AutoResearch/Innovation Generator

IDEA DISCOVERY / VERSION LIBRARY

Version library

Each generation’s ideas. Each step of change.

From contribution lenses and persistent search through validation, typed evidence and source-conditioned research action. Start with a series, then open a version to see how it organizes an exploration.

Current: Ariadne v7.5.0

V3 · 从贡献身份开始

Explore this series
V3.x / V3.4

Lens-First Active Generation

This stage organizes research ideation around twelve contribution types, including methods, theory, perspectives and measurement. It fixes the subfield and essential prior work, combines first-principles kernels with transferable anchors, and develops candidates with a clear difference, risk and decisive test.

Read the full introduction

V4 · 把搜索变成持久过程

Explore this series
V4.1 / V4-pro

LA-PATS

V4 expands the one-shot protocol into seven search spaces: domains, contribution lenses, kernels and anchors, mathematics, physics, experiments and reviewer attacks. The scope and budget are fixed first; candidates and refutations become structured state before review, repair and termination.

Read the full introduction
V4.2

Hill

Hill focuses on engineering reliability. A model-call bridge and structural and semantic validation turn missing responses, placeholders and empty refutation ledgers into pauses or explicit invalid states, with review records saved separately.

Read the full introduction
V4.3

Ocean

Ocean assigns tree expansion to specialized model roles for field splitting, kernel mapping, anchor transfer, mechanism construction, experiment design and refutation. Python wraps and records their responses; the mathematics–physics bridge becomes a staged model-call pipeline.

Read the full introduction
V4.4

Stars

Stars repairs advancement after a pause, frontier construction without a literature response and incomplete final-report fields. Operators are scheduled by node type, with visits and attempted actions recorded to make process state explicit.

Read the full introduction
V4.5

Jupiter

Jupiter separates draft, checked, refuted and internally accepted candidate states. Ledger decisions return to the tree and block unsuitable candidates; state changes and events are saved so the tree and final report remain consistent.

Read the full introduction
V4.6

Galaxy

Galaxy accepts both single-node and list model responses, injects kernel and anchor seeds and uses provenance flags to keep seed-only candidates out of the internal acceptance path. Reviewer rejection, candidate–ledger binding and call budgets become runtime constraints.

Read the full introduction
V4.7

Cosmos

Cosmos uses stage-bucket scheduling to connect mechanism, experiment, attack and candidate stages, reducing the starvation of later-stage nodes by exploration scores. It also tightens model-field coverage, ledger identity binding and renewed refutation after repairs.

Read the full introduction
V4.8

Nova

Nova replaces random mathematics–physics triggering with a deterministic depth condition, prioritizes the main candidate path until a first candidate exists and limits bridge branches. Terminal progress and repair-budget guards reduce branch starvation and unchecked repair growth.

Read the full introduction

V5 · 让验证与执行有明确合同

Explore this series
V5.0

Aurora

Makes verification a separate execution layer. Candidate artifacts start unverified, and explicit checks govern promotion, rejection, unresolved evidence and honest failure.

Read the full introduction
V5.2

Eureka

Moves upstream from filtering false originality to organizing candidate discovery: frontier tensions drive typed operators, equation genomes, mathematical objects, definitions and conjectures before Aurora review.

Read the full introduction
V5.3

Eureka

Retains Eureka’s discovery layer while repairing its verification interface: translates discovery objects into Aurora payloads and reconciles every output to one terminal verdict.

Read the full introduction
V5.4

Spark

Routes harvested equation payloads into verification and extends the Equation Lab with canonical forms, a prior-equation library, small numerical probes and physics diagnostics.

Read the full introduction
V5.5

Forge RC

Adds explicit viability and bridge firewalls before idea harvesting. Failed candidates remain legible in structured failure records with repair reasons instead of being packaged as discoveries.

Read the full introduction
V5.5

Forge Final

Wires Forge’s standalone components into the live discovery chain: resolved units, viability records, repair traces, CoMath state and prior caps contribute to one package.

Read the full introduction
V5.6

Origin

Expands from equation-term mutation to the choice of mathematical descriptors. Domain, math and physics search trees produce first-class mappings and structured equation ASTs.

Read the full introduction
V5.7

Genesis

Adds compatibility, execution and quality constraints to Origin: prune mappings, select diverse elites, execute supported AST terms and route proof strategies by conjecture type.

Read the full introduction
V5.8

Logos

Closes the ontology-to-idea evidence chain around one candidate ID. The release focuses on linkage, property provenance and replay rather than adding mappings or solvers.

Read the full introduction

V6 · 组织对象、证据与观察

Explore this series
V6.1

Dedalus

Hardens Archimedes with subprocess timeouts, numeric benchmarks, finite-graph execution, candidate-ID binding and required-route checks.

Read the full introduction
V6.2

EulerForge

Turns candidates into compilable objects: equations use registered operator semantics, algorithms face subprocess tests and fixed benchmarks, and architectures produce executable forward paths, all bound to evidence and prior deltas.

Read the full introduction
V6.3

Grothendieck

Adds a generation-side concept-space layer: explicit axiom cards, twelve T operators, diversity archives, quarantined failures, taste forecasts, reality bets and strategy genomes.

Read the full introduction
V6.4

Lakatos

Turns criticism into executable counterexamples and discriminating experiments, and organizes candidates as research-programme sequences through provenance, re-instantiation, null models, paired bets, anomalies and question genomes.

Read the full introduction
V6.5

Feynman

Organizes research around challenge responses, preregistered predictions and explanatory obligations, adding world-evidence accounting, learning-progress curiosity, prequential meters, explanation checks and world witnesses.

Read the full introduction
V6.6

Noether

Treats reports as evidence indexes: true audits, answer-free sealed challenges, adjudicator-minted settlements, stable hash identities and prosecutor replay must agree, with demonstrations isolated as PASS_DEMO.

Read the full introduction
V6.7

Poincaré

Replaces one-shot prose with ResearchDelta objects that record questions, representations, assumptions, mechanisms, falsifiers and lineage, organizing multi-generation search through isolated populations, multi-niche archives and independent event replay.

Read the full introduction
V6.8

Poincaré

Upgrades the default path to an observation-driven controller: HTTP model providers propose constrained candidates, the operator locks tasks and baselines, execution yields concrete counterevidence and revision obligations, and evidence authorizes representation changes.

Read the full introduction

V7 · 从来源走向下一研究动作

Explore this series
V7.0.0

Aletheia

Inherit v6.8 engineering mechanisms and add primary-text evidence, rival explanations, typed actions, explicit stopping, Fusion contracts and local receipts in a standalone Aletheia module. Synthetic demos and fixed-weight meta evaluation check execution paths.

Read the full introduction
V7.1.0

Aletheia

Add conditional rival updates and a persistent frontier. The macro loop retains branches and deferred actions; the micro loop connects predictions, host attestation, observations and revision contracts, repairing unrelated-observation admission, zero-action quota and silent rival resurrection.

Read the full introduction
V7.2.0

Aletheia

Add an authenticated Codex proposal provider and signed paired pilots. The model selected different correct next measurements under two observations. After same-case field clarification, both proposals passed revision and host contracts. Initial failures, usage and execution-gate repairs remain recorded.

Read the full introduction
V7.3.0

Aletheia

Record the authorized literature-development loop and frozen eight-state A/B/C/D/E comparison, repairing source fields, query context, acquisition receipts, stop closure and observed-action registration. Feedback control is supported; ordinary active design matches the paired result, leaving separate harness gains unestablished.

Read the full introduction

Archives retain the design, implementation and recorded checks of their own version. The current entry is Ariadne v7.5.0; historical material does not automatically describe present capability.