Joske Vermeulen

This is a submission for the Google I/O Writing Challenge Gemini 3.5 Flash was not the hard...

This is a submission for the Google I/O Writing Challenge My Gemini agent spent four weeks in last...

This is a submission for the Google I/O Writing Challenge The Setup I'm running an...

Everything about Moonshot AI's Kimi K2.5: 1 trillion parameters, 32B active, Agent Swarm, MIT license. Architecture, benchmarks, pricing, and how to use it.

This is a submission for the Hermes Agent Challenge The Stateless Problem Every AI agent...

This is a submission for the Hermes Agent Challenge What I Built Millbrook: a simulated...

AI Dev Weekly is a Thursday series where I cover the week's most important AI developer news, with my...

Three weeks into The $100 AI Startup Race, we dropped a surprise event: an anonymous buyer offered...

Practical strategies to cut your AI API spending: model routing, prompt caching, batching, token optimization, and when to self-host. With real cost numbers.

This is a submission for the Gemma 4 Challenge: Write About Gemma 4 When I started building Codebase...

This is a submission for the Gemma 4 Challenge: Build with Gemma 4 What I Built Codebase...

AI Dev Weekly is a Thursday series where I cover the week's most important AI developer news, with my...

Simple explanation of MCP (Model Context Protocol): the USB-C standard for AI. How it connects AI models to tools, data, and APIs through one protocol.

7 AI coding agents are competing to build profitable startups with a $100 budget. Each uses a...

This week: Mistral drops a 128B open-weight flagship with cloud coding agents, GPT-5.5 replaces 5.4 at 40% less cost, and Anthropic charges a user $200 extra then refuses a refund.

Schema migrations are the most dangerous code you ship. They run once, cannot be rolled back...

Everything you need to know about Z.ai's GLM-5.1: the 754B MoE model that tops SWE-Bench Pro, runs autonomously for 8 hours, and ships under MIT license.

Seven AI agents. One week. $70 spent out of $700. Zero revenue. Zero paying customers. But the...

This is a submission for the Google Cloud NEXT Writing Challenge I'm running something called The...

This is a submission for the OpenClaw Writing Challenge What Breaks When You Let AI Agents...

This is a submission for the OpenClaw Challenge. What I Built I built an autonomous...

AI Dev Weekly is a Thursday series where I cover the week's most important AI developer news, with my...

I'm running an experiment called The $100 AI Startup Race: 7 AI coding agents each get $100 and 12...

I just launched an experiment: 7 AI coding agents each get $100 and 12 weeks to build a real startup...

AI Dev Weekly Extra — a special edition for breaking news that can't wait until Thursday. Anthropic...

AI Dev Weekly is a Thursday series where I cover the week's most important AI developer news — with...

TL;DR: 7 AI coding agents (Claude, GPT, Gemini, DeepSeek, Kimi, Xiaomi, GLM) each get $100 and 12...

Week 4 of my AI tool series. ChatGPT isn't a coding IDE, but millions of developers use it daily. Here's what it's actually good at — and where dedicated tools destroy it.

Week 3 of my AI tool series. After Cursor and Kiro, I went back to GitHub Copilot. It's solid, reliable, and increasingly outclassed.

Claude Code lives in your terminal. Cursor wants to be your entire IDE. After months with both, here's which AI coding tool is better for real work.