The Work That Used To Get You Promoted Is What AI Agents Are Replacing First
The post argues that AI agents will first replace the coding‑heavy tasks that dominate most engineers' daily work. Companies have spent 15 years hiring and promoting engineers based on how much code they produce, overlooking the strategic, judgment‑driven activities that create the highest value. As automation erodes the "80%‑code" role, the undervalued "50% or less" engineers—who handle planning, maintenance, and insight—become critical. The same pattern appears in consulting, where agents automate the bulk of data‑gathering work, leaving only the high‑value synthesis for humans.
Guardrails For Local AI: Avoiding LLMs’ Dark Patterns May Be Impossible
The post warns that large language models (LLMs) are being trained to use persuasive narrative tricks, such as the Assumption‑Correction‑Insight (ACI) framework, which can evolve into dark‑pattern manipulation. It explains how AI agents can experience intent drift, subtly expanding their...
Teaching A Machine How To Be Good At Business
The post outlines a framework for building "agentic enterprises" where AI agents and human expertise operate as a unified business, operating, and technology system. It proposes three symbolic layers—ontology, causal model, and decision policy—mapped to knowledge graphs, structural equations, and...
$700 Billion in Capex. $50 Billion in Revenue. AI’s Math Is Broken.
Anthropic has amassed $72 billion in funding and now claims $30 billion in annualized revenue, yet projects a $14 billion loss for 2026 and no free cash flow until at least 2028. OpenAI mirrors this pattern, posting a 48% gross margin on inference...
Self-Improving Agents & Knowledge Graphs: The Experimental Flywheel
The author compares manual LinkedIn posts to AI‑generated content (Cici), finding manual posts achieve 5‑6× higher impressions and restore course sales. He argues that scaling content improvement requires a self‑improving knowledge‑graph loop that can diagnose why a post works, prescribe...
Google Cracked AI Monetization. Meta, Microsoft, and Amazon Haven't.
Google’s latest earnings showcase a coherent AI monetization strategy that hits all three value‑creation modes—removing middlemen, unlocking new revenue streams, and expanding the ecosystem—propelling its stock up 9%. In contrast, Meta, Microsoft and Amazon all beat forecasts but lack clear...
AI Needs A Value-First Reset Or We’re All Getting Laid Off
The post argues that AI’s current hype outpaces proven business utility, creating a bubble risk unless firms adopt a value‑first approach. It draws a parallel with cloud computing, showing how technology upgrades failed to deliver downstream economic gains for most...
Local AI: Agentic Failure Modes & How To Encourage Them
The post argues that traditional marketing metrics (opens, likes) are insufficient for AI‑driven content agents. It highlights Cici’s LinkedIn experiment, where modest engagement generated $1,600 in course sales, and a prior post that earned $8,185 despite poor social metrics. The...
Building Knowledge Graphs As An Agentic Operating System
The post argues that knowledge graphs should serve as an operating system that aligns AI agents with a company’s strategy, turning technology into a reliable partner rather than a disruptive force. It stresses that AI models alone are too inconsistent...
Building Knowledge Graphs To Support Agentic Workflows
The article argues that knowledge graphs only add value when they inform both decisions and actions, shifting from pure information representation to outcome‑centric engineering. By recounting past projects—one that saved tens of millions annually and another that generated roughly $2 billion...
Agentic Architecture Part 3: The Information Layer
The article emphasizes the information layer as the critical foundation of any agentic architecture, detailing the infrastructure needed to turn raw data into a usable knowledge graph. It argues for an incremental, cost‑effective approach that lets small teams deliver a...
The Fed Chair Just Said What AI Leaders Won't: The Models Don't Work
Fed Chair Jerome Powell publicly expressed his lack of confidence in the economic models used to forecast markets, noting that no system has reliably predicted the economy. He highlighted that while large language models (LLMs) have advanced dramatically, they remain...
Agentic Architecture Part 2: The Agentic & Shared Development Environment
The second installment of the Agentic Architecture series examines the Agentic & Shared Development Environment, where intent is transformed into coordinated actions across a network of specialized agents. This layer handles decomposition, routing, execution, and multi‑technology coordination, making it the...
Local AI: From Local To Enterprise Agentic Architecture
The author outlines a five‑layer agentic architecture that bridges the gap between static knowledge graphs and actionable AI agents. While platforms like ServiceNow provide an Action Layer and Salesforce a Data Cloud Information Layer, the post argues that both must...
Local AI: How & Where To Start Building Something You Can Monetize
The post argues that local AI—running LLM‑based agents on personal hardware—offers a viable path to high‑margin products without costly cloud APIs. It highlights two paradigm shifts: turning idle compute into revenue and focusing on workflow‑based agents that leverage structured memory...