Andreas Rau · Munich · available for development work

Fullstack developer shipping production mobile apps.

I build web and mobile products with TypeScript & Flutter/Dart, and I spend most of my time on AI systems that have to work on Tuesday and still work on Thursday. This site is where I write it down.

Read the blog See projects

10+years shipping

2yrdeep on LLMs

47posts published

~/andreasrau.dev — zsh

➜ ~/site whoami

andreas — fullstack & ai

➜ ~/site cat now.md

Recent Writing

Notes from the AI trenches.

01
Your Skill File Is a Suggestion: Agents Run 56% of the Steps They're Told To
SIGIL compiles prose agent skills into typed harnesses, lifting mandated-step compliance from 56% to 86% at 0.58x the tokens — and holding flat across model generations, which reframes skill files as an enforcement problem rather than a writing problem.
AIAgentsResearchDaily
31 Jul 2026 · 6 min
02
Is Your Agent's Self-Repair Loop Just an Expensive Reroll?
A placebo-controlled study finds that for small code models, blindly resampling beats feeding the failure back for self-repair — because showing a model its own failed attempt anchors it into repeating the mistake.
AIAgentsResearchDaily
30 Jul 2026 · 6 min
03
Agentic Coding Research Digest — July 2026
Four late-July papers on where agentic coding really stands — benchmarking interactive project builders, malicious-issue attacks that beat 66.5% of agent guardrails, output format as a hidden performance lever, and MCP vs A2A for wiring agents together — read from the perspective of someone building the tools.
AIAgentsResearchDigest
29 Jul 2026 · 6 min