ilpanich / axiam / 29698939951

85%

push

github

docs(claude_dev): expand plan E1 — implement & validate the SDK client benches

The benchmarks/sdk/ harnesses are effectively stubs: no bench has ever produced
a validated axiam.sdk-bench/v1 "ok" record, and kotlin/swift/c/cpp are pure
emit_pending scaffolds. Replace the one-line E1 item with concrete sub-tasks:

* E1.1 (Sonnet, per language) — validate & repair the 7 code-bearing benches
  (rust, python, typescript, go, java, csharp, php) against a live seeded
  target until each emits a spec-conformant OK record.
* E1.2 (Sonnet, per language) — implement the 4 stub benches (kotlin, swift,
  c, cpp) per their TODO.md, gated on the corresponding SDK being buildable.
* E1.3 (Sonnet) — cross-SDK run on p0+p2, overhead-vs-wire-baseline table in
  the report, and README status claims corrected to measured reality.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L9wexJBZaB7DbhFrc9vtt1

1682 of 1859 branches covered (90.48%)

Branch coverage included in aggregate %.

27337 of 33351 relevant lines covered (81.97%)

71.07 hits per line