> 9% $V 0% -> 4% $V Visa was my only new position, $HHH was a notable increase, and $JPM was a notable reduction.

Why $MKL looks cheap. Tom Gayner publishes his intrinsic value methodology right in the 10-K — Item 7, Capital Performance. Classic Buffett-Munger discipline: better to be roughly right than precisely wrong. The stickman: Start with the earnings side. Take adjusted operating income — $2.30B in 2025. Strip out dividends from the equity portfolio — $156M — and interest on cash — $228M. What's left is roughly adjusted earnings: $1.92B. He then takes a 3-year average — $1.60B — and applies a multiple. Then the balance sheet side. Add equity securities at fair value — $13.0B — and cash and short-term investments — $6.0B. Subtract debt, preferred, and noncontrolling interests — $4.8B. You get $14.2B. Divide by 12.59M shares. Tom shows three multiples — 8x, 12x, 16x: 8x → $2,143 12x → $2,651 16x → $3,158 Stock today: ~$1,779. Below his most conservative anchor. Where Tom is being conservative: Even Tom's own smoothed 3-year average earnings figure has compounded at 16.5% per year over the last five years — through COVID, through the 2023 reserve charge, through everything. Balance sheet value has compounded too — $9.0B in 2020 to $14.2B in 2025. That's a 9.5% CAGR, with faster growth over the last two years as the equity portfolio compounded and cash built. The math on multiples matters. A 12x multiple at a 12% required return only needs roughly 4% growth to be justified. Markel's actual growth has been multiples of that on both engines. Now invert it. At $1,779, the market is paying about 5x for the operating businesses after backing out the marked-to-market balance sheet. Five times. That is a very low look-through multiple for a business that just put up record earnings. There is one important caveat: Markel is not a simple industrial company with excess cash sitting on the balance sheet. Some of the investment portfolio supports the insurance business and its liabilities. So the balance-sheet value is not identical to excess cash at a software company. But that is also the point of the Markel model: insurance float, disciplined underwriting, public equities, and operating businesses all compound together. And it's not just the math. You get Tom Gayner running it. You get the Markel culture — nearly a hundred years of underwriting discipline and reserve conservatism. And you get optionality across three engines: insurance, public equities, and operating businesses, all feeding each other. Buffett taught us that this kind of optionality increases your chances of getting a date on Saturday night. Tom's putting his money where his mouth is — 223K shares bought at a $1,894 average in 2025. Bought another $134M in Q1 2026. The key underwriting question is whether recent operating performance is durable. The key valuation question is whether you are careful not to double count investment income that is already reflected in the investment portfolio. But even after making those adjustments conservatively, the setup looks unusually attractive. The 10-K shows you exactly how Gayner thinks. Do the math yourself. $MKL looks cheap. Not because the math is complicated — but because the market may be underpricing a boring, durable compounder hiding in plain sight. One specific item to flag from the Q1 2026 10-Q: Markel disclosed a collateral shortfall on a State National fronting relationship. Management has engaged outside actuaries and stated it is not material to capital position. Worth watching in subsequent filings, but not a thesis-breaker.