The Gray Files

Could the AI investment boom outpace the profits it is expected to generate? Erika Barker examines Michael Burry’s AI bubble thesis through capital spending, chip depreciation and stock valuations. This January 2026 episode explores the financial assumptions behind data centers and the companies building them. AI capital expenditure and return on invested capital Server useful lives, depreciation and reported earnings Data center electricity demand and the risk of stranded assets Investment relationships among Nvidia, OpenAI and Microsoft Valuation, stock-based compensation and passive investing The discussion presents a bearish investment thesis and scenarios, not established findings of fraud or a prediction of market timing. Company relationships and accounting choices require assessment on their own evidence. This episode is commentary and education, not personalized investment advice. Explore Erika’s research and writing at ErikaBarker.ai.

What can your digital footprint reveal, and who gets to interpret it? Erika Barker explores digital surveillance, data brokers and the use of algorithms to build profiles from online activity. The conversation connects data privacy with the systems used by institutions and businesses to assess people. Topics include social media monitoring, facial recognition, opaque risk scoring and the limits of deleting information online. The episode discusses Palantir, Clearview AI and LexisNexis as part of a broader examination of the data economy. It also considers the right to be forgotten and the tensions between privacy, accountability and access to information. Originally published October 2025. For more research and writing, visit ErikaBarker.ai.

How do political messages become part of someone’s identity? In the final installment of the Attention Wars series, Erika Barker explores political polarization, persuasion and the human need to belong. The episode considers how language and group loyalty can make disagreement feel personal. A narrative about Charlie and Rachel frames the discussion of conflict within relationships. The broader conversation examines political messaging, media incentives and the costs of treating people as opposing teams. It asks how listeners can recognize persuasive techniques and examine their own assumptions. Research and writing: ErikaBarker.ai.

How do attention, emotion and context shape what we believe? Erika Barker continues the Attention Wars series with a discussion of persuasion, memory and social media. The episode uses Derren Brown’s demonstrations as a starting point for exploring influence and the stories people tell about it. Topics include environmental priming, emotional associations, marketing and claims made about neurolinguistic programming. Demonstrations and contested theories should not be treated as proof that platforms can reliably rewrite a person’s mind. The conversation asks how to recognize persuasion and think more critically about the messages competing for attention. Research and writing: ErikaBarker.ai.

Why does online disagreement so often become personal? Erika Barker opens the three-part Attention Wars series with an examination of recommendation algorithms, the attention economy and political polarization. The episode connects today’s social media incentives with the longer history of propaganda and disinformation. A narrative about twins Charlie and Rachel introduces the effects of conflict on family relationships. The discussion explores confirmation bias, outrage, echo chambers and how platforms learn from engagement. It considers these as interacting influences on behavior, rather than a single explanation for every person’s beliefs. Research and writing: ErikaBarker.ai.

When does concentrating power improve a system, and when does distributing it make the system stronger? Erika Barker explores centralization and decentralization across technology, economics and institutions. Examples from history and biology frame a discussion of efficiency, resilience and accountability. The episode examines internet platforms, corporate concentration and the promises of Bitcoin, Ethereum and decentralized organizations. It considers why systems that begin with distributed control can develop new concentrations of power. The discussion closes with questions about AI, work and the design of institutions. Research and writing: ErikaBarker.ai.

What separates a data scientist from a machine learning engineer, and where does prompt engineering fit? Erika Barker concludes the data science series with a look at changing roles in the AI industry. The episode connects technical work with the judgment required to use models responsibly. Topics include deepfakes, misinformation, digital surveillance and the consequences of delegating decisions to algorithms. The discussion considers where automation can help and where human interpretation remains essential. It offers a perspective on a changing profession rather than a promise about any particular career path. Research and writing: ErikaBarker.ai.

How do machine learning models turn data into predictions? Erika Barker introduces common algorithms through examples from fraud detection, recommendation systems and risk assessment. This episode is a starting point for listeners who want to understand the methods behind AI. Linear and logistic regression Decision trees and random forests Support vector machines and kernel methods Clustering with k-means and DBSCAN Dimensionality reduction with PCA and t-SNE Neural networks and deep learning Overfitting and the bias-variance trade-off The examples illustrate how different methods approach different problems. Real systems may combine several techniques and require validation on appropriate data. The discussion focuses on concepts rather than a coding tutorial. Research and writing: ErikaBarker.ai.

How did statistical analysis become part of everyday decision-making? Erika Barker opens a three-part data science series with a history of the field and its influence on institutions. The episode connects early statistical thinking with today’s models, recommendations and risk scores. Topics include John Graunt, Thomas Bayes, Francis Galton, the legacy of eugenics and the use of measurement in systems of discrimination. The discussion then turns to Moneyball, commercial analytics and Palantir. It asks who designs algorithmic systems, how they are used and what accountability should look like. Research and writing: ErikaBarker.ai.

How can simple rules produce surprising behavior? Erika Barker introduces emergence and complex adaptive systems through examples from AI, biology, online worlds and financial markets. The episode examines why a system’s behavior can be difficult to predict from its individual parts. Examples include flocking, ant colonies, the World of Warcraft Corrupted Blood event, multi-agent learning and the 2010 flash crash. A discussion of the MegaSyn experiment raises questions about dual-use AI research; computer-generated candidates should not be confused with manufactured or experimentally verified chemical agents. The broader conversation considers feedback, adaptation and unintended consequences. The episode asks what these ideas mean for system design. It explores small experiments, monitoring and the limits of prediction. These examples offer ways to think about complexity rather than a universal formula for controlling it. Research and writing: ErikaBarker.ai.