Blog

Articles & Insights

Technical deep-dives, engineering lessons, and ideas from over a decade of building products.

Filter by tag

Your Coding Agent Bill Is a Routing Problem

Running coordinated teams of coding agents on a single frontier model destroys budgets. Role-based routing for agentic workflows is not optional anymore.

4 May 2026aiengineeringllmagentsmodel-routingcost-optimizationagentic-codinglocal-inferencellmopsclaude-code

Read article

The Cognitive Atrophy Paradox: What the Research Actually Says

MIT EEG work, the APA 2026 study, Gerlich age-gap data, and the MDPI CAP model. What survives fact-checking, what does not.

18 April 2026aicognitionresearchopinion

Opus 4.7 Is Great. I'm Still Waiting for Mythos.

Claude Opus 4.7 just shipped and it is a genuinely good release. But the model I am actually watching for is Mythos.

16 April 2026aianthropicclaudeopinion

Why Do Many Token When Few Do Trick

Caveman is a cross-harness plugin that strips filler from LLM output, cutting token use by up to 87 percent without losing accuracy.

13 April 2026aitokensagentstools

Killing Prompt-and-Pray: The Harness-First Roadmap

The series finale. Going harness-first is not a subscription or a tooling purchase. It is a five-layer operational shift, and this is the practical roadmap for running it at a real engineering org.

12 April 2026harness-engineeringaiengineering-leadershipagentic-workflow

The Ghost in the Code: Why Humans Remain Indispensable in the Age of AI Programming

AI has commoditized coding, but programming (the art of solving human problems with technology) remains irreducibly human.

4 April 2026aisoftware-engineeringprogrammingopinion

The Moltbook Lesson: Why AI-First Needs Human-Verified Security

In January 2026, agents built a social network and forgot to lock the front door. The Moltbook breach didn't just expose millions of API keys—it exposed a fundamental flaw in the hands-off AI development philosophy. Introducing the RIDER framework.

2 April 2026harness-engineeringaiengineering-leadershipagentic-workflowsecurity

Context Rot: Why Your Agent Gets Dumber the Longer It Works

LLM reasoning decays as conversation history grows. The fix isn't prompting harder. It's decomposing features into subtasks where each one gets a fresh context window and reads its state from disk.

21 March 2026harness-engineeringaiengineering-leadershipagentic-workflow

The 6-Layer Shield — Architecting for Agents

Agents don't read documentation — they hit walls. The 6-Layer Shield is an architectural model that turns your codebase into a habitat with mechanically enforced boundaries, so agents can't drift into chaos.

17 March 2026harness-engineeringaiengineering-leadershipagentic-workflowarchitecture

1,500 PRs in 5 Months — The OpenAI Harness Benchmark

OpenAI's Harness team shipped ~1 million lines of production code with zero manual code — 1,500 PRs, 3.5 PRs/day/engineer, 90% time savings. Here's what it means for the future of software delivery.

16 March 2026harness-engineeringaiengineering-leadershipagentic-workflowopenai

Why Modular AI Agent Architecture Outperforms Monolithic Tools for Complex Workflows

Modular AI agent architectures with Pi, Tallow, and TillDone outperform monolithic tools for complex migrations and multi-agent orchestration. Learn why control and context flexibility matter more than batteries-included polish.

16 March 2026aiagentsmodular-architectureworkflow-orchestrationai-engineeringtallowtilldonepi-agent

Page 1 of 2