The Breakthrough ingestion contract
Every app in the Prepare 4 VC Suite writes founder evidence into Oasis. This is the shared kernel that keeps that evidence honest: what counts as an observation, what is only a claim, and who is allowed to say a Breakthrough happened.
Full source: docs/canon/OASIS-BREAKTHROUGH-SOURCE-CODE-v1.0.md
Distinctions that must never collapse
FOUNDER != SCORE OBSERVATION != CLAIM != FACT != BREAKTHROUGH BREAKTHROUGH_CASE != CONFIRMED_BREAKTHROUGH SESAC != TAJ != SIX_NEEDS FOUNDER_STATE != FOUNDER_EVOLUTION != BREAKTHROUGH SELF_REPORT != OBSERVED_BEHAVIOR NOT_OBSERVED != OBSERVED_ABSENCE DERIVED_EVIDENCE != INDEPENDENT_EVIDENCE TEMPORAL_SEQUENCE != CAUSATION
An implementation that cannot preserve these distinctions is not compliant with the standard, regardless of how good its dashboards look.
A Breakthrough, in one sentence
A Breakthrough is a consequential reality-alignment shift that changes how a founder operates and survives contact with subsequent evidence.
Formally: an evidence-supported transition in which a founder's operative model of a consequential reality becomes materially better aligned with the strongest available evidence, produces meaningful operational expression, and remains observable beyond the initiating context while remaining revisable.
Four objects, in order
REAL WORLD / FOUNDER ACTIVITY
|
v
OBSERVATION immutable, provenance-bearing record
| N:M
v
CLAIM falsifiable interpretive proposition
| N:M
v
BREAKTHROUGH CASE governed investigation container
|
v
ADJUDICATION governed evaluation against requirements
|
v
CASE STATUSIdentity, time, provenance, derivation, uncertainty, contradiction, version and limitations wrap all four. They are kernel services, not optional decoration.
Status describes evidence, never a founder
CANDIDATE -> OPEN |-- insufficient evidence ---------> UNRESOLVED |-- material evidentiary conflict -> CONTESTED |-- evidence does not support -----> NOT_SUPPORTED v PROVISIONALLY_SUPPORTED | reality alignment + consequentiality + operational | expression + subsequent evidence + persistence supported v CONFIRMED -> (later material evidence) -> SUPERSEDED
Insufficient warrant — not false.
Evidence materially supports incompatible propositions.
Case unsupported by available evidence — not founder failure.
Strongest warranted current judgment — revisable.
What producers must send
- SELF_REPORT
- BEHAVIOR
- DECISION
- ACTION
- ARTIFACT
- COMMUNICATION
- EXTERNAL_RESULT
- THIRD_PARTY_OBSERVATION
- LONGITUDINAL_CHANGE
- OBSERVED_ABSENCE
- HUMAN_OBSERVED
- SYSTEM_OBSERVED
- SELF_REPORTED
- THIRD_PARTY_REPORTED
- COMPUTATIONALLY_DERIVED
- OBSERVED
- REPORTED
- INFERRED
- UNKNOWN
Application names are never evidence types. Send source_system: RATE_MY_PITCH with evidence_type: BEHAVIOR — never RATE_MY_PITCH_EVIDENCE. Claim bearing is one of SUPPORTING, CONTRADICTING, CONTEXTUAL, AMBIGUOUS. Constructs are referenced explicitly: SESAC.SELF_EFFICACY, SESAC.SITUATIONAL_AWARENESS, SESAC.CLARITY, TAJ.TRUST, TAJ.ANTICIPATION, TAJ.JOY, NEED.*
Seven rules the hub enforces
Observations are appended, never rewritten.
Measurement is not interpretation.
Competing claims survive together.
No app confirms a Breakthrough on its own.
Storing the record grants no authority.
The governed protocol decides, not the loudest system.
CONFIRMED can become SUPERSEDED.
Provenance law: three derived observations are not three independent observations. Oasis must not manufacture confidence by circulating one observation through several applications. Contradiction law: if Compass records a founder reporting they abandoned an ICP and Rate My Pitch later observes them defending it, both survive — no summarizer deletes the conflict for a cleaner story. Limitation law: every producer states what it cannot know.
What each app may and may not do
May: Emit self-report and diagnostic observations, set baselines, propose SESAC / TAJ / Six Needs claims, nominate cases.
Must not: Convert self-report into observed fact, or declare a Breakthrough from a score or a single assessment.
May: Emit observed pitch behavior, adaptation behavior, communication observations and artifacts; compare longitudinally.
Must not: Assume pitch-context behavior generalizes, or equate polish with Clarity, confidence with Self-Efficacy, excitement with Joy.
May: Emit communication and relationship-behavior observations and longitudinal outreach change.
Must not: Infer internal transformation from message style. OBSERVED_COMMUNICATION != INFERRED_MOTIVATION.
May: Emit interventions, facilitator observations, founder reports, artifacts and longitudinal program evidence.
Must not: Imply causation from sequence. INTERVENTION_BEFORE_CHANGE != INTERVENTION_CAUSED_CHANGE; causal attribution defaults to UNKNOWN.
May: Preserve identity, observations, the provenance graph, claims, cases, adjudications, contradictions and supersession history.
Must not: Become an oracle merely because it stores the evidence. CUSTODY != ADJUDICATION.
AI proposes, staff confirm
Oasis agents may nominate Breakthrough cases, assemble evidence and draft requirement findings — SUPPORTED, UNRESOLVED, CONTESTED or NOT_SUPPORTED across reality alignment, consequentiality, operational expression, subsequent evidence and persistence. A draft adjudication is never a determination. Only Prepare 4 VC staff sign an adjudication that moves a case to CONFIRMED, and every signed adjudication remains preserved when a later one supersedes it.
Agents preserve UNKNOWN rather than fabricate completeness, never infer a Breakthrough from scores, emotional state, evolution phase or external success, and never produce a numerical Breakthrough probability.
Program scores are not Breakthrough determinations
The Breakthrough Program's entry and current program scores and the pressure response arc are legacy program-progress indicators. They describe cohort movement inside a program window. They carry no Breakthrough authority, cannot open or confirm a case, and are never presented as evidence of transformation.
Writing evidence into Oasis?
The hub API, scopes and OIDC flow live in the developer guide. This contract governs the payloads you send through it.