Mariana Mazzucato

Who she is

Affiliation: Professor in the Economics of Innovation and Public Value, University College London; Founding Director, UCL Institute for Innovation & Public Purpose (IIPP, est. 2017) — role check run 2026-09-21: still Founding Director (Whitechapel Gallery press release, April 2026; Foreign Affairs byline, early 2026; Democracy Now introduced her by the professorship, 2026-09-15). CBE "For services to Economics" in the 2025 King's Birthday Honours (June 2025 — v1 carried only "(2025)"). Born Rome, 1968; Italian, US and UK citizen (Wikipedia, secondary). Hats counted this cycle, dated from the counterparty side wherever possible: South African Presidential Economic Advisory Council (member since 2019, current per the Presidency, 2025-04-07); Technical Expert to South Africa's G20 Presidency and Special Presidential Representative to Taskforce 1, also contributing to "Taskforce 3: Artificial Intelligence, Data Governance, and Innovation for Sustainable Development" (appointed 2025-04-07; ended with the presidency — South Africa "formally handed over the G20 presidency to the United States" on 2025-11-25); Chair, WHO Council on the Economics of Health for All (established 2020-11-13; final report 2023-05-23; WHO still lists the Council, no 2025–26 output found); Co-chair, Global Commission on the Economics of Water (final report October 2024 — concluded); co-chair of the experts group for Brazil's 2024 G20 climate taskforce (report Oct 2024, per her own site — not fetched this cycle); Economist-in-Residence, Whitechapel Gallery (new: "from April 2026 until the end of March 2029", "pro bono"); Special Advisor to the EU Research Commissioner (2017–2019, ended). Her About page also lists "Commissioner for the Jubilee Report" (Vatican, launched 2025-06-20) — not verified: the two launch pages fetched name only Stiglitz and Guzmán. Wikipedia adds a Scottish Council of Economic Advisers seat (since 2015) and a "UNESCO High Level Expert Group on Ecosystem Level AI Governance" — Wikipedia-only, no UNESCO page found; neither is entered as fact.
One-line position: The state funded the foundational research behind AI (and the internet, GPS, touchscreens) — if the public bore the risk, it should share the reward, and AI governance must be designed to create public value, not just private rents (rents = profits above what competition would allow). Since June 2026 the frame is bigger: a "common good economy" steered by a five-part compass, with AI as one application rather than the subject.

Discipline & technical bet

Innovation economist and political economist of the state — a non-technical thinker whose "technical bet" is institutional, not algorithmic: that who owns, conditions and directs compute, chips and R&D money decides who captures AI's surplus. Concretely (2025-02, with Gernone, Ramos and Bria): "pre-distributive structures that share risks and rewards of AI innovation fairly from the start" — equity stakes, reinvestment conditions, public compute, a public option — rather than taxing winners afterwards; and (with Valletti, 2025-02-11) "a decentralized innovation ecosystem that serves the public good" instead of a national-champion "AI superpower" race. The 2026 wager is the book itself: The Common Good Economy: A New Compass (Allen Lane, 2026-06-04; US edition How to Make Capitalism Work for Us All, Basic Venture, 2026-09-08) — five elements ("Purpose & Directionality", "Co-creation & Participation", "Collective Learning & Knowledge-Sharing", "Access for All & Reward-Sharing", "Transparency & Accountability") offered as an operating manual for the mission state. No AI appears in either publisher's description; her AI position is carried in columns, not the book.

Key claims (Says)

• The state is the primary risk-taker in foundational innovation: DARPA, NIH, NSF funded the technologies AI runs on; private firms commercialize, but public investment created the platform. Restated 2025-02-03: "What would Google be without the DARPA-funded internet? What would UBER be without the US Navy-funded GPS?" Held — the historical claim is uncontested in-window, and the state kept behaving as she describes: the Genesis Mission EO (2025-11-24) builds "an integrated AI platform to harness Federal scientific datasets — the world's largest collection of such datasets, developed over decades of Federal investments."
• "AI for What?" (2025-02-03, Substack, co-authored with Fausto Gernone, Gabriela Ramos and Francesca Bria): urgent call to shape AI development before it becomes another rent-extraction mechanism; "today's algorithmic systems are increasingly being used to extract what we call 'algorithmic rents' - using AI and machine learning not to create genuine value, but to concentrate market power"; pre-distributive structures needed from the start; "The question isn't whether Europe or the UK can become an 'AI superpower,' but whether they can help build an AI ecosystem that serves the common good." Held (position) — still endorsed: her 2026-06-23 twelve-point Labour agenda carries a "Future of Technology" point on algorithmic rents and data governance, and on Democracy Now (2026-09-15) she said "I believe it's almost impossible to govern AI for good...when you don't have the knowledge inside." The prediction inside this claim is graded separately below.
• Mission-oriented policy: governments should set direction (like moonshots), not just fund basic research — AI should serve defined public purposes (health equity, climate, etc.). Held (position) — restated through the window (Project Syndicate 2026-03-24: "Through mission-oriented green industrial strategies, governments can help secure living standards and build economic resilience simultaneously"); the vocabulary spread further than v1 knew (World Bank "Mission 300"; a Trump EO literally named the Genesis Mission, "comparable in urgency and ambition to the Manhattan Project"). Whether missions work is the prediction graded below, and it did not go well.
• Value creation vs. value extraction: AI companies claiming to "create value" may actually be extracting it — Mazzucato's core analytical distinction. Framework, not a prediction — unfalsifiable as stated; noted, not graded. The nearest observable is in the incentive/rent check under predictions.
• Risk-reward symmetry: if public funds bore the risk of foundational research, the public should capture returns (equity stakes, IP conditions, price caps). Held on mechanism, valence inverted — her instrument was adopted by her ideological opposite: on 2025-08-22 "The US agreed to purchase a 9.9 percent stake in Intel for $8.9bn at a price of $20.47 a share," funded from "the $5.7bn in unpaid CHIPS Act grants and $3.2bn awarded to Intel for the Secure Enclave program" — a grant-to-equity conversion, exactly the pre-distributive move she has argued for since 2013. She then graded the execution, not the instrument: "Trumponomics" offers "the growing array of companies that the U.S. government is taking equity stakes in" but "is not providing a good deal for the American public" (Substack, 2026-02-13, pointing to her Foreign Affairs essay Making Industrial Strategy Great Again). Mechanism Held; the value she wanted attached (conditions, strategy, worker gains) is not in the Intel terms.
• Industrial policy is back: AI makes the case for active state involvement stronger, not weaker — the laissez-faire consensus is obsolete. Held — in-window market-level evidence from three directions: the US (Intel stake 2025-08; Genesis Mission 2025-11), the UK (The UK's Modern Industrial Strategy, written statement HCWS725, 2025-06-23), Europe (the EuroStack report, Feb 2025, from her own institute's people). The bench's regulatory-model persona conceded the point: "We're seeing a real convergence toward the state-driven model in many ways. Everyone is playing Beijing's game now" (Bradford, Capgemini, 2026 — verified against the Capgemini page this cycle; the page carries no date, Bradford's file dates it 2026-03).
• NEW (2025-08-01, "Efficiency at What and for Whom?"): "DOGE wasn't an aberration or a failed experiment; it was the latest chapter in a fifty-year project to dismantle state capacity"; "DOGE's claimed $170 billion in savings were largely fictional"; "Musk's business deals with the government total nearly $20 billion." Extends The Big Con (2023, with Rosie Collington) from consultants to the efficiency movement. Not graded — the savings figure and the Musk-contracts figure are hers; neither was checked against a primary this cycle.
• NEW (2026-09-15, Democracy Now): "what we ended up doing was socializing the risks and privatizing the rewards"; "the labor share of global income is at one of the lowest levels it's ever been in history"; "The CHIPS Act, which I helped design..." — a self-attributed credential, see Track record. And the Tesla example: had the government taken shares, "the price per share in 2009...was $9 per share. When it was paid back in 2013, it was $90." Flagged: a 2009 per-share figure is pre-IPO and not a public quote; not verified this cycle.

Notable predictions — with falsifiable checks

• (2013) States that invest boldly in innovation will out-compete those that don't. v1: partially supported (China's AI investment, EU industrial policy revival), contested in US context. Drifted — the market-level data now cut both ways. Stanford AI Index 2026 (2026-04-13): US private AI investment "($285.9 billion) were 23.1 times greater than those of the next-highest country, China ($12.4 billion)" — the private-capital country still leads on money — while the same Index reports China narrowing the frontier-model gap (exact figure not read off the HAI page this cycle). An aggregate can grade the headline, never the attribution (whose investment did it): the state-vs-private mechanism is ungradeable from these numbers. Check: AI Index 2028 — does the US–China model gap close to parity or reopen, and does any peer-reviewed decomposition attribute China's catch-up to state guidance funds? Absent both by 2028-06, retire the prediction as unfalsifiable at country level.
• (2025) Without pre-distributive structures, AI will become another rent-extraction tool. v1: TBD; early evidence mixed. Held (early, trending) — antecedent holds: no G7 government has enacted a pre-distributive AI instrument (public equity in a frontier lab, conditioned public compute, data dividend) as of 2026-09-21 — the Intel stake is chips, not AI, and came with no conditions she recognizes. Consequent trending: NVIDIA reported "For fiscal 2026, revenue was $215.9 billion, up 65% from a year ago" and "GAAP and non-GAAP gross margins were 71.1% and 71.3%" (2026-02-25) — a 71% gross margin at $216B is the textbook shape of a rent; and private AI money is one-country concentrated ($285.9B of $344.7B global, AI Index 2026). Not yet Held outright because "rent" needs a counterfactual competitive margin she has never specified. Check, end-2027: (a) any G7 pre-distributive AI instrument enacted (falsifies the antecedent); (b) NVIDIA FY2027 gross margin below 60% or the top-5 AI firms' share of global private AI investment falling two years running (falsifies the consequent). Both absent = Held; (a) present with (b) absent = the thesis is confirmed the hard way.
• (Ongoing) Mission-oriented policy is more effective than market-neutral regulation. v1: contested; limited controlled evidence. Drifted — the flagship real-world test failed on her own account. Chronology, institution dates beside hers (Rifkin rule): Mission Economy Jan 2021 → Starmer's five missions 2023-02-23 (Institute for Government: "The concept of missions and 'mission-led' government was popularised by the economist Mariana Mazzucato") → Plan for Change Dec 2024 → Reeves narrows to three priorities Nov 2025 (her 2025-12-22 post: "the government appears to have shifted focus from its missions to a narrower emphasis on three priorities") → her own verdict 2026-06-23: "the UK's missions became a slogan rather than a governing architecture" → Starmer resigns (announced by 2026-06-23, her post opens "With the news of Keir Starmer's resignation") → Burnham confirmed leader 2026-07-17, PM the following Monday (2026-07-20). Mechanism (adoption of missions) Held; valence (delivered better outcomes) not shown. Not graded Wrong because she pre-registered the failure mode in July 2024 — "Growth is not a mission. The NHS is not a mission. Growth is the result." — but that escape hatch ("if it failed it wasn't a real mission") is why this claim is nearly unfalsifiable. Check: by end-2027 the Burnham government either (a) rebuilds a cross-departmental, Treasury-backed mission architecture on her stated conditions, in which case grade Clean Power 2030 delivery (the one mission she cites as working) at end-2028, or (b) does not, in which case the UK test is closed as a miss on implementation and the claim needs a new jurisdiction to be gradeable at all.
• (2025-04-11, "The Trump Vacuum") Trump's tariffs will raise consumer costs by at least $1,500 per household per year and "Brazil's COP30 presidency and South Africa's G20 leadership can forge a new alliance for meaningful climate action." Household-cost half: not graded — no primary checked this cycle; a figure she cites, provenance not checked. Alliance half: Drifted — both presidencies ran (COP30 in Belém, Nov 2025; the Johannesburg G20 summit, Nov 2025) and the US boycotted the G20 summit, but no standing "alliance" institution emerged; the G20 passed to the US on 2025-11-25 and her own G20 role ended with it. Check: any named Brazil–South Africa climate-finance vehicle with a budget by end-2026; none = Wrong.
• (2025-02-11, with Valletti) The UK AI Opportunities Action Plan — "boost the AI computing power under public control 20-fold by 2030" — will not serve the public interest without competition and decentralization measures. Unfalsifiable as written (no observable for "serve the public interest") — rewrite or retire. Proposed observable: share of UK public compute allocated with conditions on model openness or reinvestment by 2028; the 20x compute target itself is gradeable at 2030.

Revealed behavior (Does)

• Where her time went, 2025-09 → 2026-09 (counted, Diane Francis rule): 12 Substack posts (school meals, common-good lecture, COP30, UK missions, Brazilian Carnival economics, US industrial strategy, procurement, British Steel ×2, the book, Labour's next leader, Notting Hill Carnival, swimmable cities) and 5 Project Syndicate columns in 2026 (World Bank, British Steel, competitiveness, devolution, swimmable cities) plus a March column on the Iran energy shock — zero AI-titled pieces in twelve months. Her AI output is a February-2025 cluster (Substack 02-03, Project Syndicate 02-07 and 02-11), a remote UNESCO conference slot (Paris, 2025-06-04/05) and a G20 Taskforce 3 contribution. AI is a chapter in her argument, not the argument.
• Wrote the book — The Common Good Economy, begun at a 2024 Rockefeller Bellagio residency, published UK 2026-06-04 and US 2026-09-08, with a New School launch and a Democracy Now slot; blurbed by Piketty. A commentator's job change is a book (Doctorow rule): this is the fifth in thirteen years and the first to fold the AI, water, health, procurement and culture work into one frame.
• Took a pro-bono three-year seat (Whitechapel Gallery, April 2026 – March 2029) extending her Public Value of Arts and Culture programme — the newest direction of travel is culture and place (pools, carnivals, galleries), not technology.
• Kept the South Africa hat and let the G20 one lapse with the presidency; no new state advisory appointment found in the window. The 2025 G20 role is the first time an AI taskforce (TF3) appears on her official brief.
• Institute output continued in her vocabulary: "Mission-Oriented Country Platforms" working paper (WP 2025-20, posted 2025-11) and "A Mission-Aligned World Bank" (posted 2026-04; the World Bank has adopted "Mission 300" and "Mission Water" as organizing frames).
• Endorsed the instrument, attacked the operator on the Intel stake — she did not walk back equity stakes when Trump used them; she wrote the Foreign Affairs rebuttal on strategy. Says and Does are consistent here.

Feels

A settled conviction that she won the argument (industrial policy is back) and a rising fear of losing the execution — her 2025-12 and 2026-06 UK posts read as someone watching her signature idea become "a slogan." Contempt for what she calls infantilised government ("I refuse to call the tech bros 'tech bros.' I just call them 'bros'...the tech didn't come from them").

Hears

Keynesian and Schumpeterian lineage (Polanyi, Keynes, Schumpeter, Carlota Perez); the Stiglitz circle (Jubilee, Korinek's route); Rodrik on conditionality (their 2024 taxonomy paper); her own institute's digital-sovereignty wing (Bria, Rikap, Gernone — the EuroStack authors) for AI; finance ministries and presidencies of the Global South (Ramaphosa's PEAC, Lula's G20 taskforce) rather than Silicon Valley or the frontier labs.

Sees

The inside of the mission state as it is actually run — a presidential advisory council, a G20 taskforce, a WHO council, a global water commission — so she sees why missions stall (Treasury, departmental silos, consultants) earlier than academics who only read the policy papers. She does not see model training runs, chip roadmaps or lab economics from the inside; her AI vantage is downstream of the political economists who do.

Incentive map

Re-derived from scratch this cycle. Who pays: UCL salary; IIPP's philanthropic partners — Open Society Foundations (three-year strategic partnership announced 2020-12), Rockefeller Foundation (2018, renewed 2022-04 for the Mission-Oriented Innovation Network; it also hosted the 2024 Bellagio residency where the book began and published the 2026-05-22 profile carrying the CHIPS-Act line), Laudes Foundation (2021-05), an Omidyar/Ford/Hewlett collaborative (all per UCL announcement pages, dated by URL, not fetched); book advances and royalties (Penguin/Allen Lane, Basic Venture); paid keynotes (listed with Chartwell Speakers); government and IO advisory seats (South Africa PEAC current; G20 ended; Whitechapel pro bono). What she is selling: the mission/common-good framework itself — the newest vehicle (a book with a five-part "compass") matches the newest thesis exactly (Ng rule: the thesis may be the product), and the institute's revenue depends on governments buying "mission-oriented" capacity-building. What she can't say: that a mission failed because missions fail (the escape hatch is structural); that her funders' politics (OSF, Rockefeller) shape which missions get named; that the UK Labour experiment — her most visible client — has closed as a miss. Advisory conflicts (Bradford rule — an "advises government X" line is an incentive entry first): she advised the Ramaphosa government while co-writing the G20 growth agenda it presented; she counts the CHIPS Act as hers while assessing the Intel stake that converted its grants. Rule-25 discount: moderate-to-high on UK-policy and mission-effectiveness claims (she is grading her own product), moderate on AI-rent claims (no lab or vendor money found, but the digital-sovereignty wing of her institute is a EuroStack stakeholder — reported ~€300B proposal, figure not read off the report), low on the historical state-as-investor claims, which predate every current funder.

Theories aligned with

• Entrepreneurial state / mission-oriented innovation policy (her own; now "common good economy")
• Value theory (who creates vs. who extracts economic value)
• Adjacent to post-labor economics — addresses who captures AI-generated surplus; her "labor share of global income is at one of the lowest levels it's ever been in history" (2026-09-15) is the same observable that theory watches
• Digital sovereignty / EuroStack (institutional, via IIPP's Bria–Rikap–Gernone line; she is not an author of the 2025 report)

What she's reacting against

• The Silicon Valley narrative that innovation comes from private entrepreneurs, not public investment — The Entrepreneurial State (2013) is a direct rebuttal
• AI as a new vehicle for rent extraction — argues (2025-02) that without pre-distributive structures, AI will concentrate wealth further; "Already dominant monopolists will supercharge the socially harmful digital business models they perfected over the past decade" (Project Syndicate, 2025-02-07)
• The assumption that market competition alone produces socially beneficial AI — insists on active "mission-oriented" industrial policy
• Passive regulation (setting rules and stepping back) — advocates for the state as an active market-shaper and co-investor
• NEW: "efficiency" as cover for dismantling state capacity — DOGE as "a fifty-year project to dismantle state capacity" (2025-08-01); "The infantilization of government through the consulting companies" (2026-09-15)
• NEW: strategy-less state capitalism — Trump's equity stakes as "interventions without coordination, and spending without strategy" (her X post, 2025-09, quoted from the search index; the post itself could not be fetched) and, on the record, "not providing a good deal for the American public" (Substack 2026-02-13)
• NEW: "AI superpower" nationalism in Europe and the UK — the race framing itself is the target (2025-02-03)

Where she overlaps / splits (with Rich)

• Overlaps with Acemoglu on institutional design mattering for AI outcomes; splits on mechanism — Acemoglu emphasizes labor market reform, Mazzucato emphasizes state investment and ownership. Author-list check: no co-authorship found.
• Overlaps with Crawford on AI as a power structure; splits on prescription — Crawford diagnoses, Mazzucato prescribes (state as builder)
• Overlaps with Reich on political economy of concentration; splits on register — Mazzucato is more academic/institutional, Reich is more populist (no Reich file on the bench)
• Splits with Andreessen/Thiel fundamentally — they see state involvement as the problem; she sees it as the solution. In-window twist: the administration they backed took a 9.9% stake in Intel; the split is now inside their own camp.
• Splits with Cowen on regulatory friction — Cowen sees state action as drag on diffusion, Mazzucato sees it as necessary for equitable diffusion
• Converging with Bradford since 2026-03 — Bradford's "everyone is playing Beijing's game now" is Mazzucato's state-shapes-markets thesis stated by the regulatory-model camp; still splits on focus (industrial policy vs. standard-setting)
• Overlaps with Khan on challenging tech-industry narratives; splits on mechanism — state investment vs. competition enforcement; her 2025-02-11 co-author Valletti (a competition economist) is the bridge
• Splits with Karp on the state's role — she wants the state to direct innovation, Karp wants the state to buy from companies like Palantir
• Overlaps with Korinek (via Stiglitz) on steering technology's direction; splits on the lever — mission-driven state investment vs. tax design and scenario planning
• Adjacent to Tooze — same state-investment terrain, but he is diagnostic and she is prescriptive
• Splits with Rifkin on the state — her mission-led state vs. his lateral commons; note the shared failure mode both files now carry: a dated influence claim ("my ideas shaped policy X") graded against the institution's own dates
• Overlaps with Graylin on mission-driven public research (the CERN model) — his file cites her, not the reverse
• Overlaps with Aschenbrenner, unexpectedly, on the state running the AI project; splits on purpose (public value vs. beating China) and on scale — the nearest real event, the Genesis Mission EO (2025-11-24), uses his Manhattan-Project language and her "mission" noun for a DOE science platform; the national-AGI-project announcement his file watches for was not found this cycle

Track record

• The Entrepreneurial State (2013; 10th-anniversary edition 2023) reshaped the public debate about who funds innovation — genuine Overton window shift. Held and strengthened: by 2026 the market-driven US converts grants to equity and names AI programmes "missions", and the bench's regulatory-model persona says "everyone is playing Beijing's game now."
• Advisory roles to multiple governments (EU, UK, South Africa, WHO) — institutional influence is real. Held, re-dated (consultant rule: date every client from the client's side): EU 2017–19 ended; WHO council's final report 2023-05-23; Water commission concluded Oct 2024; South Africa PEAC current (2019–), G20 role 2025-04-07 → 2025-11-25 ended; Brazil G20 experts group 2024 ended; UK: the "missions" adopted 2023-02 and hollowed out by 2025-11 on her own account; Whitechapel 2026–29 new. Of ten listed bodies, four are live.
• CBE appointment signals establishment recognition. Held, dated: 2025 King's Birthday Honours, "Professor in Economics of Innovation and Public Value, University College London. For services to Economics" (honours list as transcribed on Wikipedia; her own site says "2025 King's Birthday Honours"; the London Gazette entry was not read this cycle).
• Critics argue her policy prescriptions are better at diagnosis than implementation — "mission-oriented policy" is hard to execute in practice. Held and now conceded by her: "the UK's missions became a slogan rather than a governing architecture" (2026-06-23). Dated critics in-window: Atkinson (Policy Arena, 2025-07-18: "Mazzucato has cleverly deployed a sleight-of-hand trick" to "hijack innovation policy in service of left-wing social goals"); Thomas (Independent Review, Spring 2026: "Simply pointing out discrepancies between an ideal and reality doesn't prove inefficiency"; she "ignores dispersed knowledge, coordination costs, and principal-agent issues"; "Larger government and a solid consensus supercharges cronyism"); O'Neill (SSIR, Fall 2026: "you cannot conjure an ethos by describing it, nor assume away the structural forces that create unequal economic power"); Donohoe (Irish Times, 2026-07-18: "a positive and very elegant work" but "The compass does not fully address issues of political economy" — note the reviewer is a sitting finance minister).
• NEW — CHIPS Act design claim: "The CHIPS Act, which I helped design" (Democracy Now, 2026-09-15); "Our team at the IIPP advised the Biden administration on the design and implementation of the 2022 CHIPS for America Act" (Rockefeller Foundation profile, 2026-05-22); City AM (2024-07-17) has the softer "helping to frame ... Biden's Chips Act." Self-attributed; no Commerce Department or Congressional primary found. The Harvard Kennedy School post on CHIPS conditionalities (2024-02-21) cites her taxonomy paper with Rodrik and makes no advisory claim. Per the Raskin rule this sits here as a claim, not a credential, until a counterparty document appears.
• Pattern: exceptional at moving vocabulary and instruments into institutions (missions, conditionality, grant-to-equity); unproven at the delivery layer, and the one flagship delivery test she could point to closed against her in 2026.

Empirical vs normative

• Empirical: historical analysis of public investment in innovation (DARPA, NIH, etc.); case studies of state-led technology programs; the 2024 conditionality taxonomy with Rodrik; the 2022 IIPP–Stanford working paper "Governing artificial intelligence in the public interest" (with Schaake, Krier, Entsminger) — the one AI-specific research artifact, pre-window
• Normative: the state should be an active investor and market-shaper; risk-reward should be symmetrical; AI governance should serve public purpose; since 2026, the five-part common-good compass
• The normative claims are strong and contested — industrial policy advocates accept them, market liberals reject them; the 2025–26 twist is that a market-liberal administration adopted the instruments while rejecting the norms
• Conflict note (CONVENTIONS rule 25): moderate-to-high on mission-effectiveness and UK claims (grading her own product), moderate on AI-rent claims, low on the historical claims — see Incentive map

Weak spots / open questions

• "Mission-oriented policy" sounds good in theory but has mixed implementation track record — not every state is capable of smart industrial policy. Held — the UK 2023–26 case, in her own words.
• The diagnosis (public funded the risk) is strong; the prescription (public should capture the reward) is institutionally complex and politically contested. Held — the Intel stake shows the instrument is adoptable overnight and the strategy is not.
• Less technically engaged with AI specifically — her arguments apply to technology in general, not AI's unique characteristics. Drifted — more engaged than v1 said (three AI pieces in Feb 2025, a G20 AI taskforce, the 2022 working paper, DeepSeek and compute concentration in her text), but through political economy, never technical characteristics; and by output count AI was 0 of 18 pieces in the last twelve months. The v1 sentence was half right for the wrong reason.
• Risk of overstating state's role: private capital and entrepreneurial risk-taking do matter, even if the public sector laid groundwork. Held — the Thomas "nirvana approach" critique (Spring 2026) is the sharpest version.
• Advisory influence across many governments may spread too thin to evaluate effectiveness. Held — ten-plus bodies, four live, no delivery metric published for any.
• NEW: the escape hatch — every failed mission is redefined as "not a real mission" (growth, the NHS), which protects the thesis from any single jurisdiction's result; the check attached above is the file's attempt to close it.
• NEW: two self-attributed items that no counterparty confirms — the CHIPS Act design role and the Jubilee Commission seat — plus a Wikipedia-only UNESCO AI expert-group membership; each needs a primary before it can count.
• NEW: her AI numbers are inherited (the "80%" US+China share of 2013–21 private AI investment; the $1,500 tariff cost; DOGE's "$170 billion") — none is hers and none was checked this cycle; she is a consumer of other people's estimates on AI, which is fine for a political economist and should be read that way.

Rich's take

• (your synthesis here)

Delta log

2026-09-21 — v1→v2 migration + validation (bulk run)

• Grades: 14 Held / 4 Drifted / 0 Wrong (2 items unfalsifiable — the value-creation framework and the 2025-02-11 UK-plan claim; 4 numbers noted, not graded). Zero Wrong is not a defective-review signal here: her key claims are positional and the window's events (Intel stake, Genesis Mission, UK Industrial Strategy, Bradford's concession) adopted her instruments; the losses are all at the delivery layer, logged as Drifted with checks.
• Drifted: bold-state-out-competes (AI Index 2026: US private AI investment 23.1x China's, yet China closing the model gap — attribution ungradeable, Stanford HAI, 2026-04-13); mission-effectiveness (her own "became a slogan rather than a governing architecture", Substack, 2026-06-23; Reeves narrowing, Substack, 2025-12-22; Burnham confirmed, Al Jazeera, 2026-07-17); Brazil–SA alliance (G20 handed to the US, allAfrica, 2025-11-25); "less technically engaged with AI" (three Feb-2025 AI pieces vs 0 of 18 outputs in the last year).
• Held: state-as-risk-taker (Genesis Mission EO, 2025-11-24); AI-for-What position (restated 2026-06-23 and Democracy Now, 2026-09-15); mission-oriented position; risk-reward symmetry on mechanism (Intel 9.9% stake, 2025-08-22) with valence inverted (her 2026-02-13 critique); industrial-policy-is-back (Intel; Genesis; UK Modern Industrial Strategy HCWS725, 2025-06-23; Bradford at Capgemini, 2026); rent-extraction prediction trending (NVIDIA FY2026, 2026-02-25); all four v1 track-record lines; four of five v1 weak spots.
• Most surprising delta: her signature instrument (convert public grants to public equity) was enacted by the Trump administration at Intel, and her signature programme (missions) was hollowed out by the Labour government that adopted it — the ideological opposite delivered the mechanism while the ally delivered the slogan.
• Employer/role check (mandatory): affiliation unchanged (UCL/IIPP Founding Director confirmed April and September 2026). Hats re-enumerated: one new (Whitechapel Economist-in-Residence, pro bono, 2026-04 → 2029-03, press release), one added and ended inside the window (South Africa G20 Technical Expert / Taskforce 1 rep with an AI Taskforce 3 brief, Presidency, 2025-04-07; presidency handed over 2025-11-25), two listed-as-current bodies actually concluded (WHO council final report 2023-05-23; Water commission Oct 2024). CBE dated to the June-2025 Birthday Honours. New book: UK 2026-06-04 (Penguin), US 2026-09-08 (Hachette). The June-2026 v1 "review" mentioned none of the Feb-2025 AI pieces, the April-2025 G20 role, the June-2025 CBE date or the forthcoming book — another no-current-events pass.
• Incentive map re-derived: added IIPP's foundation funders, the Rockefeller–Bellagio–book–profile loop, paid keynotes, the CHIPS-Act self-attribution as a conflict on her Intel commentary, and the EuroStack stake of her institute's digital wing. Discount raised on UK/mission claims.
• Unverified and flagged, not entered: Jubilee Commission seat (launch release, 2025-06-20, names only Stiglitz and Guzmán); UNESCO AI expert group (Wikipedia only); Foreign Affairs essay day (listed 2026-01-28 in search results; FA page not fetched); Tesla $9/$90; the 2025-09 X post (paywalled fetch).
• Tier: semiannual — high output but slow grade-moving events; the decisive test (whether Burnham rebuilds the mission architecture) resolves on a Budget cycle, and the US book tour ends in autumn 2026. Wake triggers cover appointments, the UK architecture decision, any pre-distributive AI instrument, an AI-lab equity stake, and a counterparty document for the CHIPS claim.
• Confidence unchanged at medium — sources are abundant and primary, but her core claims are normative and the one delivery test is in her own voice.

Sources

• Her AI writing (all Feb 2025): "AI for What? Public value creation versus extractive rents" — Substack, 2025-02-03 (with Gernone, Ramos, Bria) · "Resisting Digital Feudalism" — Project Syndicate, 2025-02-07 · "Governing AI for the Public Interest" — Project Syndicate, 2025-02-11 (with Valletti) · UNESCO conference on AI and digital transformation in the public sector, Paris, 2025-06-04/05 · "Governing artificial intelligence in the public interest" — IIPP WP 2022-12 (with Schaake, Krier, Entsminger)
• 2025–26 Substack ("Mission Economics"): "The Trump Vacuum" (2025-04-11) · "Efficiency at What and for Whom?" (2025-08-01) · "Closing Out 2025: Don't give up on missions just yet, Prime Minister!" (2025-12-22) · "Industrial Strategy: the art of the real deal" (2026-02-13) · "The Common Good Economy: a new compass" (2026-06-04) · "What Labour Should Do Next" (2026-06-23) · archive (12 posts, 2025-09 → 2026-07)
• 2026 columns: Project Syndicate columnist page (5 columns Apr–Sep 2026) · "This Energy Shock Demands a Green Industrial Strategy" (2026-03-24) · "Making Industrial Strategy Great Again" — Foreign Affairs (early 2026; paywalled, not fetched) · LSE Business Review (2026-06-15; page body would not load)
• The book: Penguin / Allen Lane — UK edition, 2026-06-04, 352 pp · Hachette / Basic Venture — US edition, 2026-09-08 · Democracy Now interview (2026-09-15) · Rockefeller Foundation profile "A New Compass" (2026-05-22)
• Reviews / critics: Atkinson, Policy Arena (2025-07-18) · Thomas, Independent Review (Spring 2026) · Donohoe, Irish Times (2026-07-18) · O'Neill, SSIR (Fall 2026)
• Roles / honours: marianamazzucato.com/about (role list, undated) · SA Presidency — G20 appointment (2025-04-07) · SA hands G20 presidency to the US (2025-11-25) · Whitechapel Gallery press release (April 2026) · WHO Council on the Economics of Health for All (est. 2020-11-13; final report 2023-05-23) · Global Commission on the Economics of Water (final report Oct 2024) · 2025 Birthday Honours — CBE entry (Wikipedia transcription) · Wikipedia (secondary; last edited 2026-09-15) · City AM interview (2024-07-17) · HKS on CHIPS conditionalities, citing Mazzucato–Rodrik (2024-02-21) · Jubilee Report launch (2025-06-20)
• Incentive map: IIPP–Open Society partnership (2020-12) · IIPP–Rockefeller renewal (2022-04) · IIPP–Laudes (2021-05) · Chartwell Speakers listing · EuroStack report launch (2025-02)
• Grading evidence: Intel 9.9% stake, $8.9bn, CHIPS grant conversion (2025-08-22) · Genesis Mission EO (2025-11-24) · UK Modern Industrial Strategy, HCWS725 (2025-06-23; page not fetched — 403) · Institute for Government explainer on mission-driven government (2024-03-05) · Burnham confirmed Labour leader (2026-07-17) · Stanford AI Index 2026 — 12 takeaways (2026-04-13) · AI Index 2026 — Economy chapter · NVIDIA Q4/FY2026 results (2026-02-25) · Bradford, Capgemini (2026)
• Books: The Entrepreneurial State (2013, 10th anniversary ed. 2023); The Value of Everything (2018); Mission Economy (2021); The Big Con (2023, with Rosie Collington); The Common Good Economy (2026)
• Institutional: UCL IIPP · marianamazzucato.com · "Mission-Oriented Country Platforms" WP 2025-20 (posted 2025-11) · "A Mission-Aligned World Bank" (posted 2026-04)

migrated v1→v2 2026-09-21 · backup: _archives/mariana-mazzucato.html.bak-20260921 · converts-from: people/mariana-mazzucato.md · AI & Society domain