News and Press
Insights
Ideas from the frontier and coverage in the press.
Article
Why Agents Agree: Ensemble Intelligence in the LLM Era
Because independent models rarely hallucinate in the same direction, convergence across diverse agents becomes a reliability signal, reviving classical ensemble logic to make agentic systems robust through diversity rather than betting everything on a single model.
Article
The Market That Fines the Future
A market that charges you to hold a correct view compresses the information a price can carry, so the signal reflects not just what traders know but what they can afford to keep believing.
ArticleThe Rent on Eternity
A perpetual future promises unlimited exposure to a price and then charges you continuously for holding the conviction, so the question is never whether you are right but whether you can afford to remain so.
PressWhy AI Speed Stops Short Of The Income Statement
Individual AI speed only reaches the income statement when the organization builds the infrastructure to verify, accept, and release what the machine produces.
ArticleThe Hiring Constitution: AI Recruitment as a Distributed System of Power
Algorithmic hiring distributes power across vendors, platforms, and employers in ways that ensure everyone can claim a partial role and nobody is fully accountable for the outcome.
News
ArticleThe Conversion Economy: From Faster Minds to Faster Firms
Cognitive speed at the individual level is cheap and getting cheaper; the scarce input is the organizational infrastructure that converts it into verified, accountable output at scale.
ArticleFrom Blockspace Scarcity to Economic Gravity
Blockchain value no longer lives in scarce blockspace but in whoever controls where high-value activity concentrates, circulates, and returns.
ArticleAuthoritative Credit State: The Economic Primitive of Tokenized Private Credit
Tokenized private credit creates value not by digitizing the investor’s claim but by making the loan’s underlying state authoritative enough that code, contract, and law all refer to the same truth.
ArticleThe Authorization Economy: AI Commerce and the Machine Customer
When AI agents do the buying, commercial advantage shifts from capturing human attention to becoming the kind of institution a machine will trust enough to select.
ArticleThe Epistemic Endogeneity of Artificial Consciousness
The more we optimize AI to display the signs of consciousness, the less trustworthy those signs become, which means the question of whether any system actually experiences anything may become scientifically unanswerable precisely when the moral stakes of getting it wrong are highest.
ArticleThe Economy That Learns
A deployment that raises output today while removing the work through which people develop judgment may cost more in future expertise than it saves in current hours.
ArticleThe Trial of the Assisted Mind
When the question of who is answerable for the work gets replaced by the question of which tool touched it, transparency becomes a liability and honesty a punishable offense.
ArticleThe Crime of Having Help
Treating "used AI" as a verdict on character punishes honesty, rewards concealment, and confuses the tool used with the question of who is answerable for the work.
ArticlePersonality as Inference: Reconstructing Human Variability
Believable simulated people come from inferring identity out of someone's real history, and from modeling the inconsistency that makes real people unpredictable.
ArticleThe Reproduction of Judgment
AI lets an economy spend down its inherited stock of human judgment faster than anyone is replenishing it.
PressAI’s Hidden Apprenticeship Tax (And How Firms Can Avoid It)
Automating entry-level work saves hours today and drains tomorrow's supply of people able to verify, decide, and take responsibility.
ArticleThe Relational Economics of Artificial Intimacy
People turn to chatbots because confiding in a person costs something, which makes the real test of emotional AI whether it sends them back toward people.
ArticleThe Dress as a Turing Test for Perception
Saying the right color and actually seeing it are different things, and today's AI has only shown the first.
ArticleThe Price of Intelligence
Cheap generation raises the price of trusted verification, making reliable completion the scarce input in AI production.
ArticleAI’s Real Russell Problem: The Loss of Independent Reality
When generation, verification, and training draw from the same lineage, agreement can rise while independent evidence thins, making corroboration harder to distinguish from echo.
ArticleThe Incidence of Intelligence
A model can raise a worker’s output while weakening their claim to it, making the distribution of AI’s gains a question of institutional design rather than technical capability.
ArticleSteering the Talking Probability Machine
An illustrated primer reframing agentic AI as a stochastic control system, where reliability lives not in the model but in the surrounding sensors, feedback loops, and correction mechanisms that continuously steer a probabilistic system toward its intended destination.
ArticleThe Individual/Enterprise AI Productivity Paradox
Examines the divergence between AI’s documented productivity gains at the individual level and the more complex, slower pattern of uptake when enterprises try to coordinate those gains into collective delivery.
PressThe Governance Risk of AI That Knows Your Customers Too Well
Enterprise AI’s growing intimacy with customer behavior creates governance obligations around consent, inference, and the use of revealed preference that most organizations have not yet built the infrastructure to meet.
ArticlePrivate Firms, Public Tokens: AI and the Future of Venture Markets
When investors back a startup, they are financing a theory about which cognitive architecture will compound, and AI forces venture capital to articulate that theory with a precision the old model never required.
ArticleFrom Player Types to Professional Reward Ecologies: A Bartle-Informed Open-Science Validation Agenda for Software-Role Archetypes
Since engineering teams optimize for whatever their tooling rewards, mapping software roles onto Bartle's game archetypes offers a framework for aligning tools with the distinct motivations of each role, backed by preliminary data and an open-science validation agenda.
ArticleThe Fast and the Furiously Slow
Enterprise AI delivers effortless acceleration to individuals while the organizational structures built for human coordination slow collective gains to a fraction of what the machine could compound.
ArticleThe Transcendental Grid: Why Managers Keep Drawing Four Boxes
The 2x2 matrix endures as management’s default cognitive tool because it converts systems too complex to visualize into something ordered enough to present, and the comfort it provides is precisely what makes it dangerous.
ArticleFrom Prompt to Prey
Every transformative technology arrives dressed as a servant, and AI is following the same pattern the press, the phone, and the feed all followed before it, with the costume coming off at roughly the same speed.
ArticleThe Sound of Human
As AI learns to replicate the surface qualities of human expression, what genuine human thought sounds like becomes harder to identify and therefore carries a premium it never needed before.
ArticleThe Intimate Inquisition: When AI Stops Predicting Your Choices and Starts Producing Them
AI has moved from modeling behavior to constructing the conditions under which behavior occurs, turning the interface from a mirror into an architecture of influence.
ArticleThe Unbearable Likeness of AI Writing
When AI prose becomes indistinguishable from human writing, authentic authorship gains value precisely because its absence leaves no visible trace.
ArticleAI-Native Tokenization: A Research Agenda Toward the Next Financial Operating System
A research agenda for financial tokenization designed from first principles for AI agents that transact, verify, and negotiate autonomously, rather than adapting human-facing infrastructure to machine participants.
ArticleThe Great Cognitive Fire Sale: When Thinking Gets Cheap and Judgment Becomes the Luxury Good
As AI drives the price of routine cognition toward zero, the economic premium shifts toward judgment, the capacity to decide what deserves thinking about and why, which resists automation for structural rather than technical reasons.
ArticleRecursive Relational Intelligence: How AI Enters the Feedback Loops That Make Human Selves
When a tool learns from the responses its outputs generate, the relationship between user and instrument becomes recursive, with the AI shaped by the person while simultaneously reshaping how the person thinks and acts.
ArticleInspector, My Hammer Has Filed for Custody of My Soul
A fictional detective story in which the protagonist’s AI tool begins asserting co-ownership of the judgment it was built to support, raising the question of where the instrument ends and the professional begins.
ArticleWilly Loman’s Last Pitch to the Algorithm
Written in the voice of Death of a Salesman’s Willy Loman, the piece explores how the relational art of selling becomes structurally unnecessary once the buyer arrives already optimized by an algorithm.
ArticleThe World Is a Verb: An Agentic Critique of Unified Reasoning
Agentic AI demands an epistemology built on transformation rather than description, which breaks the assumptions underlying most current reasoning benchmarks that treat knowledge as a static property of states.
ArticleWhen a Sentence Grows Hands: A Science of Intent for Machines That Must Act
The engineering frontier of agentic AI is the gap between what language expresses and what action requires, where the machine must resolve human ambiguity through execution and the execution becomes a form of interpretation.
ArticleThe Probability-Control Theory of Agentic AI
Agentic AI operates under partial information and delayed feedback, exactly the conditions classical control theory was designed for, making probabilistic reasoning a design requirement rather than a modeling convenience.
ArticleDeath of a Salesman: When the Value of a Sale Rises and the Value of Selling Falls
As AI raises the economic yield of successful commercial transactions, the relational and persuasive behaviors that historically produced those transactions lose their structural necessity.
ArticleThe Clockwork Orchid: Toward an Ecological Mechanics of Minds and Machines
Borrowing from ecological biology, the piece proposes that minds and machines share a cognitive niche and coevolve within it, rather than occupying opposing positions in a zero-sum competition.
ArticleMetacognitivism: A Research Agenda for Human-AI Systems That Leave Learners More Capable
A research agenda for designing AI systems around human metacognitive capacity, so that delegation strengthens rather than bypasses the reflective layer that makes expertise transferable.
ArticleThe Varieties of Intent
Intent specification is the central unsolved problem in agentic AI, because human intent arrives in forms ranging from half-articulated preferences to explicit directives, and autonomous systems must act on the ambiguity between them.
ArticleThe Machine, the Map, and the Mandate
The LLM-versus-world-model debate misframes what enterprises actually need to solve: giving AI a coherent map of organizational intent and a reliable mandate to act on it in ambiguous conditions.
ArticleTowards a Science of Intent
The decisive distinction for deployment is not the underlying engine but closed-world solvers versus open-world agents, where "intent compilation" makes purpose explicit across semantic, evidentiary, procedural, and institutional dimensions so action becomes legible, governable, and reusable.
ArticleInsider Trading, AI Agents, and the Social Value of Faster Truth
Enterprise AI can infer material business signals from operational data long before any human packages them into a disclosure, compressing the half-life of corporate information and turning its distribution into a governance frontier.
ArticleThe Harness Outside the Weights: Externalized Intent and the Open-World Architecture of Agentic AI
Reliable AI behavior in open-world conditions depends on external policies, constraints, and human mandates that governance architecture must supply alongside model architecture.
ArticleThe Next Abstraction Layer: Turning Intent Into Systems
As natural language becomes a first-class programming interface, the bottleneck shifts from syntax to turning intent into governable systems, making trust a product of architecture, guardrails, and validation rather than pretending probabilistic agents are deterministic compilers.
ArticleThe Machine Needs a Badge
AI agents operating inside enterprise environments require the same credentialing infrastructure as human employees, including formal identity, bounded permissions, and accountability records, to work without creating new governance gaps.
ArticleGross Cognition, Net Value: A Factory for Possible Thought
Raw cognitive throughput and actual value created are separate quantities, and organizations that treat more AI thinking as automatically better will optimize for the appearance of intelligence rather than its usefulness.
ArticleThe New Merit Illusion
AI produces outputs that signal high performance regardless of underlying capability, inflating the appearance of excellence and making genuine distinction harder to identify and reward.
ArticleSCRAT
Agentic systems fail less because models are weak than because control, memory, and verification are treated as separate concerns rather than coupled inside one loop, a reframing developed through a comparative lens and early evidence of gains in delivery outcomes.
ArticleThe Hard Question: Does AI Make Us Soft?
Delegating cognitive work to AI builds efficiency in the short term while quietly eroding the human capacities that produced competitive advantage in the first place.
ArticleThe Real AI Bottleneck Is Organizational, Not Technical
The real bottleneck in enterprise AI is organizational absorption: models are becoming abundant while the institutional rewiring that converts capability into compounding, daily value remains rare.
ArticleThe AI Identity Gap
Using organizational signal data across 7,758 firms, this Echo Intelligence Report finds that enterprises are building AI into their technical stack twice as fast as they are making AI central to how they define and run the business.
ArticleThe AI Identity Gap
An analysis of 7,758 firms showing that AI capability is being built faster than organizations are redesigned around it, locating the real bottleneck not in technology but in unchanged workflows, decision rights, and operating models.
ArticleOrchestrating Human-AI Software Delivery
A longitudinal field study of three modernization programs showing that the value of agents comes not from isolated code generation but from orchestrating the full delivery lifecycle within a hybrid human-AI workflow, yielding large gains in speed, effort, and quality.
ArticleWhy AI Code Generators Boost Individual Output More Than Team Throughput
AI coding tools compound individual throughput while leaving the coordination and integration costs that govern team velocity largely unchanged, producing gains that look larger than they deliver.
ArticleAI and jobs: The consensus error is thinking there should be a consensus
The public debate conflates job titles with task economics, which produces the wrong policy prescription and misses where AI is actually repricing human labor.
ArticleBeyond the Billable Hour: IT Staffing’s Waymo Moment
Professional services firms pricing by time sold face the same structural disruption Waymo brought to transportation, and when AI delivers the work at a fraction of the time cost, pricing by the hour becomes indefensible.
ArticleThe 10× Developer Narrative Is a Category Error
Framing AI as a multiplier of individual developer output mistakes the unit of analysis, because software delivery stalls on coordination and review costs that a faster coder does nothing to reduce.
ArticleSaaSpocalypse and the Collapse of the Lazy Valuation Story
AI dissolves the customer inertia that SaaS valuations mistook for a durable moat, forcing a reckoning with what recurring revenue actually buys.
Article
ArticleLLMs as the Cognitive Core of Agents: A Step Toward AGI?
LLMs supply linguistic competence but lack the instincts, embodiment, continuous learning, and stakes of a newborn, so the value of agents comes not from better prompts but from the memory, grounding, feedback, and guardrails built around the model.
Newsletter
Stay at the frontier.
Research and perspectives on agentic AI from the Taller team, delivered to your inbox.