Technical Expertise & Experience Sharing

01

10 min read

Claude Code as a back-office: wiring Drive, Gmail and Trello to actually run your company

Plugging Claude Code into Google Drive, Gmail and Trello to audit what's actually happening in a startup: 107 files reorganized, emails that never arrived, a Trello board reconciled with reality. Including the OAuth traps nobody documents.

02

11 min read

Claude Fable 5: what the official demos don't tell you

Fable 5 migrates 50 million lines of Ruby in a day. That's true. But while prepping a training module, I ran into the flip side the announcements skip: 500k to 1M tokens per session, 19 timeouts out of 33 tasks, and guardrails that trip at the faintest hint — to the point that Anthropic had to publicly apologize 48 hours after launch. The real manual for a model you commission instead of steer.

03

12 min read

chrome-devtools MCP from WSL: driving (and auto-launching) a Windows Chrome

Setting up the chrome-devtools MCP in WSL2 to drive a real Windows Chrome: the hostAddressLoopback networking prerequisite nobody documents, why --autoConnect doesn't work cross-OS, and a PreToolUse hook that launches Chrome on its own at the right moment.

04

9 min read

Claude Code Remote Control: resume your WSL sessions from your phone

How to drive Claude Code sessions stored in WSL from your phone using the remote-control command, and why DISABLE_TELEMETRY=1 silently blocks the feature even on Max plan.

05

23 min read

Self-hosted family SFTP: exposing a NAS without port forwarding using Pangolin + Newt + SFTPGo

Four non-techy relatives need access to folders on my NAS from anywhere, simply, with a mainstream Android app. The old Internet-exposed SFTP is gone. Full story of replacing it with a Pangolin + Newt + SFTPGo stack: no public IP on the NAS, outbound WireGuard tunnel, raw TCP SFTP coming out of a 3€/month VPS Lite. Plus two hours wasted blaming my ISP when the culprit was sitting one panel over.

06

15 min read

Kernel panic on a Debian NAS: recovering a truncated libc without rebooting into the void

One February morning, my Terramaster F6-424 NAS running Debian 13 won't boot. Immediate kernel panic on the very first userspace exec. No recovery mode, single kernel installed, aggressive hardening: the system has no way to catch itself. Full story of the diagnosis (libc.so.6 turned into "ISO-8859 text" instead of ELF), the recovery from a live USB (with the apt↔libc chicken-and-egg trap), and the 2026 best practices I applied so it never happens again. Bonus: a second incident in May that revealed the real root cause.