BERKSHIRE HATHAWAY — THE SHORT VERSION

August 22, 2026  •  No position, no compensation, no MNPI 

This is an example of the Free Read document that I offer to organizations.

Here is the whole argument in two pages. The long version runs to two hundred and eight points across ten headings (available by request), and if you only read this page you will not miss the thesis. The company does not have a brand problem. Its brand is probably the strongest in American finance. It has a retrievability problem and a reachability problem, and both are cheap to fix.

WHAT ACTUALLY BROKE

1. The marketing department was one man, and he retired on January 1, 2026. One letter a year, one crowd every May, no budget. Nobody replaced the function because nobody ever wrote it down as a function.

2. There is no front door. No contact form, no stated acquisition criteria, no named human being, no origination tracking. Sellers assume there is a process they are not part of. There is no process.

3. Nobody can quote you. Sixty years of the best business writing in America sits in PDFs with no HTML, no anchors, no structured data, and no sitemap. When somebody asks a chatbot about you at eleven at night, the answer comes from Wikipedia. A lot of those answers still name the wrong CEO.

4. 390,000 employees, no careers page. Every subsidiary pays full price for talent and none of them borrow the parent’s name. At the operating company that most needs machine learning engineers, the employer rating is 2.7. That is a wage premium nobody has ever seen invoiced.

WHAT I WOULD DO THIS QUARTER

1. Turn on the instruments. Search Console, Bing Webmaster Tools, an XML sitemap, meta descriptions. Assign one named person to read them monthly. Two days of work, and it has never been done.

2. Fix the plumbing, not the look. Navigation, a footer, breadcrumbs, permanent URLs, mobile reflow, and a link palette that does not invert thirty years of convention. Every item on that list is invisible on a desktop screen. The page should look exactly the same and behave completely differently.

3. Publish the letters as HTML. Every one since 1977, with an anchor on every section, alongside the PDFs. Three weeks. Then the meeting transcript within seventy-two hours, every year, forever.

4. Write the facts page. Founding, headquarters, segments, employee count, leadership, tickers. Wikipedia should not be the authoritative source of facts about a trillion-dollar company, and right now it is.

5. Build the seller page. Criteria, size ranges, industries you want and industries you do not, a named person with a real email address, and the permanence promise in writing. Then staff the inbox. An unanswered inquiry is worse than no page at all.

6. Market to the advisors. Estate planners, CPAs, and regional M&A attorneys learn about a succession event a year or two before a banker does. A few hundred of them control the deal flow you say you want. Nobody markets to them. One forwardable page and a plain quarterly email covers it.

7. Interview ten lost sellers. Owners who sold to somebody else. Outside interviewer, four questions, no defending the loss. Price is the polite answer, not the true one. Costs a few thousand dollars and will be the most valuable page in next year’s plan.

8. Tell the energy story. Regulated rate base is the safest way anyone owns the AI buildout, and you are barely mentioning it. Publish the large-load tariff, the queue position, and an honest energization date. Procurement teams can plan around a published constraint. They cannot plan around enthusiasm.

WHAT I WOULD NOT DO

1. Do not redesign the website. The ugly page is the ad. It generates earned media, gets taught in design courses, and says something no campaign could buy. Change what is under it and leave the surface alone.

2. No parent-level brand advertising. It would cost a fortune, contradict the frugality that is the brand, and solve a problem you do not have.

3. Do not brand the subsidiaries. Every operator asked said the same thing: putting the holding company on the window subtracts trust. They work because they are themselves.

4. Do not try to out-advertise Progressive. They win on segmentation and rate refresh speed and they say so publicly. Advertising cannot outrun a pricing disadvantage. Compete on transparency instead, which is empty ground.

5. Do not buy an AI company. No edge in evaluating one, brutal multiples, and you already own the better exposure through the utilities.

WHO TO HIRE (if not already there under different titles)

1. A head of AI search visibility, this quarter. One person, one quarter to first results, highest return per dollar in the entire plan.

2. A chief communications officer at the parent by Q4 2026. A writer and an operator, not a campaign person.

3. A seller-facing marketing lead in the first half of 2027, owning the funnel, the advisor channel, and the origination system. That is three hires and a small team. Do not build a department; it would be resisted, and correctly.

THE QUESTIONS I CANNOT ANSWER FROM OUTSIDE

1. Does origination still work without one man’s telephone? This is the whole ballgame. You can answer it internally in a quarter by counting inbound approaches by source.

2. What is the one sentence that survives the founder? Two separate hundred-person panels could not produce it. A journalist is going to write it, and they will pick the version that is easiest to write.

3. What does the talent tax cost? Nobody has ever put a dollar figure on the weak employer ratings, and it is knowable.

4. What is the tell for drift? The real succession risk is not a bad decision. It is the slow replacement of judgment by procedure, and it shows up in the org chart long before it shows up in results.

Sequence it: retrievable, then reachable, then hireable. Two senior hires and eighteen months, for a rounding error against one quarter of buybacks. The aesthetic stays. The plumbing changes. The reputation stops depending on a person.

Built on public information only. Not investment, legal, or tax advice.