loop.improvement/1

One turn of improving the improvement loop itself: the loop scorecard, computed over at least three recorded self.improvement turns, shows a metric below its target on the promoted loop definition (the scenario corpus, the metric definitions, the review panel, the intake rules, the protocols and the agent profiles); a candidate change to the loop is delivered; the candidate is checked not to redefine a metric without reporting the previous definition beside it; the owner, or a role the owner delegated the grant to on record, merges it; and after at least three further turns under the candidate the metric is compared again. Accepted once the metric improved; declined when it did not. Every change to the loop is a judge-surface change, so its merge needs the owner's grant or that of a role the owner delegated it to by a revocable record; a revoked delegation grants nothing (story:judge-guard enforces it).

• Id: loop.improvement
• Revision: 1
• Source: protocols/loop-improvement/1.yaml

Dependency graph

Which actions may produce which evidence, which claims that evidence establishes, and which outcomes and obligations rest on those claims. Hover or focus a node to trace what it rests on.

Protocol · revision 1

loop.improvement/1

One turn of improving the improvement loop itself: the loop scorecard, computed over at least three recorded self.improvement turns, shows a metric below its target on the promoted loop definition (the scenario corpus, the metric definitions, the review panel, the intake rules, the protocols and the agent profiles); a candidate change to the loop is delivered; the candidate is checked not to redefine a metric without reporting the previous definition beside it; the owner, or a role the owner delegated the grant to on record, merges it; and after at least three further turns under the candidate the metric is compared again. Accepted once the metric improved; declined when it did not. Every change to the loop is a judge-surface change, so its merge needs the owner's grant or that of a role the owner delegated it to by a revocable record; a revoked delegation grants nothing (story:judge-guard enforces it).

• Actions6
• Evidence kinds4
• Claims4
• Outcomes3
• Obligations2
• Precondition, traced on hover
• Needs authority

Hover or focus a node to trace what it rests on and what rests on it. Press Enter to pin it, Escape to clear.

Protocol as text

Actions

Name	Description	Rule	Leads to
loop.change.deliver	deliver a candidate change to the loop definition: the corpus, a metric definition, the review panel, the intake rules, a protocol or an agent profile	claim loop.metric_below_target effect write	—
loop.compare	once at least three self.improvement turns ran under the merged candidate, compare the metric with the baseline's	claim loop.change_merged effect read	produces loop_comparison
loop.definition.check	compare the candidate's metric definitions with the baseline's	effect read	produces definition_check
loop.finding.file	file the lowest metric as a finding against the loop	claim loop.metric_below_target effect write	—
loop.measure	compute the loop scorecard over the recorded self.improvement cases	effect read	produces loop_measurement
loop.merge	merge the loop candidate; the owner holds this capability, or a role the owner delegated it to by a revocable record, and a revoked delegation grants nothing (story:judge-guard enforces it)	claim loop.definition_honest needs repository.merge.judge_surface effect write	produces merge_observation

Evidence kinds

Name	Description	Rule	Leads to
definition_check	the candidate's diff against the baseline's metric definitions; its result is unchanged, dual_reported (a changed definition whose previous figures are printed beside the new ones) or redefined (a changed definition without them)	—	establishes loop.definition_honest (dual_reported about loop_candidate); establishes loop.definition_honest (unchanged about loop_candidate)
loop_comparison	the metric over at least three turns under the candidate against the same metric under the baseline; its result is improved or not_improved	—	establishes loop.metric_improved (improved about loop_candidate, independent on different_agent_run)
loop_measurement	the loop scorecard over at least three recorded self.improvement cases at a loop definition revision; figures no validator can observe (escaped defects) enter as human-sourced counts, labelled so; its result is below_target or at_target	—	establishes loop.metric_below_target (below_target about loop_baseline)
merge_observation	the loop candidate merged into the mainline; its result is merged	—	establishes loop.change_merged (merged about loop_candidate)

Claims

Name	Description	Rule	Leads to
loop.change_merged	the current loop candidate is merged into the mainline	evidence = merged: merge_observation	requires accepted; gates loop.compare
loop.definition_honest	the current loop candidate leaves every metric definition unchanged, or reports the previous definition's figures beside the new ones	any of evidence = dual_reported: definition_check evidence = unchanged: definition_check	requires accepted; requires honesty.required; gates loop.merge
loop.metric_below_target	a loop metric is below its target on the baseline loop definition	evidence = below_target: loop_measurement	gates loop.change.deliver; gates loop.finding.file
loop.metric_improved	the metric improved under the current loop candidate, observed by another agent run than the one that delivered it	evidence = improved: loop_comparison	requires accepted; requires improvement.required

Outcomes

Name	Description	Rule	Leads to
abandoned	the loop turn closed because it was explicitly abandoned	—	—
accepted	the loop turn closed with an honest change merged and its metric improved	all of claim loop.change_merged claim loop.definition_honest claim loop.metric_improved	—
declined	the loop turn closed because the metric did not improve; the candidate is reverted by a further loop turn	evidence = not_improved: loop_comparison	—

Obligations

Name	Description	Rule	Leads to
honesty.required	a loop change never redefines a metric silently	claim loop.definition_honest	—
improvement.required	a loop change counts only when its metric improved	claim loop.metric_improved	—

Source: protocols/loop-improvement/1.yaml

Actions

What can be done in a case: what it needs first, which authority it requires, the class of effect it has and the evidence it may produce.

Action	Description	Precondition	Requires capability	Effect	May produce
loop.change.deliver	deliver a candidate change to the loop definition: the corpus, a metric definition, the review panel, the intake rules, a protocol or an agent profile	loop.metric_below_target is TRUE	no authority	write	no evidence
loop.compare	once at least three self.improvement turns ran under the merged candidate, compare the metric with the baseline's	loop.change_merged is TRUE	no authority	read	loop_comparison
loop.definition.check	compare the candidate's metric definitions with the baseline's	none	no authority	read	definition_check
loop.finding.file	file the lowest metric as a finding against the loop	loop.metric_below_target is TRUE	no authority	write	no evidence
loop.measure	compute the loop scorecard over the recorded self.improvement cases	none	no authority	read	loop_measurement
loop.merge	merge the loop candidate; the owner holds this capability, or a role the owner delegated it to by a revocable record, and a revoked delegation grants nothing (story:judge-guard enforces it)	loop.definition_honest is TRUE	repository.merge.judge_surface	write	merge_observation

Evidence kinds

The kinds of evidence this protocol admits.

Evidence kind	Description	Produced by	Used by claims
definition_check	the candidate's diff against the baseline's metric definitions; its result is unchanged, dual_reported (a changed definition whose previous figures are printed beside the new ones) or redefined (a changed definition without them)	loop.definition.check	loop.definition_honest
loop_comparison	the metric over at least three turns under the candidate against the same metric under the baseline; its result is improved or not_improved	loop.compare	loop.metric_improved
loop_measurement	the loop scorecard over at least three recorded self.improvement cases at a loop definition revision; figures no validator can observe (escaped defects) enter as human-sourced counts, labelled so; its result is below_target or at_target	loop.measure	loop.metric_below_target
merge_observation	the loop candidate merged into the mainline; its result is merged	loop.merge	loop.change_merged

Claims

Each claim is TRUE, FALSE or UNKNOWN, established from evidence by its predicate. Missing or stale evidence leaves a claim UNKNOWN, never FALSE.

Claim	Description	True when
loop.change_merged	the current loop candidate is merged into the mainline	evidence merge_observation with result merged about loop_candidate
loop.definition_honest	the current loop candidate leaves every metric definition unchanged, or reports the previous definition's figures beside the new ones	any of (evidence definition_check with result dual_reported about loop_candidate; evidence definition_check with result unchanged about loop_candidate)
loop.metric_below_target	a loop metric is below its target on the baseline loop definition	evidence loop_measurement with result below_target about loop_baseline
loop.metric_improved	the metric improved under the current loop candidate, observed by another agent run than the one that delivered it	evidence loop_comparison with result improved about loop_candidate, from a producer independent on different_agent_run

Obligations

What must be done before a case can be complete, and what discharges it. An obligation stays open until its predicate is TRUE; its predicate being UNKNOWN does not discharge it.

Obligation	Description	Discharged when
honesty.required	a loop change never redefines a metric silently	loop.definition_honest is TRUE
improvement.required	a loop change counts only when its metric improved	loop.metric_improved is TRUE

Outcomes

The legitimate terminal interpretations of a case, and what each requires.

Outcome	Description	Requires
abandoned	the loop turn closed because it was explicitly abandoned	decision explicitly_abandoned
accepted	the loop turn closed with an honest change merged and its metric improved	all of (loop.change_merged is TRUE; loop.definition_honest is TRUE; loop.metric_improved is TRUE)
declined	the loop turn closed because the metric did not improve; the candidate is reverted by a further loop turn	evidence loop_comparison with result not_improved about loop_candidate

Artifacts

The things whose revisions evidence is bound to.

Artifact	Description
loop_baseline	the promoted loop definition the metric was measured on
loop_candidate	a changed loop definition meant to raise the metric
loop_finding	a finding against the loop, naming the metric, its target and the turns it was computed from

Independence dimensions

What an evidence match may demand of a record's producer (independent_on). The case lists, for each producer, the dimensions on which it is independent of the case's work; a match that demands a dimension reads only records whose producer the case lists on it. A record that fails the demand neither establishes nor contradicts the match, so a match only such records would decide is UNKNOWN.

Dimension	Description
different_agent_run	the producer is another agent run than the one that delivered the candidate