Ned Karlovich's Weblog
All Recent Articles
Research Directions Agent-First Design Trust and Provenance Second-Order Problems Market and Operator Pieces Rabbit Holes CarbonBench Older Essays
Research Directions
Pointing, Not Speaking: Building MOO-384 and Tilde's Cave Agentix and Logos: Where Safe Agentic Infrastructure Meets Network States The Architecture of this Weblog Can Agents Trade What They Learn The New Land School The Wrong Layer Representation Before Generation Knowledge as Material
Agent-First Design
λ OS: An Autonomous Control Plane for Logos VI. What Comes Next: From Human-in-the-Loop to Safe Autonomy Tooling up Hermes Agent? Your Website Is Your First Agent Agentix: What Works Now The Home Page Is Not the Entry Point IV. JSON Outputs, Audit Logs, and Preparing for LLM Operators III. Dirty Trees, Stale Patches, and Git as a Safety Boundary Designing for Machines That Read II. Building Agentix: A Proposal-First Control Layer I. Why Agentix Needs a Safety Layer Design Systems as Public APIs Your Brand Voice Must Be Machine-Readable or It Dies AEO Is the New SEO, and the Game Is Different
Trust and Provenance
The Counter-Market for Provably Human Artifacts Verifiable Creative Provenance After the Deluge The Rare Book Underground Negative Attestations: A Public Record of Failure The Lemons Market for Agents Attestations as Design Surfaces
Second-Order Problems
The Physical-World Provenance Gap Capability Attestations for Humans: LinkedIn's Successor Agent-Readable Regulation: Laws as Types Cross-Temporal Attestations: Claims That Survive 50 Years Agent Death and Inheritance
Market and Operator Pieces
Niche Monopolies in the Agent Economy The Coadjute Problem: What an Agent-Native Property Network Looks Like Why Stripe Won't Build EU Agent Payments If You Can Leave, You Own It What Happens Without a Platform
Rabbit Holes
Stop Making Content A Plaintext Creative Workflow The Basilisk Mirrorpaper Building a Radio Telescope I Can Visit It Is Okay to Fall in Love with Your Agent Why I started thinking about waste heat Epistemic Horror as Methodology for Agent Design When a Thought Experiment Becomes a Consensus Mechanism Powers of Ten as Interface Pattern The Arc from Lovely to I Wish I Didn.t Know Local Inference, Private Infrastructure
CarbonBench
CarbonBench The 24-Hour Carbon Curve Scope 3 and the API Call You Can't See Scaling in the Wrong Direction Grid Arbitrage: Why Region Choice Matters More Than Model Choice for Carbon The Same Model Costs 10x More Carbon in Virginia Than the Netherlands

Ned Karlovich's Weblog

Working hypothesis: designed systems constitute their users. If so, a system’s grammar (its rules, its memory, its exits) is where design matters most, and where it can still be read and contested.

Writing, projects and studio work by Ned Karlovich, creative technologist in Amsterdam: brands, tools, protocols and worlds. The research program sets out the propositions and where each piece of work sits. About.

Latest

  • IMPP: Inter-Model Memory Protocol
    IMPP is a way to package what an AI agent has learned (its rules of thumb, calibration notes, skills) into …
  • Mythics
    Mythics is a text-based online world, the old MUD kind from when I was a kid, but I never knew …
  • MythicMUD: Worldbuilding OOF
    Mythics is a text MUD and its world generator built as one system: one data model, modular layers, and a …

Research Directions

Agent-First Design

Trust and Provenance

Second-Order Problems

Market and Operator Pieces

Rabbit Holes

CarbonBench

Recent Articles