Which Parts Of An Unfamiliar App Should AI Explore First?

A guide based on REA shows how to direct coding agents toward specific app behaviors, verify their findings and avoid treating guesses as evidence.

A Skeptic’s First Contact With Kubernetes (2024)

A 2024 infrastructure engineer’s first hands-on look at Kubernetes focuses on control loops, service networking and the tradeoffs behind its complexity.

Two days, one plan and a team of coding agents: inside the Gewerkton relaunch

Gewerkton replaced its whole look in two days, with AI coding agents carrying out the work from written plans: web app, phone app, Studio, icons, a website in forty languages and regional films. How it was checked, what broke and what the films cost.

Show HN: Petal, Open-source CoScreen/ScreenHero Finally Free

Petal presents a free desktop app for collaborative screen sharing and remote control, with cloud, self-hosted and build-your-own options.

What A Clock-Driven Close Looks Like In “Nightmarket — Mercado De Medianoche”

A ThorstenMeyerAI.com field guide describes how time-driven visual changes make a digital night market appear to wind down.

How I Put Jev’s Visual AI Through A 30-Image Trial

A 30-image trial tested Jev 1.14 and 1.13 on headline matching and product identification after TypeSafe added opt-in image support.

Iosevka Is An Open-source, Sans-serif And Slab-serif Monospace Font

Iosevka is an open-source typeface family with monospace, slab-serif and quasi-proportional variants for coding, terminals and documents.

Building Resilient AI-powered Software, And Planning For When Things Go Wrong

An excerpt from Sam Newman’s new book covers disaster recovery, AI model dependencies and planning for outages in software systems.

A Retirement Countdown And The Lessons Of Building With AI

A developer says an AI coding agent helped build a free retirement countdown app in five days, while he directed, tested and approved the work.

FrontierCode: How To Navigate Maintainer Review Of AI Contributions

Artificial Analysis tested 23 models on FrontierCode, where Claude Opus 5.5 led but passed every blocking check in 57.7% of attempts.