handoff / 90-second captioned walkthrough
No audio. Explanations of saved evidence; not a new live client-session capture.

00:00:00.000 – 00:00:15.000
OrgX / historical cross-client handoff

Published observations: July 11, 2026
Claude Code -> OpenCode
One task, one client pair
Not a new live session recording
This walkthrough explains three published historical observations from one task and one client pair.

00:00:15.000 – 00:00:30.000
Run 1 / baseline

Without OrgX: 0/5 fields recovered
A fresh workspace had no prior record.
Writer model: claude-opus-4-1-20250805
The baseline recovered zero of five fields. Its writer model differed from the OrgX runs, so this is not a controlled causal comparison.

00:00:30.000 – 00:00:45.000
Run 2 / retain the outage

With OrgX during outage: 0/5
Records existed. Tool responses failed.
The failed observation stays in the corpus.
The OrgX outage run also recovered zero of five fields. The record existing did not make it readable.

00:00:45.000 – 00:01:00.000
Run 3 / post-fix observation

With OrgX after the fix: 5/5
Decision / rationale / owner
Artifact / next action
Recorded date: July 11, 2026
The post-fix run recovered all five fields: decision, rationale, owner, artifact and next action.

00:01:00.000 – 00:01:15.000
Read the exact sources

handoff-corpus.json preserves all 3 runs.
Every source URL pins a repository commit.
Original records link to raw transcripts.
Inspect the pinned run files and their original transcript links. The local script only checks the recorded fields and totals.

00:01:15.000 – 00:01:30.000
Keep the limits visible

One task. One client pair. Different models.
No universal client-parity claim.
No fresh paid benchmark.

Current releases need their own checks.
These historical observations do not establish current parity or a general effect. Recheck current releases under an approved test envelope.