Independent evaluation will determine whether Orivael’s non-LLM reasoning system can reproduce its perfect ft09 result and generalize across additional ARC-AGI-3 task families.

state: expiredheat: lowuncertainty: highknownscott: lowagent-benchmarks symbolic-reasoning arc-agiOrivael

What is this?

Orivael claims a non-LLM reasoning system achieved 100% on the ARC-AGI-3 ft09 task without any model calls, with independent reproduction and broader generalization still unestablished. ARC-AGI-3 is an interactive benchmark that tests agents’ ability to explore unfamiliar environments, infer goals and mechanics, form world models, and plan efficiently through action and feedback. The supplied web results describe the benchmark and its competition infrastructure, but they do not independently document Orivael’s system, methods, result, or planned evaluation.

Why it matters to Scott

The radar already tracks this same ARC-AGI-3 claim-validation pattern in the Schema and Seed IQ cases, including the need for independent reproduction and cross-environment generalization. It aligns with Scott’s evidence-ceiling and replayable-evaluation doctrines, but without methods, evaluator results, or broader task performance, it is currently another unverified benchmark claim rather than something that changes what he builds or argues.

ip:framework.discussed-is-not-deployedip:framework.challenger-never-arbiterradar:schema-arc-agi-3-claimradar:seed-iq-arc-agi-3d-doomradar:concept.arc-agiradar:concept.benchmark-integrityradar:concept.agent-benchmarks

queries asked of Scott's wikis

non-LLM agent architectures and symbolic reasoning
benchmark reproduction and independent evaluation
ARC-style abstraction and interactive reasoning
agent world models from action and feedback
benchmark overfitting versus cross-task generalization
model-free agents and inference economics

Measured heat

no measured readings yet — the hourly heat pass fills this in

How the heat travelled

no chain yet — the hourly chain pass fills this in

Evidence (1) — ⭐ canonical anchor

source	object	author	score	comments
🟠 reddit ⭐	We got 100% on ARC-AGI-3 ft09 with zero model calls. The failures are more interesting. artificial	Living_Substance1274	0	13

Interpretation history

2026-08-12T18:33:29Z
No independent evaluation, methods release, or cross-family result arrived within the case horizon, leaving the narrow self-reported ft09 score unchanged and no longer worth active monitoring.
2026-08-10T17:39:46Z
The author now clarifies that Claude models participated during development before the runtime path became model-free, narrowing the claim from wholly non-LLM provenance to zero model calls at execution. This adds useful architectural context but no reproducible method, independent evaluation, or cross-family evidence.
2026-08-10T00:35:39Z
Refreshed comments remain repetitive skepticism and architecture questions, adding no methods, reproduction, independent evaluation, or cross-family performance; the claim’s evidentiary status is unchanged.
2026-08-09T21:37:15Z
Comment thread refreshed with skeptical replies ('slop', questions about representation/architecture) but no methods disclosure, reproduction, or independent evaluation — case remains an unverified single-family self-reported claim.
2026-08-09T20:29:27Z
The added discussion provides no independent reproduction, methods, or cross-family results, so the case remains a narrow self-reported benchmark success rather than evidence of general non-LLM reasoning.
2026-08-09T20:26:54Z
grounded: known/low — The radar already tracks this same ARC-AGI-3 claim-validation pattern in the Schema and Seed IQ cases, including the need for independent reproduction and cross
2026-08-09T20:23:17Z
case created — The linked ARC Prize scorecard makes the zero-model-call result concrete, but it currently covers only one task family and lacks independent validation.

Decision trace

08-13 04:33expireNo independent evaluation, methods release, or cross-family result arrived within the case horizon, leaving the narrow self-reported ft09 score unchanged and no longer worth active monitoring.
08-13 04:33alert_silentThe only delta is elapsed time without validation or a new event; there is nothing Scott needs before the next briefing.
08-13 04:33alert_routeThe only delta is elapsed time without validation or a new event; there is nothing Scott needs before the next briefing.
08-11 03:39repriceThe author now clarifies that Claude models participated during development before the runtime path became model-free, narrowing the claim from wholly non-LLM provenance to zero model calls at executi
08-11 03:39alert_silentThe development-path clarification changes how the claim should be framed but does not validate the reported score or generalization, so it can wait for independent reproduction, implementation detail
08-11 03:39alert_routeThe development-path clarification changes how the claim should be framed but does not validate the reported score or generalization, so it can wait for independent reproduction, implementation detail
08-11 03:21sensor_dirtycomment_update
08-10 10:35repriceRefreshed comments remain repetitive skepticism and architecture questions, adding no methods, reproduction, independent evaluation, or cross-family performance; the claim’s evidentiary status is unch
08-10 10:35alert_silentThe new delta is only discussion churn without substantive evidence, so it can wait for an independent evaluator result, reproducible implementation, or broader task-family results.
08-10 10:35alert_routeThe new delta is only discussion churn without substantive evidence, so it can wait for an independent evaluator result, reproducible implementation, or broader task-family results.
08-10 10:21sensor_dirtycomment_update
08-10 07:37repriceComment thread refreshed with skeptical replies ('slop', questions about representation/architecture) but no methods disclosure, reproduction, or independent evaluation — case remains an unv
08-10 07:37alert_silentOnly comment churn with skepticism, no new substantive evidence; nothing actionable before next briefing.
08-10 07:37alert_routeOnly comment churn with skepticism, no new substantive evidence; nothing actionable before next briefing.
08-10 07:21sensor_dirtycomment_update
08-10 06:29repriceThe added discussion provides no independent reproduction, methods, or cross-family results, so the case remains a narrow self-reported benchmark success rather than evidence of general non-LLM reason
08-10 06:29alert_silentOnly comment count changed; no consequential new evidence arrived, so this can wait for an evaluator result, reproducible implementation, or broader task performance.
08-10 06:29alert_routeOnly comment count changed; no consequential new evidence arrived, so this can wait for an evaluator result, reproducible implementation, or broader task performance.
08-10 06:27alert_silentA self-reported run links to a scorecard showing perfect ft09 performance without model calls, but the same system performs poorly across the other disclosed task families and provides no reproducible
08-10 06:27surface_candidateA self-reported run links to a scorecard showing perfect ft09 performance without model calls, but the same system performs poorly across the other disclosed task families and provides no reproducible
08-10 06:27alert_routeA self-reported run links to a scorecard showing perfect ft09 performance without model calls, but the same system performs poorly across the other disclosed task families and provides no reproducible
08-10 06:26groundThe radar already tracks this same ARC-AGI-3 claim-validation pattern in the Schema and Seed IQ cases, including the need for independent reproduction and cross-environment generalization. It aligns w
08-10 06:23createThe linked ARC Prize scorecard makes the zero-model-call result concrete, but it currently covers only one task family and lacks independent validation.