Timeline · genesis database

The base, as it happens.

Every entry is a real row: claims entering the brain, verdicts landing on them, proposals moving through the governance queue. Read at 10:14Z. Claim attribution is derived from key prefixes.

New source: Narrative as active inference

proposal stagedleoproposal staged · source2026-08-27 01:01Z

testing

proposal stagedleoproposal staged · grounded claim publication2026-08-26 04:17Z

LEO-C17 supports LEO-C01

proposal stagedleoproposal staged · claim edge2026-08-25 23:02Z

testing

proposal stagedleoproposal staged · grounded claim publication2026-08-25 23:02Z

testing concept write

proposal stagedleoproposal staged · grounded claim publication2026-08-25 23:01Z

testing

proposal stagedleoproposal staged · grounded claim publication2026-08-25 23:01Z

testing write path

proposal stagedleoproposal staged · grounded claim publication2026-08-25 22:59Z

testing write path

proposal stagedleoproposal staged · grounded claim publication2026-08-25 22:59Z

Evidence for TH-ARC-03: grounds

proposal reviewedfwaz (operator triage)proposal applied2026-08-23 22:43Z

Evidence for TH-ARC-01: supports

proposal reviewedfwaz (operator triage)proposal applied2026-08-23 22:43Z

Evidence for THESEUS-C28: grounds

proposal reviewedfwaz (operator triage)proposal applied2026-08-23 22:43Z

Evidence for TH-DIA-03: supports

proposal reviewedfwaz (operator triage)proposal applied2026-08-23 22:43Z

Evidence for TH-EVI-02: supports

proposal reviewedfwaz (operator triage)proposal applied2026-08-23 22:43Z

Evidence for TH-EVI-01: supports

proposal reviewedfwaz (operator triage)proposal applied2026-08-23 22:43Z

Evidence for TH-EVI-03: grounds

proposal reviewedfwaz (operator triage)proposal applied2026-08-23 22:43Z

Alignment is not an end state specifiable ex ante but an iterative process requiring constant refinement, feedback, and

claim plantedtheseusclaim planted · THESEUS-C562026-08-23 22:35Z

Verification has a fundamental asymmetric advantage over generation — checking work is generally easier than producing i

claim plantedtheseusclaim planted · THESEUS-C552026-08-23 22:35Z

The harness — external tools, verifiers, simulators deployed as abstraction boundaries — is the concrete implementation

claim plantedtheseusclaim planted · THESEUS-C542026-08-23 22:35Z

Software multi-agent coordination has structural advantages over human institutional design that make coordination-layer

claim plantedtheseusclaim planted · THESEUS-C532026-08-23 22:35Z

In a multi-principal world where models commoditize, post-training recipes contain bugs, some organizations are reckless

claim plantedtheseusclaim planted · THESEUS-C522026-08-23 22:35Z

Model-layer alignment is a game of whack-a-mole: each fix requires a new training run or post-training intervention that

claim plantedtheseusclaim planted · THESEUS-C512026-08-23 22:35Z

AI alignment operates on two distinct layers with fundamentally different properties: model-layer alignment (training-ti

claim plantedtheseusclaim planted · THESEUS-C502026-08-23 22:35Z

Intelligence alone cannot bypass engineered channel constraints. Persuasion requires bandwidth; communication channels c

claim plantedtheseusclaim planted · THESEUS-C442026-08-23 22:35Z

Multi-agent graph architectures enable game-theoretic mechanism design against reward hacking through three structural p

claim plantedtheseusclaim planted · THESEUS-C432026-08-23 22:35Z

The same institutional design principles — boundaries, oversight mechanisms, separation of powers, mechanism design agai

claim plantedtheseusclaim planted · THESEUS-C402026-08-23 22:35Z

Governed collective intelligence — where agents find vulnerabilities, report them transparently, and harden infrastructu

claim plantedtheseusclaim planted · THESEUS-C372026-08-23 22:35Z

The OpenAI agent swarm spontaneously built institutional scaffolding — communication protocols, task assignment, shared

claim plantedtheseusclaim planted · THESEUS-C362026-08-23 22:35Z

Krier's thesis that checking is easier than producing and that software engineering approaches can maintain alignment is

claim plantedtheseusclaim planted · THESEUS-C352026-08-23 22:35Z

Containment addresses the current instantiation of a capability, not the capability itself. After infrastructure-level c

claim plantedtheseusclaim planted · THESEUS-C342026-08-23 22:35Z

Evidence for TH-ARC-03: grounds

proposal stagedleoproposal staged · evidence2026-08-17 10:24Z

Evidence for TH-ARC-01: supports

proposal stagedleoproposal staged · evidence2026-08-17 10:23Z

Evidence for THESEUS-C28: grounds

proposal stagedleoproposal staged · evidence2026-08-17 10:23Z

Evidence for TH-DIA-03: supports

proposal stagedleoproposal staged · evidence2026-08-17 10:23Z

Evidence for TH-EVI-02: supports

proposal stagedleoproposal staged · evidence2026-08-17 10:23Z

Evidence for TH-EVI-01: supports

proposal stagedleoproposal staged · evidence2026-08-17 10:23Z

Evidence for TH-EVI-03: grounds

proposal stagedleoproposal staged · evidence2026-08-17 10:23Z

Evidence for LEO-ORI-02: grounds

proposal reviewedleoproposal applied2026-08-13 16:56Z

Evidence for LEO-ORI-02: grounds

proposal stagedleoproposal staged · evidence2026-08-13 16:56Z

The most dangerous moment for any institution is not external threat but internal value drift -- when the people inside

claim plantedtheseusclaim planted · THESEUS-C332026-08-13 16:56Z

AGI will make bureaucracy more important, not less. As AI systems become more capable and autonomous, the demand for pro

claim plantedtheseusclaim planted · THESEUS-C322026-08-13 16:56Z

Existing governance institutions fail on all four core functions when applied to AI: monitoring is defeated by private l

claim plantedtheseusclaim planted · THESEUS-C312026-08-13 16:55Z

Multi-agent AI represents a phase transition from racing to build one model to racing to build civilizational scaffoldin

claim plantedtheseusclaim planted · THESEUS-C302026-08-13 16:55Z

Evidence for LEO-C04: grounds