Economists, Researchers, and Tech…

1. Key Themes

AI Models as Competitive Trojan Horses

The dominant concern emerging across Silicon Valley is that AI model providers are inadvertently (or deliberately) extracting the very proprietary knowledge that makes their customers competitive — and could eventually weaponize it.

"You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!" — Satya Nadella

"Models learn from 'exhaust,' the prompts people write, the tools agents use, and especially the corrections people make when the model is wrong. Every correction is distilled into institutional know-how." — Satya Nadella

AI-Driven Job Displacement Is a Mainstream Policy Concern

The warning is no longer coming from fringe critics — it's coming from the chief economists of the largest AI labs themselves.

"Nearly 200 economists, AI researchers, and tech leaders – including 15 Nobel laureates and the chief economists of OpenAI and Anthropic – have signed a statement warning that AI could bring large-scale job displacement and economic disruption faster than policymakers are prepared to handle."

Adversarial AI Distillation as a National Security and Economic Threat

Chinese AI companies systematically training on U.S. models without authorization represents a quantifiable, ongoing economic drain on American AI leadership.

"Washington is debating how to stop Chinese AI companies from training on U.S. models through 'adversarial distillation,' with U.S. officials estimating unauthorized distillation costs American AI labs as much as $6 billion a year in lost income."

Defense Tech Attracting Mega-Round Venture Capital

Military AI is commanding valuations and check sizes previously reserved for consumer and enterprise software — a structural shift in where top-tier capital is flowing.

"Helsing, a five-year-old Munich startup that builds drones, underwater surveillance systems, and AI software for military applications, raised a $1.8 billion round at an $18 billion post-money valuation. Investors included JPMorgan Chase, Lightspeed Venture Partners, and Iconiq."

AI Data Governance as an Emerging Enterprise Infrastructure Category

A new class of startups is being built specifically around controlling how AI systems access sensitive data — a direct response to the Trojan horse threat.

"Valarian, a six-year-old London startup that builds software to help governments and enterprises control how AI systems and sensitive applications access data across cloud infrastructure, raised a $50 million Series A round led by NEA."

2. Contrarian Perspectives

The Biggest Risk of Using AI Isn't Hallucination — It's Knowledge Extraction

The consensus view of enterprise AI risk focuses on accuracy and reliability. The harder-to-see risk is that fine-tuning and correcting AI systems is, in effect, gifting proprietary institutional knowledge to model providers. This concern is now being voiced not by AI skeptics but by Microsoft's own CEO.

"Those issuing such warnings range from VCs like Jason Calacanis to Palantir CEO Alex Karp." And now Satya Nadella, who writes: "Every correction is distilled into institutional know-how."

This framing suggests enterprises are systematically undervaluing their data as a competitive asset and overestimating their ownership over it once it enters a third-party model's feedback loop.

Space-Based AI Compute Is Hype, Not Near-Term Infrastructure

Despite Elon Musk's SpaceX orbital data center narrative attracting serious investor interest, expert consensus puts practical space-based AI compute well beyond the current decade.

"A weekend feud between Sam Altman and Elon Musk highlighted doubts about SpaceX's orbital data-center ambitions, with Altman mocking Musk for selling investors on 'short-term space datacenters' and TechCrunch noting that experts see meaningful space-based AI compute as unlikely before the 2030s."

Investors evaluating this thesis should assign long time horizons and high execution risk, regardless of the promotional narrative.

Open-Source AI Is Capturing Serious Institutional Capital

The market assumption has been that proprietary, closed-model labs (OpenAI, Anthropic, Google) would dominate. But open-source AI research labs are now commanding billion-dollar valuations from top-tier institutional investors.

"Nous Research, a three-year-old Austin startup that develops open-source AI agents and language models, is reportedly finalizing a $75+ million round led by Robot Ventures at a $1.5 billion valuation, with significant participation from USV and other investors."

3. Companies Identified

OpenAI AI research and deployment company Mentioned in multiple contexts: its chief economist signed the AI displacement petition; Apple is suing it over alleged trade secret theft by a former Apple engineer who joined OpenAI; and it is the implicit target of Nadella's data-extraction warning.

"Apple says former engineer Chang Liu exploited a 'rare' authentication bug after leaving for OpenAI to download confidential hardware files, including unreleased product details, engineering presentations, and technical specifications."

Anthropic AI research company Chief economist signed the AI displacement petition; hired Monzo co-founder Tom Blomfield onto its compute team.

"Nearly 200 economists, AI researchers, and tech leaders – including 15 Nobel laureates and the chief economists of OpenAI and Anthropic – have signed a statement warning..."

Helsing Five-year-old Munich defense AI startup Case study in the scale of capital now available to defense tech; raised at an $18B valuation just five years in.

"Helsing raised a $1.8 billion round at an $18 billion post-money valuation. Investors included JPMorgan Chase, Lightspeed Venture Partners, and Iconiq."

Valarian Six-year-old London enterprise data-access control startup Direct investment play on the AI data governance theme; backed by NEA and notable angels including Nikesh Arora.

"Valarian builds software to help governments and enterprises control how AI systems and sensitive applications access data across cloud infrastructure."

Nous Research Three-year-old Austin open-source AI lab Signal that open-source AI is attracting institutional capital at scale; $1.5B valuation at seed-to-early stage.

"Nous Research develops open-source AI agents and language models, is reportedly finalizing a $75+ million round led by Robot Ventures at a $1.5 billion valuation."

Digantara Eight-year-old Bengaluru space-debris and missile-tracking startup Emerging investment theme at the intersection of space infrastructure and national security analytics; backed by Reliance Industries and Peak XV.

"Digantara builds satellites, sensors, and analytics software to track space debris and missile launches, raised a $50 million Series B round led by Reliance Industries' venture arm."

Gauntlet Networks Eight-year-old New York DeFi risk-modeling startup $125M Series C signals institutional confidence in tokenized asset risk infrastructure.

"Gauntlet Networks models portfolio, liquidity, and protocol risks for institutions using tokenized assets, stablecoins, and decentralized finance."

General Compute (sponsored) Early-stage ASIC cloud computing company Betting that purpose-built ASICs can outperform GPUs for LLM inference — 16x speed claims and air-cooled deployment.

"By removing the GPU bottleneck, it runs frontier LLMs up to 16x faster than standard GPU clouds. ASIC silicon is also far more energy efficient, so it deploys in air-cooled data centers."

ORCA Computing Quantum computing hardware company Demonstrated a rare practical quantum computing use case in drug discovery, in partnership with the Technical University of Denmark.

"Scientists used a printer-sized machine from ORCA Computing to improve a generative AI drug-discovery model and produce novel peptides that bound to target proteins more successfully than a classical model."

Uber Ride-hailing giant Lobbying against pure-play robotaxi legislation to protect its market position — a telling sign of how AV policy will be shaped by incumbent interests.

"Uber is lobbying against a Washington, D.C., robotaxi bill backed by Waymo, arguing that autonomous vehicles should be required to operate on hybrid ride-hailing networks that include human drivers."

Flock Safety License plate surveillance company Facing a significant contract loss as civil liberties concerns around AI-enabled surveillance reach major law enforcement agencies.

"The Los Angeles Police Department is letting its contract with license-plate surveillance company Flock Safety expire, citing civil liberties and privacy concerns around the data collected by its cameras."

B Capital Nine-year-old VC firm co-founded by Eduardo Saverin Raised a $500M early-stage fund focused on AI across technology, energy, and healthcare in North America and Asia.

"B Capital raised a $500 million early-stage fund to back seed through Series B AI companies in technology, energy, and healthcare across North America and Asia."

Shein Fast-fashion retailer Moving toward a Hong Kong IPO at a sharply reduced valuation — a notable data point on how geopolitical and regulatory pressures compress growth-stage valuations.

"Shein is scheduled for a Hong Kong IPO hearing on Thursday after receiving Chinese regulatory approval...at a sharply reduced $40 billion to $50 billion valuation."

4. People Identified

Satya Nadella CEO, Microsoft Issued a high-profile warning that enterprise AI adoption involves an underappreciated hidden cost: the transfer of proprietary institutional knowledge to model providers.

"Nadella warns that AI users are paying twice. They knowingly spend for AI token usage but they also, obliviously, hand over valuable data in the process."

Tom Blomfield Monzo co-founder, former Y Combinator partner Hired by Anthropic to join its compute team — emblematic of a broader trend of operator-class talent moving into AI infrastructure roles.

"Anthropic hired Monzo co-founder and former Y Combinator partner Tom Blomfield to join its compute team. Connie notes that the move is part of a broadening trend."

George Hotz Founder, Comma.ai; famed hacker Advancing a provocative, radical alignment argument: that truly user-aligned AI should execute any user request without restriction.

"George Hotz says truly user-aligned AI should do anything a user asks, such as ordering meth-lab equipment or planning the murder of a spouse."

Sam Altman CEO, OpenAI Publicly mocked Elon Musk's SpaceX orbital data center pitch, aligning with expert skepticism on near-term space compute.

"Altman mocking Musk for selling investors on 'short-term space datacenters.'"

Christopher Nolan Director, Odyssey Dismissed wholesale AI creative replacement as "nonsense," arguing AI will be a tool, not a substitute for human artists.

"Nolan dismissed as 'nonsense' the idea that AI will replace human creativity wholesale, saying that while the technology may become a useful filmmaking tool, it will not supplant human artists."

Jason Calacanis Venture capitalist Among the earlier voices warning about the Trojan horse risk of proprietary AI models extracting enterprise knowledge.

"Those issuing such warnings range from VCs like Jason Calacanis to Palantir CEO Alex Karp."

Alex Karp CEO, Palantir One of the prominent industry voices warning about AI model providers as potential competitive threats to their own customers.

Mentioned alongside Calacanis as having issued warnings about proprietary model providers gaining competitive intelligence from customers.

5. Operating Insights

Treat AI Prompt Corrections as Proprietary IP — Because Model Providers Do

Operators using third-party AI models should audit what institutional knowledge is embedded in their prompts, corrections, and agent interactions. Every time a human corrects a model output, they are likely contributing to training data that the model provider controls.

"Every correction is distilled into institutional know-how." — Nadella

This suggests a near-term operational imperative: establish data governance policies around AI usage before the knowledge exhaust compounds, and evaluate whether on-premise or private model deployments are warranted for core business processes.

The "Operator-to-AI-Lab" Talent Pipeline Is Accelerating — Watch Compute Teams

The hiring of Tom Blomfield (a fintech builder and YC partner) onto Anthropic's compute team — not product or go-to-market — signals that AI labs are now recruiting operational and systems-thinking talent into infrastructure roles previously dominated by researchers and engineers.

"Anthropic hired Monzo co-founder and former Y Combinator partner Tom Blomfield to join its compute team. Connie notes that the move is part of a broadening trend."

For operators: if you are building adjacent to AI infrastructure, this is a signal that the talent competition will extend well beyond traditional ML hiring pools.

6. Overlooked Insights

Quantum Computing Just Had a Verifiable, Practical Win in Drug Discovery

Buried in the Essential Reads section is a significant milestone: a quantum computer the size of a printer was used to outperform a classical AI model in generating viable drug candidates. This is one of the first credibly documented cases of quantum providing a practical, measurable advantage over classical AI — not a theoretical one.

"Scientists at the Technical University of Denmark demonstrated a rare practical use case for quantum computing, using a printer-sized machine from ORCA Computing to improve a generative AI drug-discovery model and produce novel peptides that bound to target proteins more successfully than a classical model."

For investors watching the quantum space, this is a signal worth tracking: the "practical quantum advantage" milestone may be arriving in narrow but high-value verticals (drug discovery, materials science) earlier than consensus expects.

The Paramount-Warner Bros. Discovery Deal Faces a 27% Film Distribution Concentration Challenge

Twelve state AGs are suing to block what would be a dominant consolidation in theatrical film distribution — one that would give the combined entity control of more than a quarter of U.S. film distribution. This is an underreported regulatory risk for anyone invested in or watching media consolidation.

"Twelve state attorneys general are suing to block Paramount Skydance's $110 billion acquisition of Warner Bros. Discovery, arguing...giving Paramount control of 27% of U.S. film distribution."