What is the harder human-capital problem beneath token capital?
Turing Post · email · 2026-06-15 Reader 🖥️
AI Labor DisruptionCoordination CostEnterprise AI Adoption & ReadinessHierarchy & Org ChartsAgencySmall Teams & CompressionTalent DensityToken Capital
Beneath Nadella's visible 'token capital vs. human capital' story lies a harder problem — the human capital big firms manufacture best (senior, politically-fluent ladder-climbers) is exactly the kind that wastes token capital, while the frontier-moving talent that would use it well is what corporate structure selects against — so the scarcest capital in the AI age is neither human nor machine but an organizational structure porous enough to hold uncorporate talent and corporate-scale compute together without flattening either.
practical · Analytical newsletter op-ed — a reaction piece that builds a structural argument on top of Satya Nadella's post about token vs. human capital. — Scoped to the ~927-word opening editorial of Turing Post FOD#156; the rest of the newsletter (news roundup, papers, model releases) is excluded from this analysis, so word_count reflects the editorial, not the ~10k-word email. Practical because it lands on a prescription (build a porous structure), but it is carried by a diagnostic argument about how corporate structure selects talent.
Structure
| Part | Role |
|---|---|
| Hook and framing (barely any stars crossed from old software into AI; reading Nadella's 56-million-view post, which matters not for its surface but for its buried layers, 'seen better from afar') | Signals that the real argument lives beneath the visible thesis and sets up the excavation. |
| Definitions: token capital (a firm's owned AI capability — models, agents, traces, evals, workflow memory, loops) vs. human capital (its people's judgment, taste, relationships, context, pattern recognition), and Nadella's claim that the two compound | Establishes the shared frame the essay will complicate. |
| The turn — the Warrior becomes The Thinker: Microsoft is dominant (Azure, Office, GitHub, OpenAI, distribution) yet its 'core warriors' are 'too stiff for the shape of this moment' | Introduces the tension inside the dominant incumbent: seniority no longer converts to success. |
| Evidence — who built the best models: bold, uncorporate people in the US and China; Demis and Jeff Dean succeed 'in spite of' Google; they are researchers, obsessives, quants and builders who 'came in sideways' and whom the ladder would have 'sanded down' | Empirical grounding for the split between ladder-climbers and frontier-movers. |
| The buried layer — WHICH human capital: the politically-fluent veteran who can move a proposal through six committees 'optimizes the org chart, not the loss curve' and wastes token capital, while the irreverent frontier-mover is the kind the structure selects against and pushes out | States the core diagnostic claim and the implicit selection mechanism. |
| The driving question and the stakes: under what conditions can corporate America hold such people without domesticating or exhausting them (or does AI reward a band/lab shape)? plus scale (70M+ in large firms, ~54% of private employment) and the slow-senior / shallow-junior trap | Scales the diagnosis into an economy-wide problem and poses the essay's central question. |
| Thesis and closing aphorism: the work of the next decade is building a structure 'porous' enough to keep uncorporate people uncorporate while giving them corporate-scale compute, data, distribution and responsibility — 'the scarcest capital is neither the human nor the machine; it is the structure brave enough to hold both without flattening either' | Prescriptive resolution that names the true scarce resource. |
Key terms
- token capital Token Capital
Adopted from Nadella: the AI capability a firm builds and owns — its models, agents, traces, evals, workflow memory and internal loops — treated as a compounding proprietary asset (a stock), not merely a spend rate. - human capital
The judgment, taste, relationships, context and pattern recognition of a firm's people — but critically split by the essay into two kinds: ladder-climbing veterans and frontier-moving obsessives, only one of which uses token capital well. - porous structure
A coinage for an organizational form permeable enough to let 'uncorporate' people stay uncorporate while giving them corporate-scale compute, data, distribution and responsibility — neither crushing nor domesticating them. - ladder creatures / uncorporate people
Two opposed talent archetypes: 'ladder creatures' who mastered the internal language of quarterly planning and climbed the org chart, versus 'uncorporate' researchers, quants and builders who 'came in sideways' and are 'allergic to permission.' - optimize the org chart, not the loss curve
A compressed contrast between the veteran who improves his political position (the org chart) and the frontier-mover who improves the model (the loss curve) — the one-line statement of which human capital wastes token capital and which uses it.
Propositions
- explicitdevelopedNadella defines token capital (a firm's owned AI capability — models, agents, traces, evals, workflow memory, internal loops) and human capital (its people's judgment, taste, relationships, context, pattern recognition) and argues the two should compound, so humans grow more valuable, not less, as machines do.conf: highextends ←How the Escalating AI Wars Benefit You · p8 — Same Nadella token-capital argument (also from 'The Reverse Intelligence Paradox'); this piece adds the 'customers pay twice' framing — dollars plus proprietary data — on top of the token-capital/human-capital compounding claim the other extraction focuses on.
- explicitSeniority no longer translates cleanly into success inside a dominant corporate structure; incentives, committees, procurement logic, internal kingdoms and the need to sound correct before knowing what is true turn senior human capital into a burden that wastes token capital in the wrong way.conf: high
- explicitThe best current AI models were built by bold, 'uncorporate' people in the US and China who succeed in spite of their corporate employers, not because of them — researchers, obsessives, quants and builders who 'came in sideways' and whom the corporate ladder would have sanded down rather than made.conf: high
- explicitThe human capital a large firm manufactures best — the senior, reliable, politically-fluent veteran who can move a proposal through six committees — is almost exactly the kind that wastes token capital, because that person optimizes the org chart, not the loss curve.conf: high
- explicitThe talent that actually moves the frontier — irreverent, allergic to permission, willing to be loudly wrong in public — is precisely the kind a rigid corporate structure like Microsoft's selects against and eventually pushes out.conf: highsupportsSkill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI · p18 — Karpathy: being inside a frontier lab compromises your independence — there are things you cannot say. This essay generalizes and supports that — corporate structure selects against and 'domesticates' precisely the irreverent, permission-averse talent that moves the frontier.
- implicitcorroboratedCorporate structure therefore functions as a selection filter — for org-chart optimizers and against loss-curve optimizers — so senior resistance to the AI transition is a structural product of how the firm recruits and promotes, not merely individual psychology or politics.conf: highrefinesThe Agentic P&L: Beyond the Empire of Headcount · p8 — That extraction attributes senior resistance to the AI transition 'primarily to psychological/political reasons, not economic ones.' This essay relocates the cause: the political fluency itself is a structural product the firm manufactures and promotes, so resistance is a selection effect of org design, not just individual psychology.supports ←What 60 Years of Data Reveals About How Men and Women Experience Leadership · p8 — Same structural-not-individual move applied to a different domain: just as senior resistance to the AI transition is framed as 'a structural product of how the firm recruits and promotes, not merely individual psychology,' the leadership-pipeline gap is shown to be a structural product of staff-vs-P&L sorting and sponsorship/coaching asymmetry, coexisting with genuinely improved individual attitudes — corporate structure, not attitude, is the selection filter.
- explicitLarge firms are caught in a two-sided trap: senior human capital is often too slow while junior human capital is not yet deep enough, leaving the firm stuck between slow knowledge and fast ignorance.conf: highextends11) AI-spesial med Vilde Opsal og Anders Sveen · p13 — Heis #13 complicates the seniors-always-win narrative (many seniors fail to retool while juniors 'run past us'). This essay extends it into a two-sided organizational trap — senior human capital too slow, junior human capital not yet deep enough — leaving the firm stuck between slow knowledge and fast ignorance.
- explicitThe firms with the most human capital by the old scoreboard — the most veterans, deepest benches and proudest org charts — may be the worst positioned, because every one of those assets is also a reason not to move, in an economy where 500+-employee firms hold roughly 54% of US private-sector employment.conf: highextendsHard truths about building in the AI era | Keith Rabois (Khosla Ventures) · p2 — Rabois: hiring aggressively adds 'ammunition without adding barrels.' This essay extends the metaphor to the AI age — the senior veterans big firms accumulate are not merely non-barrels but anti-barrels: assets that are simultaneously reasons not to move, so the deepest-bench firms are worst positioned.
- explicitThe decisive work of the next decade is not accumulating human capital or token capital but building an organizational structure 'porous' enough to let uncorporate people stay uncorporate while giving them corporate-scale compute, data, distribution and responsibility — keeping the Demises and Jeff Deans from being either crushed or domesticated.conf: high
- explicitIn the age of intelligence the scarcest capital is neither the human nor the machine, but the organizational structure brave enough to hold both without flattening either.conf: highrefinesHard truths about building in the AI era | Keith Rabois (Khosla Ventures) · p1 — Rabois makes talent 'the single most important variable' in success; this essay refines that — above talent sits the organizational structure that can hold the right talent. Talent alone, wrongly structured, wastes token capital, so the scarcest capital is the structure, not the people.
- implicitOrganizational structure is the transfer function between AI capital and outcomes: compute does not self-actualize, so whether token capital produces value or 'runs in circles' depends on who is allowed to wield it and how the org selects and holds them.conf: mediumrefinesAgentics / Tech Things: Tokenmaxxing is dead, long live tokenmaxxing · p6 — Tokenmaxxing #6 argues maximal token spend has become ROI-rational and value-driven (compounding correctness). This essay adds a prior structural condition on that value: token capital only converts to value when wielded by frontier-movers inside a porous structure; in a rigid corporate structure the same spend 'runs in circles,' so ROI-rationality is contingent on org design, not just on the inference math.
Arguments
Nadella is right that token and human capital compound, but he omits the decisive qualifier — which human capital — because the kind big firms manufacture best is precisely the kind that wastes token capital. strength: high
Explicit premises:
- The best current models were built by bold, uncorporate people who succeeded in spite of, not because of, their corporate employers.
- The politically-fluent veteran a big firm manufactures best optimizes the org chart, not the loss curve.
- The irreverent frontier-mover is the kind a rigid corporate structure selects against and pushes out.
Implicit premises:
- The value of human capital depends on its type, not just its quantity or seniority — human capital is not a homogeneous, additive stock.
- A firm's recruiting and promotion machinery is a selection filter that systematically favors one talent type over the other.
- Token capital only converts to value when wielded by frontier-movers; in the hands of org-chart optimizers it is wasted.
Inferences:
- If the talent a firm accumulates is the wasteful type and the useful type is filtered out, then 'more human capital' can lower, not raise, the return on token capital.
- Therefore the visible compounding thesis conceals an inverse risk for exactly the firms that look strongest on the old scoreboard.
The scarcest and most decisive capital in the AI age is an organizational structure porous enough to hold uncorporate talent and corporate-scale compute together without flattening either. strength: medium
Explicit premises:
- Frontier talent is crushed or domesticated by rigid corporate structure but needs corporate-scale compute, data and distribution to do its best work.
- Firms with the deepest benches and proudest org charts may be worst positioned because every such asset is also a reason not to move.
- Firms that keep promoting their best ladder-climbers keep wondering why their token capital runs in circles.
Implicit premises:
- Neither talent alone nor compute alone is the binding constraint; the binding constraint is the structure that mediates between them.
- It is possible in principle to design a structure that grants corporate-scale resources without imposing the committees and incentives that sand talent down.
Inferences:
- If both human and token capital are abundant but only a rare structure can combine them productively, then that structure is the actual scarce resource.
- Therefore the competitive winner is the firm that becomes porous, not the one that accumulates the most of either capital.
Narratives
- Satya Nadella is cast as 'the Warrior, who can crush your skull,' who suddenly sits down on a stone and turns into The Thinker, 'at the edge' — staring at the possibility that his greatest asset and his heaviest anchor are the same people.
→ Dramatizes the essay's central claim through the most powerful possible witness: even the leader of the dominant incumbent has recognized the trap, which lends the diagnosis authority and stakes. - The frontier-builders are drawn as sideways geniuses: Demis Hassabis as Ender Wiggin 'playing and fighting Google's corporate game,' Jeff Dean who 'like Chuck Norris, just looks at the model until it starts converging,' and Liang Wenfeng who ran a hedge fund before DeepSeek — succeeding in spite of their corporations.
→ Concretizes the abstract 'not ladder creatures' category with named exemplars, making the talent split vivid and the 'in spite of, not because of' claim feel evidenced rather than asserted.
Questions addressed
- ✅ What does Nadella mean by token capital and human capital, and is he right that they compound?
The essay restates both definitions faithfully and agrees they compound, but argues Nadella is 'too polite, or too strategic' to say WHICH human capital compounds — because the kind big firms manufacture best is the kind that wastes token capital. - ✅ Under what conditions can corporate America hold frontier talent without domesticating or exhausting them — or does the age of AI reward a shape that looks less like a company and more like a band, a lab, a small huddle of fanatics with infinite compute behind them?
The essay answers with the 'porous structure' thesis — a form that grants corporate-scale compute and responsibility while letting uncorporate people stay uncorporate — while leaning toward the band/lab shape as the emerging winner. The answer is named but not operationalized. - ⭕ Does this diagnosis apply to the ordinary enterprise that BUYS token capital (a telco, a bank) as much as to the frontier lab that BUILDS it?
Left unresolved. The essay slides between 'the best models were built by uncorporate people' and 'corporate America / almost every enterprise,' treating frontier labs and token-capital consumers as one problem.
Gaps
- incompleteThe lab-vs-enterprise conflation. The argument is strongest for frontier labs that BUILD token capital (models, evals, loops require irreverent frontier-movers) and weakest for the telco, bank or typical enterprise that BUYS and absorbs it. A firm that consumes AI rather than pushing the frontier may not need 'permission-averse, loudly-wrong' talent at all — for it the binding constraint is more likely absorption, integration and process legibility than porousness for mavericks. The essay treats the two as one problem and never tests whether its prescription even applies to the majority of the 54% of employment it invokes.
Read next: Read against the enterprise-AI-adoption literature that separates building vs. buying/absorbing AI capability (e.g. forward-deployed-engineer and readiness models): does a token-capital consumer need frontier-mover porousness, or is its real obstacle organizational absorption, and would importing 'uncorporate' talent even help? - incomplete'Porous structure' is named as the answer but never operationalized. The essay supplies no mechanism, org-design pattern, incentive structure, or worked example of what porousness looks like, nor how a firm hands 'corporate-scale compute + responsibility' to uncorporate people without recreating the committees and procurement logic that sand them down. The prescription is a direction, not a design.
Read next: Read toward concrete structural patterns that could fill the mechanism — small-teams/tiny-core bets, autonomous-team designs, skunkworks and lab-within-a-company structures — and test which actually preserve irreverence at corporate scale. - illogicalThe essay's own evidence strains against its thesis. Demis Hassabis built Gemini INSIDE Google, a corporate giant — 'in spite of' it, yes, but still inside it and still winning — which shows rigid incumbents CAN produce frontier results, cutting against the flat claim that structure 'selects frontier talent out.' By naming only the winning obsessives (Hassabis, Dean, Liang Wenfeng) it also survivorship-selects, saying nothing about the uncorporate people who were in fact crushed. So 'structure selects them out' is asserted more strongly than the evidence, which really shows 'structure taxes them, but some still win.'
Read next: Read comparatively for cases where incumbents DID retain and productively hold frontier talent, to separate the strong claim ('structure filters them out') from the weaker, better-supported one ('structure imposes a heavy tax they sometimes overcome').
Closes gaps elsewhere
- Your AI strategy has a trust problem, not a tooling problem · g2 — That extraction's open gap: the Anthropic/Lovable high-agency examples are AI-native startups that pre-select high-agency operators, and it doesn't address whether the model travels to a large incumbent. This essay supplies the structural obstacle — incumbents structurally select FOR org-chart optimizers and AGAINST high-agency frontier talent, so the model does NOT travel cleanly; the incumbent must first become 'porous' enough to hold uncorporate people with corporate-scale compute. (Partial closure: it names the obstacle without giving the mechanism to overcome it.)
Related documents
- Agentics / Tech Things: Tokenmaxxing is dead, long live tokenmaxxing
- Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI
- 11) AI-spesial med Vilde Opsal og Anders Sveen
- Adam Mosseri: AI is a tailwind for authenticity
- The engineering manager role is splitting in two
- Managers Are Not Overhead: They Are Infrastructure
- Hard truths about building in the AI era | Keith Rabois (Khosla Ventures)
- The Agentic P&L: Beyond the Empire of Headcount
- Designing the AI-native engineering organization
- 🧠 Community Wisdom: How AI is changing product operating models, tracking work stress with Whoop, whether you nee…