Research

Follow the evidence boundary, not a completion meter.

Each program starts with a question, selected objects, falsifiers, protocols, and an explicit publication boundary. Results appear only when the declared evidence and review gates move.

Programs
2
Model results
None yet
Completion meters
None
Local utility
Research Workbench

How to read these

Publication boundary

Each program states what may be said now, what changed most recently, and the exact gate that must move next. A listed proposal is not a result.

Atlas-64: Verifier-Guided Residual Expert Compilation

Proposal · no model result

Libraries source-verified; private Qwen structural screens complete; model execution blocked. No model has been selected, constructed, trained, inferred, or benchmarked.

FieldValue
IdentifierATLAS-64
Evidence rungSource and structural evidence only
Latest updateScrubbed landscape projection dated 2026-07-18; result state remains no model result.
Exact next gatePreregister the next Qwen private structural experiment before reading another weight value. The primary and official-parent libraries remain independently verified at 266,806,166,050 combined bytes. The general official-parent source packet still allows 0 of 16 actions and human legal review remains incomplete, while a separate Qwen-only engineering receipt admits only private storage and private weight statistics, 2 of 8 actions. Under that narrow exception, the 26-shard/1,045-tensor header inventory, one 1,073,741,824-byte tensor pilot, its 1,304-gate same-host independent implementation replay, and a 92-tensor/164-range broader structural-precision screen are complete. The broader screen processed 529,563,648 BF16 bytes across 13 shards after 12 of 12 pre-source tests, peaked at 258,720,248 coalition bytes, and caused zero swap growth. Its initial no-output operational failure and sole preflighted retry are recorded. Every metric, hypothesis disposition, retained-candidate identity, method ranking, and numeric delta remains withheld; no model was constructed or loaded, no inference ran, and no derivative was written. Next freeze a matched-global-storage comparison of compressed scale metadata and role-aware precision/group allocation against uniform controls, with exact bit accounting and no derivative write. Exact project-quantization rights and functional fidelity, target runtime admission, clean execution commit, canonical non-benchmark fixtures, prompt adapters, native model-memory receipts, local inference, training, evaluation, redistribution, and publication remain separate gates.

Attention serving systems

Methods track · no benchmark claim

A public learning and methods track for cache, batching, and serving tradeoffs. It does not claim production superiority or a completed institutional evaluation.

FieldValue
IdentifierATTN-SERVING
Evidence rungConcept and protocol development
Latest updateThe public path connects mechanisms to bounded evaluation tools; no program-level model result is published here.
Exact next gateAttach a governed experiment card and receipt-bearing evidence before promoting any performance claim.

Prepare a portable research bundle

Draft object references, a falsifier, evidence metadata, review notes, and a proposed projection in one deterministic JSON file. The workbench stays browser-local and cannot approve or execute anything.

Open the Research Workbench.