NodePrime Hillclimb — Trajectory Dashboard

generated 2026-04-24 11:11 UTC · window: last 30 days · source: operational memory (Postgres)

Trajectories — mean score per iteration

Runs

run_tag	episodes	iters	min	max	last seen
`audit_gtm_hooks_20260424`	3	3	0.6100	0.7796	2026-04-24 10:39
`audit_gtm_subjects_20260424`	3	3	0.6925	0.7197	2026-04-24 09:59

Agent activity

agent	invocations	parse_ok	avg ms
`gtm_subject_writer@1.0.0`	15	15 (100%)	24686
`gtm_hook_writer@1.0.0`	15	15 (100%)	26573
`email_outreach@1.0.0`	1	1 (100%)	40268

Meta-Evaluator patches

patch	lvl	status	fix_type	conf	target	hypothesis	decided
`d6a5760d`	L4	● pending_review	prompt_edit	0.72	`agent:gtm_subject_writer@1.0.0`	Prompt lacks position-aware guidance. For E1 (opening email), the agent generates operational subjec	—

Recent episodes

timestamp	tag	task	score	elapsed ms
2026-04-24 10:39:24	`audit_gtm_hooks_20260424`	audit_gtm_hooks_20260424:iter2:candidate	0.7780	-
2026-04-24 10:34:54	`audit_gtm_hooks_20260424`	audit_gtm_hooks_20260424:iter1:candidate	0.7796	-
2026-04-24 10:30:25	`audit_gtm_hooks_20260424`	audit_gtm_hooks_20260424:iter0:baseline	0.6100	-
2026-04-24 09:59:32	`audit_gtm_subjects_20260424`	audit_gtm_subjects_20260424:iter2:candidate	0.6925	-
2026-04-24 09:55:28	`audit_gtm_subjects_20260424`	audit_gtm_subjects_20260424:iter1:candidate	0.7165	-
2026-04-24 09:51:23	`audit_gtm_subjects_20260424`	audit_gtm_subjects_20260424:iter0:baseline	0.7197	-
2026-04-24 08:01:59	`supabase_smoke`	draft a cold outreach email to Acme Accounting — 40-person firm posting a junior	-	56806