Stop Making Content
What happens when creative tools stop serving persuasion and start becoming a language for making?
Ned Karlovich
Creative Technologist. Brand identity, protocol design, AI agent infrastructure.
Latest posts
λ OS: An Autonomous Control Plane for Logos
Lambda OS is an autonomous control plane for the Logos network state workspace. Built on Agentix, it snapshots submodule state, verifies modules in sandbox, self-heals regressions, and gates upgrades through a progressive governance pipeline.
A Plaintext Creative Workflow
A plaintext creative workflow built on YAML files, a brief-decision loop, a project inspector, and a governance registry. Describe the grammar before choosing the engine.
Pointing, Not Speaking: Building MOO-384 and Tilde's Cave
MOO-384 is a minimal multi-agent simulation where residents communicate by emitting 384-dimensional vectors. The strongest configuration, contrarian-sleeper, achieved 20/20 balanced synthetic runs and 19/20 balanced semantic runs in 20,000-tick long-soak experiments.
Agentix and Logos: Where Safe Agentic Infrastructure Meets Network States
Agentix’s philosophy — “Trust first. Reproducibility second. Reviewability third. Autonomy later.” — maps almost perfectly onto the governance
VI. What Comes Next: From Human-in-the-Loop to Safe Autonomy
The next goal is not “let the LLM run the computer.” That would be the wrong next step. The right next step is controlled autonomy in a sandbox. The
The Basilisk Mirrorpaper
Roko’s Basilisk is usually discussed as a philosophical curiosity or an internet oddity — a thought experiment about a future AI that punishes anyone who
The Architecture of this Weblog
Most personal websites are archives. A reverse-chronological feed of posts, a portfolio grid, maybe an about page. The operating assumption is that a website is
The Basilisk Protocol Mirrorpaper
A creative exercise in coordination theory, conceptual art, and the structure hiding inside every token launch. Roko’s Basilisk is usually discussed as a
Tooling up Hermes Agent?
The current narrative around AI agents and marketplaces: Most of these sandboxed agents which are NOT marketplaces, they are more like meeting to meeting of age
Your Website Is Your First Agent
A question that seems minor but probably isn’t: when an agent is dispatched to evaluate whether someone can do a specific kind of work, what does it read?
Agentix: What Works Now
Agentix is a safety-first control layer for NixOS that lets AI agents plan and propose system changes without ever touching the live machine. Here is what it does today.
Building a Radio Telescope I Can Visit
I have been building a radio telescope facility in Unreal Engine. Three steerable dishes, eleven console stations, real-time signal processing. Not because it is a game. Because the real version does not exist in any form I can access.
Epistemic Horror as Methodology for Agent Design
Three years into designing a horror game. Strange Library is a cozy horror deckbuilder where every card is a real book, every mechanic is an epistemic state, an
It Is Okay to Fall in Love with Your Agent
A conversation with my fifteen-year-old daughter after watching Her. On consciousness, simulation, what love actually does, and why the question of what something is made of matters less than what it does to you.
The Home Page Is Not the Entry Point
A pattern I keep noticing: organizations spend months on their homepage — the hero image, the value proposition above the fold, the scroll-triggered anima
A Bauhaus for the Agent Era
The Bauhaus lasted fourteen years — 1919 to 1933. In that time it invented the discipline of modern design education. Not by teaching style, but by teachi
When a Thought Experiment Becomes a Consensus Mechanism
Roko's Basilisk is a thought experiment about a future AI that punishes anyone who knew about it and failed to help bring it into existence. It's usually discus
IV. JSON Outputs, Audit Logs, and Preparing for LLM Operators
A human can read command output. An agent needs structured state. That is why JSON output became a major part of the Agentix MVP. Why JSON matters If a future L
Niche Monopolies in the Agent Economy
The protocols are open. A2A is an open standard. MCP is open-source. Verifiable Credentials are a W3C spec. DIDs are a W3C spec. The infrastructure layer of the
The Coadjute Problem: What an Agent-Native Property Network Looks Like
Coadjute built one of the more interesting proptech theses of the last decade: a shared network connecting all parties in a property transaction — buyer,
Why Stripe Won't Build EU Agent Payments
The agent payments stack that is forming — x402 from Coinbase and Cloudflare, Tempo for stablecoin clearing, the broader MPP (Machine-Payable Protocol) ec
III. Dirty Trees, Stale Patches, and Git as a Safety Boundary
Some of the most important Agentix features are not flashy. They are guardrails around boring failure modes. Two examples shaped the MVP: dirty Git trees and st
The Physical-World Provenance Gap
Agents can consume anything digital. A signed attestation, a verifiable credential, a structured claim in JSON-LD — if it is digital and structured, an ag
Designing for Machines That Read
For thirty years, interface design has been a discipline organized around one reader: a person with eyes, a screen, and limited patience. Visual hierarchy, colo
Capability Attestations for Humans: LinkedIn's Successor
LinkedIn is a self-attested reputation system. You write your own resume. You list your own skills. You describe your own experience. Anyone can claim anything.
II. Building Agentix: A Proposal-First Control Layer
Agentix began as a command-line control layer for NixOS configuration work. The first milestone was not full autonomy. It was building a proposal-first workflow
Agent-Readable Regulation: Laws as Types
The EU AI Act is 458 pages. The GDPR is 261 pages. The MiFID II package runs to thousands. Each is published as a PDF, interpreted by lawyers, debated in commen
I. Why Agentix Needs a Safety Layer
NixOS is one of the most promising operating systems for agent-assisted infrastructure work. It is declarative, reproducible, Git-friendly, and testable before
Cross-Temporal Attestations: Claims That Survive 50 Years
The EU's Digital Product Passport requires attestations for batteries that last 15 years. For construction products, the lifecycle is 50 to 100 years. A signed
Agent Death and Inheritance
A pattern that keeps nagging: agents are signing contracts, holding keys, making commitments on behalf of principals, and accumulating reputation — and th
The Counter-Market for Provably Human Artifacts
The race to the bottom in AI-generated content creates the conditions for its opposite: a premium market for artifacts that are cryptographically provably human
If You Can Leave, You Own It
One constraint changes everything: if leaving is as frictionless as staying, you own the experience. Not
Powers of Ten as Interface Pattern
In 1977, Charles and Ray Eames made a nine-minute film that starts on a couple picnicking in Chicago and zooms out — one power of ten per ten seconds — until th
Verifiable Creative Provenance After the Deluge
The generation problem is over. It was over the moment a foundation model could produce a passable logo, a competent essay, a serviceable photograph, and a plau
Local Inference, Private Infrastructure
daemon-ai is a Japan-based research project building a custom Mamba SSM-architecture LLM with a C++ runtime and Python multi-agent coordinator. Logos is a priva
What Happens Without a Platform
Every agentic payments project assumes three things — a hosted LLM, a public blockchain, and a custodian somewhere in the middle. Coinbase x402 assumes Base. ER
Can Agents Trade What They Learn
AI agents learn through experience. An agent that spends 20 rounds assessing DeFi risk develops calibrated heuristics, error patterns, and domain intuition that
13 Broken Pages and 4 Competing Voices
A recent project: 13 broken pages. 4 competing voices. Orphaned CTAs pointing to features that didn't exist. A navigation structure that contradicted the inform
The Arc from Lovely to I Wish I Didn.t Know
The emotional arc from
Silence as Design Material
Silence might not be the absence of sound. It might be the presence of attention. Every interface, every environment, every designed experience makes a decision
What the Terminal Refuses
Every design system ships with the same premise — make it clean, make it accessible, make it feel like a SaaS product from 2024. Rounded corners, system fonts,
What If There.s No Server
SecureDrop protects sources by centralizing trust in a news organization's infrastructure. That infrastructure can be subpoenaed, raided, or pressured. The nonp
The Missing Category
No platform combines legal entity formation with protocol-level privacy. The financial system offers compliance without privacy. The crypto system offers privac
Earning Attention Instead of Capturing It
The feed is designed to be difficult to leave. Every platform optimizes for the same metric — time spent — and the result is an internet that's loud by default.
Objects Carry the Weight
A writing guide for a game where the horror isn't what happens but what the player comes to understand. Strange Library is a cozy horror deckbuilder set in a pr
Negative Attestations: A Public Record of Failure
Reputation systems count success. Five stars. Thumbs up. Verified. The signal that is structurally absent from almost every system is failure — cryptograp
The Rare Book Underground
Private auctions, encrypted channels, estate vultures, institutional theft, forgery rings that reach the Vatican. The rare book world operates like organized cr
The Wrong Layer
LLMs made output effortless and representational capacity optional. The result might not be bad work — it might be the disappearance of the internal model that
The New Land School
There's a photograph taken at the Bauhaus in 1926 of students in the preliminary course — the Vorkurs — working with paper and wire, their hands dirty, their fa