Code github.com/florianbuetow/tts-mcp How TTS-MCP fits together, from your AI agent down to your speakers When you kick off a long-running task with an autonomous agent, whether it is Claude Code, Codex, or a custom runner, you usually end up locked into the terminal. You sit …
The Crossover For one or two tasks, a single session is cheaper. Past that, delegating to subagents wins depending on context length and task complexity. A single session keeps everything in one growing context and takes advantage of caching that avoids reloading the context …
Code https://github.com/florianbuetow/tts-mcp TTS MCP runs Mistral’s Voxtral text-to-speech model locally, with real-time streaming playback, an interactive CLI, a FastAPI server, and an MCP server that lets Claude Code and Claude Desktop speak. It ships 20 voices across 9 …
Download github.com/florianbuetow/claude-code Sometimes I just want to fire a prompt and forget: send Claude Code a prompt from CLI (or inside a session) and have the session close itself when the job is done. This is exactly what the Terminator plugin enables you to do. All you …
I recently joined Patrick Akil on Beyond Coding post. A podcast he created for software engineers who want to grow past writing code: technical mastery, AI, product thinking, and career growth. Over 270 episodes in, he’s interviewed CTOs, principal engineers, and tech …
As AI automates manual coding, the traditional billable hour is becoming obsolete, forcing software consultancies to start to rethink how they measure and deliver value. I argue that to not be priced out of the market, firms will eventually transition to outcome-based pricing …
What Is Guard? The Origin Story Yes, You Are Absolutely Right! Or: How I won $1000 by not reviewing any code at an international AWS AI hackathon (with guard).Continue reading … Back in February, I released Guard, an open source CLI tool written in GoLang that physically …
Preamble This conversation tracks an engineer who started professional software development in October 2022 - two months before ChatGPT shipped. He arrived from outside the field with a music bachelor’s and a few years as a working classical guitarist. Four years on, he …
Eike Rackwitz spent years as a professional classical guitarist in academia and music. When Covid hit, he decided to pivot towards software engineering. Four years later, he’s a senior software engineer who runs an LLM-powered search engine with half a million sheet-music …
Download github.com/florianbuetow/claude-code If your CLAUDE.md has grown into a monolith, every convention, every workflow rule, and every stray note inlined, it is doing the opposite of progressive disclosure. The agent loads everything at session start, burns through its …
Code is cheap - Show me the HUMAN.md Being able to write and ship code used to be a proxy for expertise. But when anyone can vibe-code a repository into existence, shipping code is no longer a strong signal of expertise. This becomes a problem when you try to assess the quality …
Anthropic just released Opus 4.7, and the behaviour changes of the new models are the ones we actually wanted. They address a range of long-standing issues many of us had. A TL;DR: Response length calibrated to task: Short answers to short questions, long answers where they earn …
Infographic: The anatomy of an AI regression Stella Laurenzo’s analysis reveals a critical regression in Claude Code for AI-driven complex engineering, correlating the redaction of “thinking” tokens with a collapse in reasoning quality. Her data from over …
Illustration: Milla Jovovich engineering in a sea of documents A short case study in AI systems engineering: Hollywood actress Milla Jovovich released an AI memory system with 12k+ stars on GitHub. MemPalace - A short case study in AI engineering and marketing. Yes, that Milla …
Infographic: Overcomming the 8 Hidden Limitations of Claude Code People were quick to analyze the leaked Claude Code codebase and the findings are startling: Anthropic is aware of Claude Code’s “laziness” and hallucinations, but the most effective fixes are …
Unless you already automated maintaining CHANGELOG.md in your projects, you might find this skill useful. The changelog skill reads your Git history and writes a proper CHANGELOG.md following the “Keep a Changelog” format with semantic versioning. One command. No …
Anthropic silently added a new /dream feature for memory consolidation in Claude Code. Unfortunately, it is not yet available to everyone. If you do not have native access yet, you can use the public Dream prompt today by installing it as a global slash command. Ask Claude Code …
In engineering, retrospectives uncover work patterns that work well and those that can be improved. I built a Claude Code plugin that performs a retrospective based on your past Claude sessions to tell you how to make your interactions with Claude Code more effective. Run …
If you are still spoon-feeding your coding agent with instructions and trying to keep it on track with prompting, you need to try specs-driven development. The issue with prompts is that they hardly capture enough of your intent. Unless you’ve specified things rigorously, …
SOLID covers class design. But architecture rot happens at a larger scale - tangled services, leaky abstractions, hidden coupling between modules, brittle failure propagation. By the time these problems surface, untangling them is far more expensive than fixing a single class. So …
How to get effective code quality reviews with Claude and Codex? With the force of five Power Rangers, of course! SOLID principles are one of the most battle-tested frameworks for evaluating object-oriented design. So I encoded them into a free plugin that gives Claude Code and …
Infographic: Guard on the Kiro Hackathon podium My open source Guard tool won 8th place in the international Kiro Hackathon hosted by Dynamous.ai and AWS Kiro! Guard is a CLI/TUI tool that protects your files from unauthorized modifications by AI coding agents. It gives you …
Logo: Yes, You Are Absolutely Right! 2001: A Space Odyssey predicted “I’m sorry, I can’t do that, Dave”. Instead, we got “Yes, you are absolutely right!” - which is equally frightening if you’re a software developer. I’m frankly …
My favorite workflow upgrade this week. I finally had enough of manually sifting through 100+ emails a day and creating custom rules in my email client. I tried several email MCP servers - they were either too complex for what I needed or the AI did things I absolutely …
From Single-Core Limits to Multi-Agent Systems Moore’s Law [1] fueled decades of computing progress by predicting the doubling of transistor density approximately every two years. For a long time, this translated into rapid gains in CPU speed-until thermal limits and the …
Introduction Large language models (LLMs) map sequences of text to other sequences of text. Given an input string, an LLM predicts the most probable continuation based on the statistical patterns it has learned during its training process. What makes LLMs remarkable is not merely …
Introduction During my time as a lead data engineer to build a predictive maintenance platform I came across the concept of the default heuristic. One of my tasks was to identify platform components that would allow us to scale storage and retrieval of time series sensory data to …
Introduction Large language models (LLMs) can behave very differently depending on how they were trained. BaseLLMs are trained purely on next-token prediction over a large corpus of text. Instruction-tuned LLMs, by contrast, are further trained to follow prompts in a more helpful …