Program menu: what we build to put Mo through its paces

Robert (session 2, on Q14): one program is not enough; build a variety of kinds to find Mo’s strengths and weaknesses and collect real data to improve the language. This is the measurement engine behind direction 28.

The menu, in rough order

#ProgramStressesMeasures
1Durable job queue + HTTP APIprocesses, supervision, deadlines, neverthe founding premise: agent builds it, Robert reads at spec altitude
2CLI tool (log analyzer)pure code, stdlib, single binarystdlib gaps; is Mo pleasant with no concurrency?
3KV store with a wire protocolbytes, sized ints, hot loopsruntime speed, overflow-check cost ([[q04-integer-types-and-overflow
4Web backend (“with Postgres” as proposed; built as notes over Fs, no database brick yet, program-4)package story, capability-scoped deps[[q17-package-management-and-supply-chain
5Agent harness (tools, budgets, retries)capabilities as permissions, deadlinesis Mo good at the thing it is for?
6Payments ledger with invariantsnever, contracts, SMT tier, fault injectionproof rate, tier-3 timings ([[q08-verification-tiers
7The Mo toolchain itself (late; program 8 on the roadmap, where program 7 is a Redis subset against Redis’s own tests)everything, at scalecompile speed on 100K lines ([[d23-compile-speed-first-class

Collected on every program

  • build time per KLOC; incremental rebuild latency
  • agent loops-to-green per task
  • contract and proof density; verified: line over time
  • lines-per-function distribution (tests the 70-line law, Q12)
  • crash reports per KLOC and time-to-autonomous-fix (direction 21)