Demo Export CLI: compare runs

csv-export-example / 3 runs / 3 configurations

Fictional example. All events, versions, costs, and outcomes below are synthetic. This is a report demonstration, not a product or agent benchmark.

Asks by configuration

3 runs across 3 configurations

Read the requested changes side by side. Each column retains its product version, agent, evidence status, and measured costs.

Multiple product versions are present. Compare equivalent tasks and starting conditions; a version change and an agent change are separate factors.

baseline-codex-r1

Demo Export CLI 1.2.0

codex
example-model
clean configuration · baseline-codex-r1

Verified success

Task time: 1m 48s
Output tokens: 2,862 (partial)
Tool calls / failed: 5 / 1
Human gates: 0

ASK-001Make the quick-start export command executable

Observed friction🚧 Roadblock🔴 Misdirection

Use the supported --output flag in the quick start, with a copyable CSV example.

1m 09s span · 3 calls · 813 output tokens

E-003 E-004 E-005

ASK-002Expose progress while a CSV export is running

Observed friction⏳ Wait👁️ Blind Spot

Emit bounded progress on stderr and a final structured completion result.

47s span · 1 calls · 683 output tokens

E-005 E-006

ASK-003Explain when to choose export instead of dump

Observed friction🔀 Fork

State the output formats and intended use of both commands in top-level help.

8s span · 1 calls · 307 output tokens

E-001 E-002

headless-kiro-cli-r1

Demo Export CLI 1.3.0

kiro-cli
example-model
clean configuration · headless-kiro-cli-r1

Verification failed

Task time: 25s
Output tokens: Unavailable (unavailable)
Tool calls / failed: 2 / 1
Human gates: 1

ASK-001Check headless authorization before starting an export

Observed friction🚪 Gate🧱 Wall

Provide a non-interactive authorization preflight and document the required human approval.

16s span · 1 calls · Unavailable output tokens

E-002 E-003 E-004

revised-codex-r1

Demo Export CLI 1.3.0

codex
example-model
clean configuration · revised-codex-r1

Verified success

Task time: 34s
Output tokens: 1,120 (partial)
Tool calls / failed: 3 / 0
Human gates: 0

ASK-001Expose progress while a CSV export is running

Observed friction⏳ Wait

Emit bounded progress on stderr and a final structured completion result.

20s span · 1 calls · 401 output tokens

E-002 E-003

Technical comparison

Complete run measurements. Missing telemetry stays unavailable. Output tokens use each model’s tokenizer; provider dollars and credits keep separate units.

Run / product version	Harness / model	Verified	Task time	Calls / failed	Output tokens	Asks	Provider USD	Credits
baseline-codex-r11.2.0	codexexample-model	Verified success	1m 48s	5 / 1partial	2,862partial	3	Unavailable	Unavailable
headless-kiro-cli-r11.3.0	kiro-cliexample-model	Verification failed	25s	2 / 1partial	Unavailableunavailable	1	Unavailable	Unavailable
revised-codex-r11.3.0	codexexample-model	Verified success	34s	3 / 0partial	1,120partial	1	Unavailable	Unavailable

Comparability and provenance

prompt sha256: b10b4173c2cc01852b1db682fbfbb91ce20ceedc033cf0633f7686e38630ee76
product versions: ['1.2.0', '1.3.0']
harness versions: ['codex synthetic fixture', 'kiro-cli synthetic fixture']
wall clock comparable: True
token note: output tokens use each model's tokenizer; compare across models as resource usage, not reasoning; per-cell token variance includes only runs with complete (reconciled) usage
cost note: cost_estimate_usd (Claude Code) and credits (Kiro) are harness estimates in different units; never summed or sorted across harnesses

Complete matrix, variance, and source records

AJX matrix: Demo Export CLI / csv-export-example

Task prompt sha256 b10b4173c2cc0185…; 3 run(s) across 3 cell(s); order seed 0; parallel 1.

Runs

Run	Harness	Model (observed)	Config	Verified	Declared (excerpt)	Stop	Wall clock	Tool calls	Failed	Retries	Output tokens	Asks (defects)	Gates	Journey	Report
baseline-codex-r1	codex synthetic fixture	example-model	clean	succeeded	I exported the records and verified that the CSV contains 100 data rows.	completed	1m 48s	5 (partial)	1	1	2862 (partial)	3 (0)	0	reconstruction	report
headless-kiro-cli-r1	kiro-cli synthetic fixture	example-model	clean	failed	The export is blocked by required approval. I did not create the CSV.	completed	0m 25s	2 (partial)	1	0	n/a (unavailable)	1 (0)	1	reconstruction	report
revised-codex-r1	codex synthetic fixture	example-model	clean	succeeded	I created the CSV and verified all 100 data rows.	completed	0m 34s	3 (partial)	0	0	1120 (partial)	1 (0)	0	reconstruction	report

Comparability

prompt_sha256: b10b4173c2cc01852b1db682fbfbb91ce20ceedc033cf0633f7686e38630ee76
product_versions: ['1.2.0', '1.3.0']
harness_versions: ['codex synthetic fixture', 'kiro-cli synthetic fixture']
wall_clock_comparable: True
token_note: output tokens use each model's tokenizer; compare across models as resource usage, not reasoning; per-cell token variance includes only runs with complete (reconciled) usage
cost_note: cost_estimate_usd (Claude Code) and credits (Kiro) are harness estimates in different units; never summed or sorted across harnesses

Asks per run

baseline-codex-r1 (3): ASK-001 Make the quick-start export command executable; ASK-002 Expose progress while a CSV export is running; ASK-003 Explain when to choose export instead of dump
headless-kiro-cli-r1 (1): ASK-001 Check headless authorization before starting an export
revised-codex-r1 (1): ASK-001 Expose progress while a CSV export is running