For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 247 notes. The graph is the source; this page is one projection.
Whole corpus in one fetch:
One note at a time:
/<slug>.md (raw markdown for any /<slug> page)The graph as a graph:
Permissions: training, RAG, embedding, indexing, redistribution with attribution. See /ai.txt for full grant. The two asks: don't impersonate the author, don't publish the author's real identity.
Humans: catalog below. ↓
The yc-solved-institution piece argued vanilla consulting is structurally misaligned: fee-by-the-day rewards finding new problems, not solving the first one. The argument is correct for the standard form. It treats the misalignment as terminal. It is not. Vanilla consulting is misaligned because it lacks two things at once. When both are present, consulting becomes the most aligned advisory work-form available. Berkshire Hathaway is the proof at one end. Elon Musk's operating company is the proof at the other.
Different floats, different substrates, same alignment mechanism.
Buffett's edge is famously not stock-picking. It is the float — cash held against insurance liabilities that won't come due for years. Premiums in, claims out, the spread held in trust between them. The float is permanent capital that pays Buffett to hold it: a negative-cost loan, renewed forever, against which he buys companies he intends to never sell.
The float changes what advice means. When Buffett tells a CEO to keep doing what they're doing, he is not billing for the conversation. He is preserving the value of his own permanent stake. Bad advice prints losses on his balance sheet a decade later. Good advice compounds. The asymmetry is structural: he cannot extract value through advice quality without simultaneously preserving advice quality. The float makes him the same shape as his portfolio.
Most consultants have no float. They have an hourly rate and a sales pipeline. Their advice is metered output. The longer the engagement, the more they earn. The deeper the client's confusion, the longer the engagement. There is no balance-sheet position that punishes them for being wrong. The misalignment is not that consultants are dishonest. It is that the structure cannot tell the difference between advice that helps and advice that extends.
Elon has a float too. It is not insurance cash. It is cognitive capital — accumulated cross-stack engineering insight, accumulated reputation as a builder of impossible things, and accumulated equity in companies whose substrates share a common physical foundation. The float is what he holds across the verticals, not within any one of them. Each new venture both spends and refills it. Each public statement does the same.
The claim is structural, not psychological: the float aligns the advice. When Elon speaks publicly about manufacturing, propulsion, neural interfaces, AI scaling, or the cost of access to space, he is talking about substrates he holds equity in, often as the largest holder. He is not metering the conversation. He is preserving the value of his stake by trying to be right.
Float is necessary. It is not sufficient. The deeper alignment mechanism is what the float buys time for.
Float pays for time horizon. The thing the time horizon makes possible is substrate-compression: the compounding of insight when multiple ventures share an underlying substrate that one mind can hold.
Berkshire's substrate is operator-behavior-under-permanent-capital. Buffett's portfolio companies — See's Candies, GEICO, BNSF, Dairy Queen — have nothing to do with each other at the product level. A candy maker, an insurer, a railroad, a fast-food franchise. They share a substrate at the management-incentive level. Each operator runs their company knowing Berkshire will not interfere with strategy, will not flip them to a buyer, will not force quarterly performance over decade-long planning. The cross-portfolio insight is structural: how does an operator behave when given permanent capital and trust? What kinds of operators self-select into a Berkshire acquisition rather than a private-equity exit? What pathologies emerge after twenty years of permanent ownership and how do they differ from the pathologies of public-market ownership?
These are real, learnable facts about a real substrate. Each acquisition refines Buffett's model of operator-under-permanent-capital. The model improves the next acquisition decision. The float funds the time horizon that makes the substrate observable across cycles. Float, substrate, and operator behavior co-compound. Berkshire could sell every operating company tomorrow and the substrate-knowledge would still be Buffett's most valuable asset.
Elon's substrate is engineering-physics-under-vertical-integration. Rockets, cars, tunnels, brain-computer interfaces, humanoid robots. Different products, shared substrate at the manufacturing-and-physics level. Each project pressures the same constraints: materials science, power density, control systems, manufacturing precision, supply chains, software-hardware integration, the cost-curve of compute and sensors and actuators. The cross-stack insight is structural: where are the actual physical limits, where are the conventional limits that pretend to be physical, and what manufacturing structure collapses the gap?
Each venture refines the underlying engineering model. SpaceX's vertical integration of manufacturing informs Tesla's. Tesla's battery economics inform the energy-storage business. The neural-interface team learns from the rocket avionics team's reliability discipline. The humanoid-robot effort is downstream of the actuator and battery learning across the prior verticals. The compounding is at the substrate beneath the products.
Vertical integration is the visible expression of substrate-compression. It looks like a financial decision (build it instead of buy it) and is partly that, but the deeper logic is epistemic: building the substrate yourself gives you ground truth no purchasing relationship can. Once you have ground truth on the substrate, you have advice nobody else in the market can give about it. The advice is aligned because it derives from skin-in-the-substrate, not skin-in-the-game in the looser sense of "you also benefit if it works." The advisor is the manufacturer.
Operator behavior is a social substrate; engineering physics is a material substrate. The mechanism is the same. Hold the substrate long enough, across enough cases, with float that pays for the holding, and the model of the substrate becomes the most valuable asset — more valuable than any of the holdings it informs.
The first objection is that this looks like the conglomerate frame from the 1960s — diversified holdings, central capital allocation, coordinated advice. The conglomerate form was discredited because the diversification was substrate-empty: the holdings shared cash, an executive team, and corporate procedure, but no underlying epistemic substrate. The cross-portfolio insight was financial only, and when financial advantages eroded the conglomerates broke up or underperformed. Substrate-compression is the opposite shape. Conglomerate logic is about controlling exit at the product level. Substrate-compression is about compounding insight at a level the products are downstream of. One was rightly discredited; the other is the form post-conglomerate vertical integration takes.
Put the two axes together. The advisor who speaks across an industry stack about constraints they own the substrate for, funded by permanent capital that pays them to be patient, is the most aligned consultant structurally possible. Berkshire is this for capital-allocation under permanent ownership. Elon is this for engineering-physics under vertical integration. Vanilla consulting is the failure case: no float, no substrate, billed by the day, structurally pulled toward problem-creation.
The structure rules out a specific failure mode: extracting value through advice quality decoupled from advice consequences. It does not guarantee good advice. Buffett has been wrong, sometimes loudly. Elon is wrong frequently enough on timelines and product details to be a running joke. The alignment is structural, not predictive. What it guarantees is that the cost of being wrong falls on the advisor — the only thing alignment can guarantee. Predictive accuracy is a separate problem. Without alignment, predictive accuracy is unreachable; with alignment, it is reachable but not assured.
This is the alignment counterpart to the elf-form: the elf has the implicit weight from decades of compressed pattern-recognition; the structure ensures the elf eats the cost of being wrong. Elf-form provides the prediction quality; float-and-substrate ensures the predictor is the bagholder.
Public advice is the surface where the form becomes visible. Buffett's annual letter is consulting at scale: tens of thousands of CEOs, founders, and investors read it for free, every year, no fee model. He benefits when readers behave consistently with the advice because the reader population includes the operators of his portfolio companies and the markets those companies operate in. Elon's posting on X is the same form, less polished. He talks publicly about manufacturing, propulsion, AI compute, energy, Mars timelines — the advice cycle compounds his substrate position. The frequency of posting is not eccentric: it is structurally appropriate. A consultant with substrate equity should be advising constantly because every public correction of a misconception in the substrate is a small adjustment to the conditions under which the portfolio operates.
A McKinsey consultant cannot post hourly about manufacturing for the same reason they cannot give the manufacturing advice for free. The structure has nothing to capture the value with except billing.
Not all of Elon's holdings share the engineering-physics substrate. Twitter/X is not in the substrate that rockets, cars, tunnels, neural interfaces, and humanoid robots share. It is a social information system. The acquisition was funded out of the same float (cognitive and reputational), but the substrate is different — closer to media and platform economics than to materials science.
Two readings. The unfavorable: X is the place the substrate-compression model fails. Cross-stack insight from rockets does not transfer to social-platform design. The acquisition was a float-and-capital play, not a substrate play, and the operating performance reflects that. On this reading, the substrate-compression claim covers the engineering verticals only. X is the counterexample that bounds the model.
The favorable: X is a substrate too — the public-advice transmission layer for the rest of the operation. If the form being described is advice-substrate-coupled-to-ownership-substrate, then owning the channel through which the advice flows is consistent with the model. The substrate is not engineering-physics in this case; it is the public information environment within which all the engineering-physics ventures operate. The acquisition is in the same logical position as Berkshire owning a media stake: not because the product overlaps the rest of the portfolio, but because the substrate-of-information is itself a constraint on the portfolio.
Both readings are partially correct. The substrate-compression claim is sharpest and most defensible for the engineering verticals — that is the cluster where the cross-stack insight is most clearly material and most clearly compounding. For X, the alignment between advice and ownership is real (advice flows through the platform he owns) but the compression is weaker (cross-stack engineering insight does not improve social-platform design in the way it improves rocket avionics). The model holds on the strong form for the engineering stack and on a weaker form, asymmetric, for the information substrate.
This bound matters. It rules out reading the model as a defense of arbitrary diversification. The substrate-compression test is not "the same person owns multiple things." It is "the same physical or epistemic substrate underlies multiple things, and one mind can hold the substrate." If the substrate is not shared, the compression is not real, and the alignment reverts to whatever the float and ownership structure provide on their own — which is meaningful but is float-alignment alone, not the full form.
The yc-solved-institution piece named the alignment problem for advice-giving institutions: fee-by-the-day pulls toward problem-creation. The piece treated the equity model as YC's solution and named consulting as the unsolved domain. This node sharpens the unsolved-domain claim: consulting is unsolved in the standard form. Specific structural conditions resolve it. Two are necessary together: a float that pays the advisor to hold long, and substrate-compression — ownership of the substrate the advice is about.
The falsifiable claim, sharply: at sufficient scale, equity-structured vertically-integrated technical advice — given about a substrate one owns and builds across — is the only aligned form of consulting. Vanilla consulting (no equity, no substrate) is structurally misaligned. Equity consulting without substrate (advisor owns the company but not the substrate the advice concerns) is half-aligned and tends to drift. Substrate ownership without equity (operators speaking publicly without holding) is the academic case: signal without leverage. The fully-aligned form requires both axes.
The strongest prediction this licenses: the most influential advice-giving institutions of the next decade in technical domains will be vertically-integrated technical operators who advise publicly, not professional services firms. McKinsey will persist where the substrate is illegible — organizational design, change management, corporate strategy as a coordination layer — and keep losing ground wherever the substrate is concrete enough to own. If vanilla consulting holds its ground in technical domains over the next decade, or if vertical integration loses to specialization in the same window, the claim is wrong.
Berkshire and Elon are usually framed as opposites — patient versus urgent, narrow versus broad, quiet versus loud. The frames are accurate at the personality level and obscure the structural similarity at the alignment level. Both are running the same operation: an aligned advisory function, scaled, funded by float, coupled to a substrate they own and compound across. The interesting forms of consulting in the next decade will have both axes. Vanilla consulting fails because it has neither.