Blog

Why Frontier Models Stay Spiky

Frontier models stay spiky because useful work is contextual and counterfactual paths are ungradable. Post-training quality is a grader problem, not a pretraining-size problem.

Blog

Why Frontier Models Stay Spiky

Can You Fire the Generalist?

Auditability Is Not Logging: What Agent Compliance Will Actually Cost You

Agent Security as Its Own Product: One Reading, Held Loosely

The Fair Use Fight Everyone Is Watching Is Not the One That Matters

Agent-First Tooling Is Not About Agents

Invert the AI Scaling Consensus: Four Constraints, One Real Fragility

Open Weights Hit Production Parity. Reliability Is Now the Harder Thing to Buy.

World Models Relocate Hallucination Rather Than Removing It

Codex Security and the Triage Problem

Self-Modifying Agents Are Version Control With Better Manners

Singapore Is Pointing AI Governance at Runtime, Not Weights

LeCun Says the LLM Revolution Is Over. The Verification Argument Is the Part Worth Arguing With.

YouTube's AI Rollout: The Label May Outlast the Toolkit

Origin Over Domicile: Why China Unwound the $2B Manus Acquisition

The Bottleneck Moved: Intelligence Is Cheap, Coordination Is Not

The Federal AI Law You're Waiting For Isn't Coming. State Enforcement Already Did.

The Cheating Crisis Is a Design Failure, Not an Enforcement Problem

27 Seconds to Breakout: Why Identity Is the Only Perimeter Left

Self-Improving Agent Skills Are a Version Control Problem, Not an Intelligence Problem

The Super App Is an Anti-Pattern: Why OpenAI's Consolidation Play Sells Modularity for Everyone Else

The Agent Bottleneck Is Memory, Not Intelligence

Agent Optimization: The SEO Discipline Nobody Is Practicing Yet

The Federal AI Law You're Waiting For Is Not Coming First. State Enforcers Are Already Here.

Anthropic's New Institute Is Not a Safety Team. It Might Be a Sales Team.

When the Model Is Free, the Hard Part Gets Harder

The Runtime Is Becoming the Product

The Money Is Going to the Engine. The Bottleneck Is the Steering Wheel.

The IDE Just Got Demoted

Stop Growing the Context Window. Start Delegating It.

The Moat Moved to the Runtime

How Paywalls Hand the Training Set to State Media

AI Disclosure Is Becoming the Baseline. Human-Only Is the Next Premium.

The Deepfake Defense Is Actually an Identity Gate

Infrastructure Is the New Moat

When Safety Moves Inside the Runtime

Treat Vision APIs as Plumbing, Not a Moat

The Agent Orchestrator Is Not the Moat

The Convenience Loop: Why AI Assistants Are Picking the Next Default Language

The Number That Matters in Nemotron 3 Super Is Not 120B

World Models Are a Bet Against LLMs, Not Proof They Win

OpenClaw Is Becoming a Platform, and That Is the Risk

The Sovereign Tier: Why Cheap AGI Will Not Save Your Margins

The Worker-Model Tier Is the Real Product

The Durable Moat in Agent Control Planes

One Throat to Choke Means No Independent Witness

The Agent That Can Stop Is the Agent You Can Trust

State AI Rules Are Becoming a Compliance Problem Before Federal Law Arrives

The Coding Agent Boom Still Needs a Steering Wheel

The GPU Cycles You Already Paid For Are Training Your Next Model

Shadow A2A: Why Local Agents Can Look Like an Intrusion

Singapore Is Moving Agent Governance From Debate to Implementation

The Zero-Human Company Still Has Humans Somewhere

Stop Summarizing Your Context. Start Querying It.

Stop Using Claude as a Search Engine

NVIDIA Is Arming Its Own Customers

Your AI Team Should Be a Shadow Company

Security Is Becoming an Agent Capability, Not a Wrapper

Enterprise AI Optimizes for the Median, and That's the Trap

The Cage Is the Point: When Algorithmic Management Becomes Political

PDF Accessibility Is the Trojan Horse for Your RAG Pipeline

The Half-Billion Dollar Clock in Defense AI

The Steering Wheel Is the Moat

Durov's Inverse: Team Velocity in the AI Era

The First to Submit Writes the Rules

The Creative Is the New Targeting

The Filing Cabinet Fallacy

The Number Cannot Be Accused of Bias

The Median Is the Trap

The Thin Layer Is the Point

The Middle Tier Is the Target

The Three-Month Gap Is the Moat

The reliability premium

APAC is writing the rules while the West debates timelines

Agent security is an identity problem, not a safety problem

Export Controls End at the API

Parameters Are a Commodity. Tool Access Isn't.

AI Tutors Need a Calculator

A2A Is the Next Shadow IT