A new analysis argues that as AI coding tools make implementation cheap, software architecture becomes more important and visible, with verification replacing code generation as the bottleneck. Dex Ho…
A developer from HumanLayer shared a workflow to keep AI agents accurate by limiting context usage to under 40% and using a three-phase process of research, planning, and implementation, with subagent…
Coding agents' verbose text output is hurting usability, and HumanLayer has released an open-source skill called show-me that prompts agents to explain code visually using component trees, call stacks…
HumanLayer released 'show-me', a coding agent skill that prompts AI agents to explain their work with compact visuals like component trees and call stacks instead of prose, available via 'npx skills a…
In a benchmark run on Thursday, Fable and Sol tied at 33.3% (10/30 strict checkpoint passes) on SlopCodeBench, with Expo and Kimi K3 following at 26.7% and 23.3%, respectively. The evaluation, conduct…
A developer has added a permanent system instruction for coding agents to communicate planned changes as call stack diffs, inspired by HumanLayer's 'Program Design' section in 'Why Software Factories …
Dex Horthy, founder of HumanLayer, has outlined a four-gate software development workflow—Product, Architecture, Program Design, and Vertical Slices—designed to make critical decisions before code is …
Claude Opus 5 passed 24% of 17 checkpoints on a subset of SlopCodeBench, quadrupling the 6% pass rates of Opus 4.8 and Sonnet 5, but achieved this by producing roughly 29,000 source lines versus about…
Dexter, CEO of HumanLayer, argues that the push for AI-driven software factories is leading to lower code quality and more outages, citing a Faros AI report showing pull-request review quality down, i…
Dex, an AI engineer, introduces the 12 Factor Agents methodology, a set of principles for building reliable, scalable, and maintainable LLM-powered software. The framework, inspired by 12 Factor Apps,…
Dex Horthy, CEO and cofounder of HumanLayer, coined the term 'context engineering' and shared lessons from building AI agents, including that shipping unread code leads to disaster within months after…
Engineers are shifting focus from model selection to harness engineering—the design of prompts, tools, and scaffolding around AI models—after observing that a decent model with a great harness outperf…
Context engineering has emerged as a critical discipline in AI development, moving beyond prompt engineering to manage an AI's working memory. Developers are grappling with the challenge that models l…
A developer argues that an AI agent is defined by its event history, or 'log', rather than by its model, runtime, or execution loop. The log, an append-only record of all user inputs, model outputs, t…
A developer has found that CLAUDE.md files exceeding 200 lines cause AI models to ignore most of the content, with the effective instruction limit being around 150 rules. The project's solution places…