Skip to content

Blog

Notes from our engineering work: AI agents, security and the systems that run in production.

Any password works, if it's long enough

A musl quirk lets any 257+ character password log in to some Alpine servers running OpenSSH without PAM. Who's exposed and what to do now.

  • DevSecOps
  • CyberSecurity
  • OpenSSH
  • CloudSecurity

Who's watching checkout at 3 a.m.?

One read-only AI agent per subsystem, plus a coordinator that pages a person once. How we'd watch production at 3 a.m.

  • AIAgents
  • SRE
  • Observability
  • DevOps

Your app, read back to you

AI agents can now drive Ghidra and IDA. What an attacker can learn from your app, and how we run that check on your own build first.

  • AI
  • CyberSecurity
  • ReverseEngineering
  • DevSecOps

Not every AI decision needs a chatbot

Decision models return typed answers with a confidence score in milliseconds. Four places they pay off: triage, guardrails, routing and alerts.

  • AIEngineering
  • AIAgents
  • LLMOps
  • Automation
  • OpenSource

All topics