Chaos with Anmol · @chaoswithanmol
Anmol Handa
Chaos with Anmol
I break prod on purpose — so you don't break it by accident.
Ex-AWS Cloud Infrastructure Architect. Today I work on reliability, resilience and chaos engineering for a large platform. Here you'll find the long versions of my 30-second reels: exact commands, caveats and sources.
Start here
I break production on purpose
The intro reel: who I am and what this channel is about.
2Multi-AZ means you're safe?
One AZ in us-east-1 was down for more than a day in May 2026. Multi-AZ is a claim until you test it.
3AI agents need a boss
My AI agent café: each agent has one job, and nothing public happens without my one-tap OK. Starter prompt inside.
What I cover
Chaos & Reliability
Game days, gray failures, error budgets, failover drills. Break it small so it never breaks big.
AI Agents in Prod
Timeouts, fallbacks, evals and human approval. What agents need before they touch anything real.
Terraform & Infra
Plans, state, drift and the AWS features that actually save you time.
Career & Build
Building with AI in public: agent offices, faceless channels, and staying relevant in tech.
Latest 30-sec Infra Fixes
Each reel is 30 seconds. Each page here is the detailed explainer: what's new, why it matters, copyable commands, caveats and a safe way to try it.
Chaos testing without being Netflix: Resilience Hub's 4 ready-made tests
AWS Resilience Hub (next gen) now ships recommended resilience tests powered by AWS FIS: AZ recovery, dependency validation, Multi-Region isolation and recovery.
IF2terraform plan taking forever? Try -minimal-refresh
Terraform 1.17 adds terraform plan -minimal-refresh: refresh only the resources with proposed changes. Plus the drift catch, and two bonus 1.16/1.17 features.
IF3Your alarms catch outages, not slowness: test gray failures with AWS FIS
AWS FIS scenario library: "AZ: Application Slowdown" and "Cross-AZ: Traffic Slowdown" simulate gray failures so you can check alarms catch slowness and practise AZ evacuation.
Latest reels
All 21 reels →Every reel gets a page: a summary, a checklist of what to do about it, and the prompts or steps where the reel promised them.
Your Terraform state file holds secrets in plain text
Reel · Nov 2, 2026
Lose Your Prod: an SRE rap about AI and on-call
Reel · Nov 1, 2026
Mere 25 AI agents ek building mein rehte hain
Reel · Oct 31, 2026
99.9% uptime = 8 hours 46 minutes of DOWN, every year
Reel · Oct 30, 2026
I never show my face: how I built this channel with AI
Reel · Oct 29, 2026
Ek tapri band… aur poora app band? Region, AZ and failover over chai
Reel · Oct 28, 2026
Articles
All articles →The SRE Interview That Actually Works: Screening for Resilience Thinking
Most SRE interviews test what candidates know. The ones that work test how they think when systems fail: a four-part framework, rubric and red flags.
99.9% uptime means 8 hours 46 minutes down. Per year.
What three nines actually allows, what one bad night does to it, and why uptime is a budget, not a promise.
Your retries can take down your own service
How retries without backoff, jitter or budgets turn a slow dependency into an outage, and the four habits that stop it.
Your system will break. The only question is whether you're watching when it does.
The chaos game-day checklist: run your first chaos experiment safely in one afternoon.
Your backup region is drifting right now. Here's the checklist that catches it.
Everything drifts: your DR region and your AI agents. A drift and failover checklist for apps and AI agents.
Build Your Own AI Agent Café, Step by Step
How I run 20 AI agents on two hours a night, and how you can start with three: eight steps and a starter prompt.
Your logs are eating your budget. Cut the bill without going blind.
Five ways to cut the observability bill, a $1.1M debug-line story, and the full log-bill checklist.
Small, reversible, predicted: three habits for safer changes
Predict before you change, shrink what you can't undo in a minute, and don't stop the review at "human error".