AGASTYA NATIVE BRAIN • PUBLIC PROVENANCE & VERIFICATION

PROVING THE NATIVE LINEAGE.

Agastya AGI v01 public evidence release R1. A DOI-backed evidence trail for native corpus, tokenizer, architecture, random initialization, checkpoint identity and independent verification—presented inside the same Zippri visual environment as the main product universe.

R1 PUBLIC EVIDENCE LIVE DOI 10.5281/zenodo.23054425 V1–V6 PROVENANCE EXTERNAL VERIFICATION NEXT
Corpus
Tokenizer
Architecture
Random Init
Checkpoint
Verification
Agastya native intelligence evidence
AGASTYANATIVE BRAIN
R1PUBLIC EVIDENCE RELEASE
Provenance → validation → independent reviewEvaluator portal ↗
AGASTYA NATIVE BRAIN EVIDENCE PROVENANCE CRYPTOGRAPHICALLY ANCHORED VALIDATION R1 MANIFESTS PUBLISHED VERIFICATION INDEPENDENT EXTERNAL REVIEW REQUESTED PUBLICATION DOI-BACKED RELEASE AGASTYA NATIVE BRAIN EVIDENCE PROVENANCE CRYPTOGRAPHICALLY ANCHORED VALIDATION R1 MANIFESTS PUBLISHED VERIFICATION INDEPENDENT EXTERNAL REVIEW REQUESTED PUBLICATION DOI-BACKED RELEASE
PUBLIC EVIDENCE SYSTEM

Evidence at a glance.

These cards summarize the current published R1 provenance package without changing the evidence scope or turning pending external gates into verified claims.

1,264,715,776Unique parameters
98,320Tokenizer vocabulary
2,529,519,595Step-0 checkpoint bytes
10.5281/zenodo.23054425Permanent R1 DOI
CANONICAL NATIVE PATH

From source to proof.

The public evidence trail mirrors the native program structure while keeping proprietary raw artifacts private.

NATIVE LINEAGE

Canonical from-scratch chain

Agastya corpus → Agastya tokenizer → Agastya architecture → random initialization → pretraining → post-training → Native Brain.

1
Corpus & DataSource admission, processing, quality, language/domain coverage and mixture calibration.
2
Native Tokenizer98,320-vocabulary BPE tokenizer with documented provenance.
3
ArchitectureNative dense architecture and exact parameter-count evidence.
4
Random InitializationFresh step-0 weights with no inherited third-party model weights.
VALIDATION & REVIEW

Evidence becomes independent proof

R1 makes the verification sequence explicit: published manifests first, independent evaluator replay next, then same-suite benchmark evidence.

V1
Artifact IdentityConfirm the reviewed release and checkpoint target.
V2
InitializationReview fresh random-init evidence.
V3
Weight AncestryInspect provenance and third-party ancestry controls.
V4–V6
Tokenizer → Count → ContinuityComplete the remaining lineage and identity checks.
INDEPENDENT VERIFICATION R1

The next external gate.

The public evidence release is the starting point; external verification is a separate evaluator activity and is not represented as completed here.

01 / PUBLIC EVIDENCE

Published

R1 evidence, manifests, checksums and citation metadata are publicly accessible.

02 / EVALUATOR

Independent review requested

V1–V6 provenance scope is packaged for an external evaluator.

03 / SAME-SUITE

Pending frozen replay

Comparable benchmark results remain pending an identical frozen evaluation harness and independent replay.

RESEARCH ACCESS

One place for every artifact.

Human-readable pages for researchers, with raw files preserved behind each page.