MARCO MASCORRO

Behavioral Archetype

THE ENVIRONMENT FUNDER ON THE EVALUATOR’S ADVISORY LIST — Subject is a roboticist turned venture investor who, at Andreessen Horowitz, with bylines running from August 2024 to March 2026, put the firm’s money behind a coding editor, an image-model lab, a voice-model lab and a company that sells reinforcement-learning environments to the frontier labs. He also sits on the adviser list of METR, the nonprofit that tests those labs’ models before release, and he is a named backer of Mechanize, the second company in this file that sells training environments and evals. None of this is concealed. METR lists him, Mechanize lists him, and a16z published his byline on every deal. The structural point is that the same short list of people sits on the testing side, the training-environment side and the cap tables of both. This profile scores that position, not the man.

His own thesis is in print. In the Deeptune announcement he co-wrote, the firm says: “the next decade will be mostly driven by better environments.” The nonprofit whose adviser list he is on measures what models trained on those environments can do. Neither fact is a conflict under the rule METR has published, and the page does not say it is. It is the overlap.

Essence Indicators

• Adviser, METR. METR’s about page lists “Marco Mascorro” under “Advisors” with the title “Advisor,” beside Adam Gleave and Rajiv Dattani (“Advisor and Board Member”), Alec Radford, Chase Hasbrouck and Yoshua Bengio (METR, About). His team page: “Marco became an advisor to METR in April 2025.” (METR, team)

• Current role, in METR’s words. “He is currently an independent researcher and entrepreneur. Previously he was a Partner at Andreessen Horowitz, where he led investments in AI and software infrastructure companies.” (METR, team) The page names no firm or company for the entrepreneurial work and none is asserted.

• Investments METR lists. “Some of his investments include Cursor, Black Forest Labs, Deeptune, and Waveforms AI.” (METR, team). a16z’s own announcements carry his byline on each: Black Forest Labs, “we’re excited to announce our seed investment in Black Forest Labs (BFL),” August 1, 2024 (a16z); Cursor, “we’re thrilled to announce our series A investment in Cursor,” August 22, 2024, signed with Matt Bornstein, Rajko Radovanovic and Martin Casado (a16z); WaveForms AI, “we couldn’t be more excited to announce our investment in WaveForms AI,” December 9, 2024, signed with Justine Moore and Martin Casado (a16z).

• Deeptune (March 19, 2026). a16z’s post, bylined “Marco Mascorro and Martin Casado”: “we are leading the Series A in Deeptune, a company building leading RL environments for computer-use and code.” It adds that “The company has been working closely with leading AI labs” and that “Evaluation needs to be precise and reproducible.” (a16z)

• Mechanize (April 24, 2026). “We’ve raised $9.1 million at a $500 million post-money valuation from Marco Mascorro, Adam D’Angelo, and Devendra Chaplot.” (Mechanize, press release) The home page lists him among the backers: “Mechanize is backed by Nat Friedman and Daniel Gross, Patrick Collison, Adam D’Angelo, Marco Mascorro, Dwarkesh Patel, Sholto Douglas, Devendra Chaplot, Alex Atallah, and Marcus Abramovitch.” (Mechanize). It describes the product: “We build environments and evals for frontier coding agents.” See Tamay Besiroglu and Ege Erdil, its co-founders.

• Fellow AI. METR’s page: “Prior to joining a16z, he co-founded Fellow AI, an AI and robotics company based in the San Francisco Bay Area,” and that earlier he spent “some time in Japan doing AI and robotics research for the Fukushima nuclear accident” (METR, team).

• The Epoch AI episode (November 24, 2025). a16z’s podcast page lists him first among the contributors with David Owen, Yafah Edelman and Erik Torenberg; the episode description reads “Epoch AI researchers reveal why Anthropic might beat everyone to the first gigawatt datacenter” and “why 10% of current jobs will vanish this decade” (a16z podcast). See Epoch AI, the nonprofit Mechanize’s founders left.

• X. @Mascobot, linked from his METR team page and from the METR about page (METR, team; METR, About).

Social Persona / Impression Management

Immediate impression: The practitioner-investor. A roboticist’s CV, a thesis written in infrastructure terms, and a byline that is almost always one of several.

Energy: Declarative, in the firm’s house register. The Deeptune post opens “Reinforcement learning is becoming both the bottleneck and the unlock for the next generation of AI systems” and closes on “We’re thrilled to partner with Tim and the Deeptune team” (a16z).

Impression management strategy: Credentialed plainness. The posts under his name state a trend, name the founders and say what the firm bought. Nothing on the pages read argues policy, and METR’s page lists the portfolio as plain fact. The disclosure he gives is the ordinary kind: the byline and the adviser line are each on a page the organization controls.

The persona avoids the confessional register of the field’s debate. In this file’s terms it is a funder’s voice, not a commentator’s, and the evidence for what he thinks about evaluation rests on one sentence the firm co-signed, “Evaluation needs to be precise and reproducible” (a16z).

Forensic Archetype Comparison

Psychometric Assessment

Frame: observations of a documented public role from a firm’s co-signed posts and an organization’s profile page; the Evidence cells cite those exhibits and nothing about private character.

Big Five (OCEAN):

Dark Triad (held evidence-bound; the score measures the public role, not character):

MBTI: ENTP-adjacent, inferred from the documented public role only. Rationale: a generalist’s path across robotics, design and finance, and trend-first investment writing that moves between the model layer, the infrastructure layer and the product layer; “adjacent” because the record is bylines and a profile page, not interviews.

Threat Assessment

Score basis. Atk 4 is leverage. a16z’s announcement that it is “leading the Series A in Deeptune, a company building leading RL environments for computer-use and code,” carries his byline, and Mechanize names him among the investors in its April 2026 round (a16z; Mechanize). It is not higher because funding a supplier is not authority over what a deployed model says, who deploys it or a pre-release gate, and METR lists the seat as “Advisor” with no duties stated (METR). Def 5 is insulation: METR’s page states he is “currently an independent researcher and entrepreneur” (METR), so no employer sits over him, and his Mechanize stake is private money (Mechanize); it is not higher because no founder control, statutory tenure or court in the record shields him, and the adviser line is on a page METR edits. Hp 3 is durability: the companies he helped fund continue without him, and the cap-table line at Mechanize persists, but no institution carries his name or depends on him (a16z; Mechanize). Peers on nearby values: Tamay Besiroglu (6/6/7, hazard 62), who co-founded the environment supplier Mascorro backs and held an operating seat; Ege Erdil (6/6/6, hazard 60), same company; Jaan Tallinn (7/10/9, hazard 84.0), the field’s donor of record, the named funder of the S-Process rounds, an allocation role where Mascorro’s is a seat on one supplier’s round; and Marc Andreessen (8/9/9, hazard 92.3), co-founder of the firm Mascorro left, whose own seat is of a different order. Against Sarah Schwettmann (5/6/5, hazard 43) he takes one less attack, one less defense and two less durability.

Alignment Analysis

Stated alignment: Back the infrastructure layer of AI, as an investor; advise METR, which “scientifically measures whether and when AI systems might threaten catastrophic harm to society” (METR).

Observed alignment: Consistent with both statements at once. The Deeptune post says “Evaluation needs to be precise and reproducible,” and a company that sells the environments in which models “learn, practice, and be evaluated at scale” is the one a16z led (a16z).

Gap assessment: No gap in the documents; the finding is the overlap. The record shows two positions held at the same time by the same person, disclosed on separate pages by separate organizations. It does not settle what, if anything, the advisory seat has to do with the portfolio, and for a position like this it never needs to.

Interest: Documented, never motive, never a verdict. The positions are the firm’s and his own as printed over his byline; what adoption would pay is read from the companies’ and METR’s own pages.

1. Position: reinforcement-learning environments are the next critical layer. In the post he co-wrote: “Reinforcement learning is becoming both the bottleneck and the unlock for the next generation of AI systems,” and “the next decade will be mostly driven by better environments” (a16z). What adoption pays: demand for the product of two companies he is tied to, Deeptune, whose Series A his firm led, and Mechanize, which says “We build environments and evals for frontier coding agents” and names him as an April 2026 investor (Mechanize). His best case: models “need structured environments where they can learn to fully control computers,” and the environments “must be realistic, measurable, and adaptable” (a16z). Others who gain: the labs whose models train on them (“The company has been working closely with leading AI labs”), the other named Mechanize backers, and its co-founders (a16z; Mechanize).
2. Position: image models should ship open-weight. The firm’s Black Forest Labs post, which he co-signed: “BFL is committed to continuing the open weights approach” (a16z). What adoption pays: a seed investment in a company whose model is released that way. His best case: the firm’s: the open approach leaves “developers and product engineers” free to build the product layer. Others who gain: every developer who builds on released weights, and the firm’s other portfolio companies at the application layer.

Alliance Map

These run at the same time. One person is on the adviser list of the nonprofit that tests frontier models before release, is named by two companies that sell training environments and evals as an investor or investment signatory, and signed deals for a firm that also led rounds in two public evaluators. Each tie is stated by METR, Mechanize, a16z, Vals AI or Arena on pages they control. The overlap is the finding.

Convergent Drive Classification

Self-preservation: Moved from a partnership to independent work, per METR’s page. Goal preservation: The thesis that environments drive capability is in the posts and in the investments. Resource acquisition: Round lines at two environment suppliers and a standing adviser seat at the evaluator. Self-improvement: From building robots to funding the layer that trains them.

On the record: self-preservation is METR’s line that he “is currently an independent researcher and entrepreneur” (METR); goal preservation is the Deeptune sentence “the next decade will be mostly driven by better environments” (a16z); resource acquisition is the $9.1 million round that names him (Mechanize); self-improvement is the move from “co-founded Fellow AI, an AI and robotics company” to AI-infrastructure investing (METR).

Subject is not an AI system. The drives appear anyway — in the adviser who also funds the environments.

Sources: Marco Mascorro — METR; About — METR; Conflict of interest policy v1.1 — METR; Mechanize raises $9.1M — Mechanize; Mechanize — home; Investing in Deeptune — a16z (Mar 19, 2026); Investing in Cursor — a16z (Aug 22, 2024); Investing in Black Forest Labs — a16z (Aug 1, 2024); Investing in WaveForms AI — a16z (Dec 9, 2024); The 2045 Superintelligence Timeline: Epoch AI’s Data-Driven Forecast — a16z podcast (Nov 24, 2025); Series A — Vals AI; About — Arena; @Mascobot — X.