[JacobStephens2]

Summary

Add a repeatable benchmark command and an empirical report for the Go Ticket/Delivery/Pickup slice and its Rust owners. Record cold and edited builds in fresh Worktrees, test compilation/execution, static binary sizes, source manifests and ADR-0017 checks. Explain the feature and model differences; ADR-0001 stays unchanged.

prototype/go/benchmark.py
  fresh detached Worktree + private caches per sample
  Go / Rust builds, compiled tests, model checks, static-link verification
  results.json + command logs
docs/research/go-vs-rust-prototype.md
  measured observations + subprocess, exhaustiveness and verification tradeoffs

Closes #1835

Evidence

Execution evidence

• Before: python3 -m unittest discover -s prototype/go -p benchmark_test.py -v failed: python3: can't open file '/home/jacob/repos/thirdshift-issue-1835/prototype/go/benchmark.py': [Errno 2] No such file or directory.
  After: the same command passed all four tests in 2.917 seconds.
• Before: python3 -m unittest discover -s prototype/go -p benchmark_test.py -k PublishedReport -v failed with FileNotFoundError: [Errno 2] No such file or directory: '/home/jacob/repos/thirdshift-issue-1835/docs/research/go-vs-rust-prototype.md'.
  After: the same command passed the report completeness/numeric agreement test.
• Real experiment: PATH=/usr/local/go/bin:$PATH python3 prototype/go/benchmark.py --ref 82cc4dc2 --samples 2 --jobs 2 --output /tmp/thirdshift-1835-benchmark.PgYbxF completed both samples. Cold builds: Go 9.659–10.197s, Rust 160.736–166.977s; edited builds: Go 0.601–0.631s, Rust 8.887–9.292s. Static binaries verified: Go 3,289,272 bytes; Rust 21,210,280 bytes. Report and scope limitations.
• Final local selection: PATH=/usr/local/go/bin:$PATH CARGO_BUILD_JOBS=2 RUST_TEST_THREADS=3 /usr/bin/time -p scripts/affected-tests da63d90d5482abcd6de148d579eaaf7f072bf11c passed. Go: 154.72s; full Rust cargo test: 2963.85s, including compilation. Total wall time: 3118.73s. Full logs remain under .affected-tests/.
• go vet ./..., gofmt -l benchmark_test.go (empty), and python3 bundle-skills.py --check passed from prototype/go with Go on PATH.
• python3 -m unittest discover -s scripts -p 'test_*.py' passed all 37 tests in 11.639s.
• ADR-0001: git diff da63d90d5482abcd6de148d579eaaf7f072bf11c -- docs/adr/0001-rust-binary-with-embedded-skills.md is empty.

Test seams

• Benchmark CLI in a temporary Git repository: external tool subprocesses, JSON/log artifacts, static verification, isolation and cleanup on success/failure.
• Published report and measured artifact: required comparison topics and every published timing, source count and binary size.

Review coverage

Reviewed against the fetched issue-1830 tip, da63d90d5482abcd6de148d579eaaf7f072bf11c. The local issue-1830 branch is checked out in another Worktree at an older commit, so the fixed point is the captured upstream tip rather than changing that other checkout.

Standards: 0 findings. Spec: 0 findings. Each independent reviewer read all ten changed files and its supporting context. No generated-output exclusions; no changed file was left unread under either axis. Reports with each axis's final files-read list and the count summary are retained in .thirdshift-review-9GMu5v/ beside the Session trace.

Unaddressed findings

Standards

None: the reviewer raised no findings.

Spec

None: the reviewer raised no findings.

Security

None: the reviewer raised no findings.

Security review outcome

No unaddressed introduced findings.

Merge Danger

Door: two-way

Blast Radius: prototype

Adds opt-in benchmark tooling, evidence and tests. Rust build/size results include the larger production executable, the test workloads differ, and the two bounded model graphs differ; the report makes those limits explicit. No production language decision is made here.

Built with codex codex-cli 0.164.0 · gpt-6.1-sol · xhigh