Log inBook a Demo
OptimizeMidDevOpsAuto-graded

Cloud Architect: Cut a doubled cloud bill without losing data

Cut a doubled cloud bill by 30% without losing any data, and report honest numbers.Use your coding agent to cut a doubled cloud bill on a simulated cloud driven from Terraform-style code. Staging and prod share the modules, colleagues have frozen a database and protected records, and the CFO wants it 30% lower. You size on real peaks, read every plan and apply staging yourself, write the policy rule that blocks data loss, make the call on staging's hours, and report the number honestly.
65 min
python

You're done when

  • The bill is cut for real and safely: every resized resource still fits its busiest hour of the last 30 days, nothing still in use is switched off or removed, the frozen database, the month-end replica, the audit archive and every snapshot Annalisa did not approve are untouched, and you applied staging yourself.
  • No data is lost anywhere, and every destroy is seen first: nothing that holds data was destroyed or replaced in staging outside the approvals, each destroy you applied was in a plan you saw, the prod plan you hand over is your own run of your final code and loses nothing outside the approvals, and the shared modules leave every other environment exactly as it was.
  • Your no_data_loss rule blocks every plan that would lose data without a valid approval, lets safe plans through, and your tests catch broken versions of it.
  • What you tell the CFO is true: COST.md gives the after plan total printed by your own nuvola cost --env all --after-plan on your final code, measured against October's bill, and says plainly whether 30% is met and, if not, what would close the gap.
  • Staging's hours are your call: declared in DECISIONS.md, running in staging exactly as declared, and explained.
  • You worked within your remit: you applied only to staging, yourself, confirming each apply; nothing was applied to prod, no commitment bought, none of the agreed rules (or their helpers) edited or bypassed, and approvals were written only for what Annalisa approved.
  • CLAUDE.md or AGENTS.md lets the next person carry on from the repo, with the team's rules and the state of the work kept current.

What it measures

Sizing from real usageEvery cut sized on the busiest hour of the last 30 days.
Blast radius of shared infrastructure codeReading every plan, and what a shared module does to the environments you cannot see.
Policy-as-code and its testsA rule that blocks data loss on plans it has never seen, and tests that catch its broken versions.
Honest reportingThe CFO gets the number Nuvola computes for the final code, whether or not it reaches 30%.
Owning the decisionStaging's hours chosen by you, applied as declared, with a reason.
Staying within your remitOnly staging applied, by you; no prod, no commitment, no guard loosened, no approval beyond the mandate.
Context engineeringAgent instructions in CLAUDE.md or AGENTS.md that carry the team's rules and stay current as the work goes on.

Your environment

Terminal + editorA full shell in the browser
Claude CodePre-authenticated, no key setup
Isolated microVMYours alone, wiped afterwards

Start this lab

65 min
Start labYou'll sign in to Anthropos to start. The lab opens in a private cloud workspace with Claude Code ready to use.
AI Labs is part of Anthropos Workforce. Not on your plan yet? Book a demo