News and Events
Insights
Ideas from the frontier, coverage in the press, and where to find us next.
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.
Steering the Talking Probability Machine
Meet us at Dreamforce Conference 2026
Article
The 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.
Meet us at Explore DDD Conference
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.
Meet us at MxD Summit 2026
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.
Towards a Science of Intent
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 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.
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.
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.
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.
SCRAT
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.
The AI Identity Gap
The AI Identity Gap
Orchestrating Human-AI Software Delivery
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.
The Hybrid Workforce
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.