WEKA is a storage software company that has renamed its product for the inference era: NeuralMesh (documentation version 5.1) is the WEKA Data Platform, and around it sit NeuralMesh Axon (the filesystem converged onto GPU servers), WEKApod appliances, the Augmented Memory Grid (key-value cache tiering to NVMe for inference engines), and, in 5.1, a Data Catalog, a WEKA App Store, and an AI Data Platform blueprint. The map reads it strong at one layer, moderate at three, and gap at six. Storage is the deepest cell, strong at 0C: one multi-protocol namespace with a native S3 service (object lock, versioning, audit) and encryption under the enterprise's key management system (KMS). Layer 1A reads moderate on a Data Catalog that indexes and queries filesystem metadata at scale and tenants two ways (Organizations and hardware-isolated Composable Clusters), with no classification or lineage. 0C is strong on software-defined storage over servers the enterprise picks from WEKA's support matrix or clouds it rents; 0A and 0B are gaps. Layer 1B is a gap: the AI Data Platform, a filesystem watcher feeding NVIDIA NeMo Retriever and Milvus, has no documented support level, and the NVIDIA ingestion pipeline inside it isn't supported under NVIDIA AI Enterprise. Layer 1C is moderate on object-store tiering (with per-path overrides) and snap-to-object; the Augmented Memory Grid, generally available to NeuralMesh customers since WEKA's March 16, 2026 release, is scored there as KV cache tiering. Layer 2A is moderate on the WEKA Operator with node-selector pins, the Cloud Deployment Manager, Terraform modules, and an open CSI driver, with no compute scheduling. Layer 2B is a gap: the AI Data Platform's agentic RAG surface is NVIDIA's reference blueprint behind WEKA's gateway, with no documented support level. Layer 2C is a gap: the platform's identity, gateway, and traces are application-scoped legs. Layer 3 is a gap: the App Store's documented catalogue is one data pipeline.
The capture is at the filesystem and nowhere below it. NeuralMesh, its client and data path, the Data Catalog, the Operator, the installers, the App Store, and NeuralMesh Observe are WEKA's; that's why 13 of the 19 components read Ceded. What lifts: the servers, drives, and adapters the enterprise buys from any vendor on WEKA's matrix; the NFS and SMB doors and a subset of S3 (buckets, policies, versioning, object lock); and the open-source CSI driver at the Kubernetes standard. GPUDirect Storage, NIXL, and Dynamo are the inference stack the enterprise brings, not a dependency the filesystem carries.
The buyer's trade: a parallel filesystem on hardware it chooses, in clouds it already rents, that can describe its own metadata and isolate tenants it can bill, in exchange for the filesystem's layout, catalog, tenancy, tiering, and control plane being WEKA's. The decision-authority readings follow: vendor / Ceded at 0C (NeuralMesh stripes per policies the enterprise set); code / Retained at 1A, where the catalog and tenant isolation execute what the enterprise configured; vendor / Delegated at 1C and 2A on documented overrides (per-path tier fetch and release; node selectors and tolerations on the enterprise's own Kubernetes); Absent at 1B, 2B, 2C, and Layer 3. What would move cells: NeuralMesh 6's replication in the second half of 2026 (1C, an argument for strong with the Augmented Memory Grid), classification in the Data Catalog (1A, to strong), a WEKA-fronted generation endpoint with model choice (2B), documented applications in the App Store (Layer 3), and WEKApod 3 shipping in Fall 2026 (nothing moves; watch-listed).
Layer-by-layer status: Layer 0A (No Compute on WEKA's Paper), Layer 0B (No Fabric on WEKA's Paper), Layer 0C (NeuralMesh on Servers You Choose, in Clouds You Already Rent), Layer 1A (A Catalog of Filesystem Metadata, With Tenant and Access Controls; No Classification, No Lineage), Layer 1B (AI Data Platform Unscored: an App Store Application With No Documented Support Level, Around NVIDIA's Reference RAG Pipeline), Layer 1C (Object-Store Tiering and Snap-to-Object for Cost-Aware Movement, and the Augmented Memory Grid's KV Cache Tiering (Generally Available March 2026); NeuralMesh 6 Replication Pending), Layer 2A (The WEKA Operator and Cloud Deployment Manager Run NeuralMesh on the Enterprise's Kubernetes and Clouds (Custom Resources, Rolling Upgrades, Auto-Scaling Groups, Composable Tenants); a Standard CSI Door; Node Selectors and Tolerations Pin Where WEKA Runs; No Compute or GPU Scheduling), Layer 2B (AI Data Platform Agentic RAG Unscored: NVIDIA's Reference RAG Server Behind WEKA's Gateway, No Documented Support Level; No General Serving, No Agent Runtime), Layer 2C (Tenant Isolation and Permission-Aware Retrieval Govern Data, Not Agents; Keycloak Inside AIDP Identifies Users; Not a Plane), Layer 3 (+1) (No Application: the App Store Deploys One Documented Blueprint, the AI Data Platform, Which Is a Data Pipeline; the Catalogue's Breadth Is Undocumented).
Assessment framework: 4+1 Layer AI Infrastructure Model. Scoring model: Decision Authority Placement Model (DAPM) — Retained, Delegated, or Ceded. Published by The CTO Advisor LLC (DBA The Advisor Bench). Author: Keith Townsend. Date assessed: October 27, 2026. Version: v1.18 - 4+1 v3: Layer 0 Sub-Layers.
WEKA is a storage software company that has renamed its product for the inference era: NeuralMesh (documentation version 5.1) is the WEKA Data Platform, and around it sit NeuralMesh Axon (the filesystem converged onto GPU servers), WEKApod appliances, the Augmented Memory Grid (key-value cache tiering to NVMe for inference engines), and, in 5.1, a Data Catalog, a WEKA App Store, and an AI Data Platform blueprint. The map reads it strong at one layer, moderate at three, and gap at six. Storage is the deepest cell, strong at 0C: one multi-protocol namespace with a native S3 service (object lock, versioning, audit) and encryption under the enterprise's key management system (KMS). Layer 1A reads moderate on a Data Catalog that indexes and queries filesystem metadata at scale and tenants two ways (Organizations and hardware-isolated Composable Clusters), with no classification or lineage. 0C is strong on software-defined storage over servers the enterprise picks from WEKA's support matrix or clouds it rents; 0A and 0B are gaps. Layer 1B is a gap: the AI Data Platform, a filesystem watcher feeding NVIDIA NeMo Retriever and Milvus, has no documented support level, and the NVIDIA ingestion pipeline inside it isn't supported under NVIDIA AI Enterprise. Layer 1C is moderate on object-store tiering (with per-path overrides) and snap-to-object; the Augmented Memory Grid, generally available to NeuralMesh customers since WEKA's March 16, 2026 release, is scored there as KV cache tiering. Layer 2A is moderate on the WEKA Operator with node-selector pins, the Cloud Deployment Manager, Terraform modules, and an open CSI driver, with no compute scheduling. Layer 2B is a gap: the AI Data Platform's agentic RAG surface is NVIDIA's reference blueprint behind WEKA's gateway, with no documented support level. Layer 2C is a gap: the platform's identity, gateway, and traces are application-scoped legs. Layer 3 is a gap: the App Store's documented catalogue is one data pipeline.
The capture is at the filesystem and nowhere below it. NeuralMesh, its client and data path, the Data Catalog, the Operator, the installers, the App Store, and NeuralMesh Observe are WEKA's; that's why 13 of the 19 components read Ceded. What lifts: the servers, drives, and adapters the enterprise buys from any vendor on WEKA's matrix; the NFS and SMB doors and a subset of S3 (buckets, policies, versioning, object lock); and the open-source CSI driver at the Kubernetes standard. GPUDirect Storage, NIXL, and Dynamo are the inference stack the enterprise brings, not a dependency the filesystem carries.
The buyer's trade: a parallel filesystem on hardware it chooses, in clouds it already rents, that can describe its own metadata and isolate tenants it can bill, in exchange for the filesystem's layout, catalog, tenancy, tiering, and control plane being WEKA's. The decision-authority readings follow: vendor / Ceded at 0C (NeuralMesh stripes per policies the enterprise set); code / Retained at 1A, where the catalog and tenant isolation execute what the enterprise configured; vendor / Delegated at 1C and 2A on documented overrides (per-path tier fetch and release; node selectors and tolerations on the enterprise's own Kubernetes); Absent at 1B, 2B, 2C, and Layer 3. What would move cells: NeuralMesh 6's replication in the second half of 2026 (1C, an argument for strong with the Augmented Memory Grid), classification in the Data Catalog (1A, to strong), a WEKA-fronted generation endpoint with model choice (2B), documented applications in the App Store (Layer 3), and WEKApod 3 shipping in Fall 2026 (nothing moves; watch-listed).
CPUs, accelerators, servers, integrated systems, OS and hypervisor, physical plant
Assessment pending
WEKA sells no compute. NeuralMesh Axon converges storage onto the enterprise's own GPU servers, but the servers and accelerators are the enterprise's purchase from any vendor. Calibration: DDN, NetApp, VAST and Everpure read gap here too; Dell and HPE sell storage and compute together.
Nothing offered and nothing inherited from WEKA at 0A: Absent.
Interconnects, switching and routing, network virtualization
Assessment pending
WEKA sells no switches. The InfiniBand or Ethernet fabric is the enterprise's choice, within WEKA's qualified-adapter matrix, and WEKApod 3's Spectrum-X networking isn't on WEKA's paper. Calibration: VAST and NetApp read moderate here because they sell cluster switches; DDN and Everpure read gap.
Absent: the fabric is the enterprise's own purchase.
Watch-listed: WEKApod 3, 'delivery beginning in Fall 2026'. If WEKA's paper carries the Spectrum-X switches at delivery, 0B gains a chip.
Arrays, storage servers and software, front-end protocols: where the data sits
WEKA's filesystem software, its client, its data path, and its installers, on hardware or cloud instances the enterprise owns. Ceded, the VAST and NetApp reading; the cloud substrate is scored on the cloud's row.
Hardware the enterprise buys from any vendor on WEKA's support list and runs NeuralMesh on. Retained, the Nutanix multi-vendor hardware reading; the matrix narrows the choice without owning it.
WEKA-engineered systems with NeuralMesh preinstalled, sold through distributors. Ceded; the third generation is watch-listed, not scored.
Standards the enterprise's hosts speak; the filesystem behind them is WEKA's, and the native WEKA client is WEKA's driver over a proprietary data path, scored in the software chip. Delegated, the NetApp storage-networking reading. GPUDirect Storage is NVIDIA's cuFile stack and sits in the NVIDIA note.
The storage half of the filesystem chip: WEKA's filesystem and protocol implementations in WEKA's layout. Ceded, the NetApp ONTAP reading.
The storage half of the S3 chip: the S3 front end over the same data. Delegated, the StorageGRID and ObjectScale reading, with the unsupported list named on the 1A chip.
The substrate half of the tenancy chip: per-tenant clusters partitioning drives, processors and memory through an SSD proxy.
Filesystem quotas and encryption at rest with the enterprise's KMS: capacity and protection of the bytes. Ceded, as the chip reads.
S3 Object Lock retention on WEKA buckets: immutability of the bytes.
Per-tenant SSD and total capacity quotas and QoS caps on throughput and IOPS: capacity and performance of the substrate.
NeuralMesh's GPU-direct read path uses NVIDIA's GDS stack; the filesystem itself runs on Intel, AMD or Grace CPUs.
WEKA sells the storage software and lets the enterprise pick the hardware under it. NeuralMesh, the WEKA Data Platform renamed and documented at version 5.1, runs on supported Intel, AMD or NVIDIA Grace servers with dedicated NVMe and qualified adapters, eight servers minimum. It also runs in the enterprise's own AWS, Azure, GCP and OCI accounts. It also converges onto GPU servers as NeuralMesh Axon and ships as WEKApod appliances. One parallel namespace serves POSIX, NFS, SMB and S3, with a GPUDirect Storage path. The buyer gets AI-throughput storage on whatever it already buys. The architect's concern is that the software is WEKA's alone: a proprietary client and data path, a qualified-adapter matrix, and an installer limited to four OEMs. Calibration: DDN, VAST and NetApp read strong on the same function.
Ceded where the grade rests. NeuralMesh stripes, places and rebuilds data across the servers the enterprise gave it; protection is fixed for the cluster, and tiering policy lives on the filesystem group and covers tier only: vendor decides, visible, not overridable, Ceded. Portability: the enterprise's servers read Retained, NeuralMesh and WEKApod Ceded, and NFS, SMB and S3 Delegated.
Watch-listed: NeuralMesh 6, 'generally available in the second half of 2026'. Named, not scored: the Andromeda partnership (July 30, 2026), a channel. WEKApod Appliances: WEKApod 3 is "available to order today ... for delivery beginning in Fall 2026"; the delivery window has opened, which reads GA under the table.
The data itself, organized and governed: the Governance Catalog that Layer 2C queries
WEKA's filesystem and protocol implementations in WEKA's layout, and the management surfaces around its S3 door. Ceded, the NetApp ONTAP reading; the keys are the enterprise's behind standard KMS interfaces.
A subset of the S3 interface over the same data; buckets, policies, versions, and locks lift to any S3 store, though server-side encryption, replication, ACLs, CORS, S3 Select, and lifecycle transitions are unsupported. Delegated, the StorageGRID and ObjectScale reading, with the gaps named.
WEKA's index of WEKA's filesystem metadata, queryable through WEKA's API. Ceded, the NetApp Metadata Engine and Dell MetadataIQ reading; it indexes paths, sizes, and ages, not content or sensitivity.
WEKA's authority and isolation models inside WEKA's cluster. Ceded.
Filesystems, protocols, tenancy, and encryption run on the CPUs and NVMe of whatever servers the enterprise chose.
WEKA's 1A is now the part of NeuralMesh that knows about the data and who may reach it. The Data Catalog (5.1 documentation) detects changes through rolling snapshots and keeps filesystem metadata in an embedded query engine. It answers questions like which directories grew more than 50 GB in a day, or which files sat untouched for 90 days. It indexes metadata, not content, and does no classification, sensitivity tagging or lineage. Access sits beside it: OIDC and LDAP authentication, IAM-style S3 policies and STS credentials, Organizations with their own admins, and an audit webhook. Quotas, encryption and Object Lock protect the bytes and read at 0C, and NeuralMesh Observe is fleet observability at 2A. Under the catalog criterion a metadata-only index answers where and how big, not what. Moderate, at the top of the band.
The Data Catalog indexes the filesystems the administrator points it at, executing without inference: code decides, visible, overridable, Retained, the Dell MetadataIQ reading. Portability: the index and query engine are WEKA's alone, Ceded; the S3 policy surface is a multi-vendor subset, Delegated.
Public evidence that moves the cell: classification or lineage in the catalog.
Low-latency retrieval for RAG — vector/hybrid search, context windows
The default profile 'requires 12 RTX PRO 6000 SE GPUs (96 GB each) across 3 nodes'; files are 'parsed, chunked, and embedded using NVIDIA NIM models', the stack includes 'the NVIDIA Nemo Retriever (NIMs, Milvus, and the ingestion pipeline)', and an NVIDIA NGC key pulls the images and weights. The retrieval engine is NVIDIA's on NVIDIA GPUs; WEKA owns the watcher, the operator, and the permission model.
WEKA's retrieval path is the WEKA AI Data Platform (AIDP), an application in the WEKA App Store, and its documentation doesn't show it as generally available. The installation guide describes the full stack as Keycloak identity, the NVIDIA NeMo Retriever "NIMs, Milvus, and the ingestion pipeline," observability, and an Envoy gateway, and its nvidia-rag-blueprint component bundles "All NIM inference services, the ingestion pipeline, and the RAG server." The guide states no version, availability, or support level for the application. The ingestion pipeline inside it is NVIDIA's NeMo Retriever Library, which NVIDIA's own documentation says "is not supported under NVIDIA AI Enterprise (NVAIE)," and the RAG Blueprint it packages is NVIDIA reference code. What WEKA adds is real: watchers that reflect every file change in the vector index through snapshot-diff, and Portable Operating System Interface (POSIX) permissions captured at ingest and enforced at query time when identity is federated with UID and GID attributes. It's packaging around a pipeline nobody supports end to end. Applying the standard: AIDP is a closed solution, WEKA's application over WEKA's storage with NVIDIA's pipeline inside, so it has to be generally available end to end, and on public evidence it isn't. The application goes to notes. Calibration: Red Hat reads gap with its retrieval APIs in preview; Cloudera gap with retrieval in Technical Preview; VMware moderate on Private AI Services retrieval that ships in a versioned release; Everpure moderate on Data Stream 1.0.0, an appliance with published release notes. WEKA has the packaging without the support level. Gap, Absent, with the platform watch-listed.
None to borrow at the layer's function on public evidence: the retrieval WEKA packages has no documented support level, and the pipeline inside it is NVIDIA reference code outside NVIDIA AI Enterprise support. Absent: nothing generally available offered, nothing inherited.
Grade moved October 24, 2026: moderate to gap in the 1B dependency audit. Watch-list, notes only: the WEKA AI Data Platform (App Store application: filesystem watchers via snapshot-diff, NVIDIA NeMo Retriever NIMs, Milvus on a NeuralMesh S3 bucket, a RAG Gateway and RAG Bridge filtering results by POSIX permissions at query time, Keycloak, twelve RTX PRO 6000 GPUs across three nodes) and the Milvus API in front of it. Permission-aware retrieval depends on identity federation and UID and GID attributes for every user (the installation guide's own caveat), and the gateway serves plain HTTP on port 80 with no TLS out of the box. The 'data catalog' entry in the 5.1 index has no content and is unscored. Public evidence that moves the cell: a documented availability or support level for AIDP that covers its ingestion pipeline, or a WEKA-owned index or retrieval engine (the VAST reading, an argument for strong).
Move/transform data — ETL/ELT, lineage, cost-aware movement, KV cache tiering
WEKA's tiering and snapshot mechanisms between its hot tier and the object store the enterprise chose. Ceded, the NetApp FabricPool and SnapMirror reading.
WEKA's KV cache tier for inference engines the enterprise runs, generally available: WEKA's March 16, 2026 release calls it "generally available to NeuralMesh customers today" (first unveiled at GTC 2025; commercial availability announced November 18, 2025, "included as a feature for NeuralMesh deployments and on the Oracle Cloud Marketplace"). The plugins are open source; the token warehouse is NeuralMesh, WEKA's: Ceded.
AIDP's ingest embeds through NVIDIA NIM models in the NeMo Retriever stack on twelve NVIDIA GPUs, the embeddings Ceded to the model under the carve-out. The Augmented Memory Grid streams KV cache between GPU memory and NeuralMesh's NVMe 'token warehouse' over RDMA and NVIDIA Magnum IO GPUDirect Storage, integrated with NVIDIA Dynamo and NIXL through WEKA's open-source NIXL plugin; named, not scored, for lack of documentation.
WEKA moves data between tiers and sites, and moves inference engines' key-value caches out of GPU memory and back. Tiering: every NeuralMesh filesystem can attach an object store (any S3-compatible store, on premises or in the cloud) and tier data from the NVMe hot tier by time-based policies set per filesystem group (a tiering cue and a drive retention period), with per-path manual overrides ('weka fs tier fetch' pulls files back to SSD, 'weka fs tier release' forces them out 'overriding standard retention policies', and the obs_direct mount mode bypasses retention), and snapshot to object storage (Snap-to-Object) for backup, disaster recovery, and cloud bursting, with encrypted filesystem keys traveling with the snapshot. Pipelines: AIDP's ingest, a fixed pipeline from a filesystem change through parsing, chunking, and embedding into a vector index, runs NVIDIA's reference blueprint stages with no documented support level, so it's in the notes. The Augmented Memory Grid ('commercially available today on NeuralMesh and OCI', November 18, 2025; validated on OCI H100 infrastructure at '1000x more KV cache capacity' and '20x faster time to first token') keeps inference key-value caches in a persistent NVMe token warehouse and streams them back to GPUs, which is the cache tiering the layer's purpose line names; WEKA's March 16, 2026 release calls it "generally available to NeuralMesh customers today," so it's scored, though the 5.1 documentation index has no page for it. NeuralMesh 6 (second half of 2026) adds 'intelligent data mobility with replication and remote caching support'. The buyer gets its cold data on cheaper storage, its snapshots somewhere else, and KV cache capacity beyond GPU memory. The architect's concern is fixity and the documentation gap. Tiering is a policy, not a pipeline; cross-site replication is a future release; and the flagship KV cache feature has a GA announcement and open-source plugins but no page in the documentation index, so an architect sizes it from WEKA's qualification criteria, not a guide. Calibration: NetApp reads moderate on a fixed AI-ingest pipeline plus SnapMirror, FlexCache, and FabricPool tiering; Everpure moderate on fleet replication plus a fixed preparation pipeline; Dell moderate on a data orchestration engine; VAST strong on a general DataEngine with user functions; Confluent and Cloudera strong on movement platforms. WEKA is NetApp's movement shape without the AI-ingest pipeline, plus KV cache tiering (DDN's Infinia NIXL backend is the other scored KV cache tier), with the replication half pending. Moderate.
Ceded at every mover, Delegated on the override. Tiering, Snap-to-Object, and the Augmented Memory Grid are WEKA's mechanisms: Ceded, the NetApp FabricPool and SnapMirror reading; the object store behind them is the enterprise's choice and is scored on its owner's row. The runtime call is NeuralMesh tiering per the filesystem group's policy, and the documentation gives the enterprise a per-path reversal of that call (fetch, release, obs_direct): vendor decides, visible, overridable per path, Delegated under the override rule, the Snowflake and Cloudera reading; the tiering policy itself is configuration, the manual release is the override.
Authority written Delegated on the documented per-path fetch and release overrides after the ChatGPT pass; NetApp's 1C (Ceded, September 2) predates the override rule. Scored October 24, 2026 under the GA evidence table: the Augmented Memory Grid had been named, not scored (ruled September 15, 2026 with Everpure's Data Intelligence: 'available now' does not equate to GA); WEKA's March 16, 2026 release announces it "generally available to NeuralMesh customers today," the owner's own GA announcement. Open question for a later pass: whether the open NIXL and LMCache plugins are a Delegated door in front of the NeuralMesh token warehouse. Watch-listed: NeuralMesh 6 replication and remote caching (second half of 2026). Public evidence that moves the cell: NeuralMesh 6 replication would argue strong on the tiering-plus-movement reading, with the Augmented Memory Grid. Removed in the 1B dependency audit follow-ups (October 24, 2026), to notes: AIDP Continuous Ingest as a pipeline (snapshot-diff change capture, parsing, chunking, embedding, indexing on NVIDIA's blueprint stages), unscored at 1B and 2B for the same reason.
GPU scheduling, quotas, RBAC, fair-share scheduling, utilization optimization
WEKA's control plane for WEKA's storage on the enterprise's Kubernetes and clouds, and its application deployment surface. Ceded, the NetApp Console and VAST Polaris reading.
A standard Kubernetes storage interface in front of WEKA volumes; the claims and classes lift to any CSI driver. Delegated, the NetApp Trident and Dell CSI Operator reading.
WEKA's observability surfaces over WEKA's clusters. Ceded, the Pure1 reading.
The WEKA Operator schedules storage containers; Kubernetes, Slurm, or NVIDIA's software schedules the GPUs beside them. In Axon the two share servers, and the deployment guide asks the enterprise to reserve cores for both.
WEKA orchestrates its own storage on the enterprise's orchestrators. The WEKA Operator (its own release cadence, x86 only; Kubernetes 1.25 and OpenShift 4.17 minimums; EKS, OKE, GKE, and AKS supported with per-provider limits) manages WekaCluster, WekaClient, WekaContainer, and WekaPolicy custom resources with nodeSelector, roleNodeSelector, and tolerations that pin which nodes run which WEKA containers: cluster creation and expansion ('The number of containers cannot exceed available Kubernetes nodes'), shrinking with a hot spare, rolling, manual, or all-at-once upgrades, drive signing by policy, a 24-hour graceful destroy window, Composable Clusters per tenant, and, since Operator 1.7.0, an embedded CSI plugin, an automatic StorageClass, and Prometheus metrics by default. The Cloud Deployment Manager at cloud.weka.io is a wizard that generates the operator commands, Helm values, and manifests for dedicated, Axon, or Composable deployments, and the Terraform modules for AWS, Azure, GCP, and OCI create the instances, auto-scaling groups, and functions that install NeuralMesh in the enterprise's accounts. The CSI plugin (an open-source project on GitHub) provisions directory-, snapshot-, or filesystem-backed volumes with quota enforcement, topology-aware placement, expansion, snapshots, and cloning, over the WEKA client or, since 2.5.0, NFS transport; WEKA recommends the Operator-managed client over the stateless client it once documented. The WEKA App Store (a Helm chart on GitHub Pages) installs the Operator, the CSI driver, and a GUI for deploying AI applications onto the cluster. The buyer gets storage that follows its Kubernetes and its clouds declaratively. The architect's concern is that nothing here schedules compute. The Operator places storage containers, the Terraform places storage instances, and the GPU, node, job, and fair-share decisions belong to the enterprise's Kubernetes, Slurm, or NVIDIA's software; the Operator's only template is 'dynamic' ('Future templates will include capacity and performance'); SMB-W needs Operator 1.11 with 5.1.20, Data Services 1.13 with 5.1.20 for quota coloring and 'Not supported for Data Catalog'. Calibration: NetApp reads moderate on a data-infrastructure control plane plus an open CSI driver; VAST moderate on Polaris; Everpure moderate on Fusion plus Portworx; Nutanix strong on Prism, ADS, and a Kubernetes platform; CoreWeave strong on GPU orchestration. WEKA is NetApp's shape with a deeper Kubernetes operator. Moderate.
Ceded at the operator, Delegated at the pin and the CSI door. The WEKA Operator, the Cloud Deployment Manager, the Terraform modules, and the App Store are WEKA's: Ceded; the Kubernetes and clouds under them are the enterprise's and scored on their rows. The CSI plugin speaks the Kubernetes CSI standard in front of WEKA volumes and is open source: Delegated at that interface, the NetApp Trident reading. The runtime call the layer asks about is placement, and here the enterprise pins it: nodeSelector, roleNodeSelector, and tolerations on WekaCluster and WekaClient name the nodes WEKA containers run on, the Dataiku named-cluster reading under the override rule; the Operator schedules inside those pins. Vendor decides the defaults, visible (Prometheus, NeuralMesh Observe), overridable per node selection, Delegated. The Hugging Face and Everpure line (a hosted flavor is a menu, not a pin) doesn't apply to node selectors on the enterprise's own cluster.
The Operator's current compatibility table: S3 from 1.7 with 4.4, NFS and audit from 1.10 with 5.1.0, SMB-W from 1.11 with 5.1.20, Data Services from 1.13 with 5.1.20 (quota coloring; not Data Catalog), ssdproxy 1.12, AlloyFlash 1.14, cluster capacity sizing 1.14.2. The stateless client is 'supported' but 'not recommended' in favor of the Operator-managed WekaClient. Authority written Delegated on the documented node selectors and tolerations after the ChatGPT pass. Named, not scored: NeuralMesh Axon Deploy's Slurm support (the Slurm scheduler is the enterprise's). Public evidence that moves the cell: compute or GPU scheduling from WEKA, which nothing describes. Placement (October 16, 2026): NeuralMesh Observe and Local WEKA Home reads here under the function audit and the hardware-management placement rule.
Model serving, agent execution, inference APIs, distributed inference
AIDP's answers come from NVIDIA NIM inference services in the NVIDIA RAG blueprint on NVIDIA GPUs; the Augmented Memory Grid plugs into NVIDIA Dynamo and NIXL (with TensorRT-LLM and LMCache) to serve other engines' KV caches. Ceded to NVIDIA under the channel-substitution rule.
WEKA serves one thing, and on public evidence it isn't a supported product. The App Store deploys the 'nvidia-rag-blueprint' ('All NIM inference services, the ingestion pipeline, and the RAG server') beside the 'aidp' components ('the AI Data Platform's GUI, Operator, RAG Gateway, RAG Bridge, and Space Manager'); users and applications 'connect to NeuralMesh AIDP through a standard UI and API to run semantic queries' and get 'agentic RAG' answers filtered by each user's file permissions. The RAG server at the center is NVIDIA's RAG Blueprint, which NVIDIA publishes as reference code and doesn't list among NVIDIA AI Enterprise software, and the installation guide gives the application no version, availability, or support level. The Augmented Memory Grid is memory for inference engines the enterprise runs (vLLM, TensorRT-LLM, Dynamo) and is scored at 1C. There's no model catalog, no bring-your-own model, and no agent runtime. Applying the standard: AIDP is a closed solution, so its answering surface has to be generally available end to end, and the reference RAG server at its center isn't supported by its owner. The September 15, 2026 ruling counted NVIDIA's blueprint on WEKA's paper under the whose-paper rule; what fails now is the blueprint's support status, not its paper. The application goes to notes. Calibration: Everpure reads moderate on Data Stream's chat endpoint, in an appliance with published release notes; NetApp gap where its AI Data Engine retrieves without generating. WEKA is its 1B reading carried up. Gap, Absent.
None to borrow at the layer's function on public evidence: the answering surface WEKA packages has no documented support level, and its RAG server is NVIDIA reference code. Absent: nothing generally available offered, nothing inherited.
Grade moved October 24, 2026: moderate to gap, superseding the September 15, 2026 ruling that held moderate on the AIDP generative surface (the whose-paper reading), when NVIDIA's documentation showed the RAG Blueprint as reference code and the NeMo Retriever Library outside NVIDIA AI Enterprise support. Watch-list, notes only: AI Data Platform agentic RAG (the NVIDIA RAG Blueprint's NIM inference services and RAG server; WEKA's RAG Gateway, RAG Bridge, and Space Manager; prompt and completion traces in Phoenix). Scored at 1C: the Augmented Memory Grid. Named, not scored: the App Store's catalogue beyond AIDP (undocumented). Public evidence that moves the cell: a documented support level for AIDP that covers its RAG server, or a WEKA-fronted generation endpoint with model choice.
Policy-driven placement and resource coordination — the Autonomy Layer
Nothing at this layer is a scored capability.
Read against the five legs, WEKA has none as a plane over agents. Identity: Organizations, Composable Clusters, S3 STS and OIDC credentials, and AIDP's Keycloak identify tenants, users, and service principals to storage and to one application; no agent principal exists. Gateway: AIDP's RAG Gateway and RAG Bridge sit between a user or application and the platform's index, looking up the caller's identity at query time and filtering results by file permissions, and its observability records prompts and completions; that's an application-scoped retrieval gateway over one platform's data, named with the platform at 1B and 2B, where it's unscored, and nothing sits in front of an agent's model or tool calls in general. Registry and orchestration: none. Observability: WEKA Home, Prometheus metrics, and the S3 audit webhook watch storage; AIDP's observability watches the platform. What WEKA gives the reasoning plane is a governed substrate: permission-aware retrieval that enforces the filesystem's ACLs on vectors at query time, prompt and completion traces inside one application, hardware-isolated tenants, and an audit trail for storage actions. The architect's concern is the same as NetApp's: data guardrails aren't agent governance. Calibration: NetApp, VAST, Dell, and Everpure read gap; HPE moderate on an IT-operations reasoning plane; Nutanix moderate on an agent gateway. WEKA is NetApp's shape with an application-scoped gateway. Gap, authority Absent (ruled September 15, 2026): the 2C floor is the Nutanix line, a GA gateway every agent and model call passes through with governance and audit over the enterprise's agent traffic; AIDP's RAG Gateway, RAG Bridge, Keycloak identity, and Phoenix traces are one application's own governance over one platform's data (the Cloudera Agent Studio and Confluent reading), and a permission filter isn't reasoning. Both reviewer passes had voted moderate.
Nothing offered as a plane over agents. The enterprise that wants agent identity, a gateway, a registry, or orchestration brings its own; WEKA's permissions travel with the data, not with the agent. Absent.
Ruled September 15, 2026: gap stands below the Nutanix 2C floor; a thin moderate had been argued on AIDP's application-scoped identity, gateway, and observability legs. Named, not scored: AIDP's permission-aware retrieval (1B), Keycloak identity and prompt-and-completion traces inside AIDP (2B), Composable Clusters (1A). Public evidence that moves the cell: an agent principal, a gateway over agent requests generally, or a registry of the enterprise's agents; nothing in the documentation describes one.
AI-powered business capabilities — business logic, workflow automation
Nothing at this layer is a scored capability; the one documented App Store blueprint is NVIDIA's RAG blueprint, scored at 1B, 1C, and 2B.
WEKA ships no application. The WEKA App Store ('browsing and deploying AI applications' from a Helm chart, with a blueprints catalogue, a credential manager, and a compatibility check) is deployment tooling scored at 2A, and the only blueprint its documentation covers is the AI Data Platform, which WEKA's own page calls 'WEKA's AI data pipeline' and which is unscored at 1B, 1C, and 2B for want of a documented support level; the catalogue's other contents aren't documented, and WEKA's software partners page lists reference architectures, not deployable applications. The buyer gets nothing to log into at this layer. The architect's concern is nil: there's no application to be captured by. Calibration: NetApp and Everpure read gap on storage beneath others' applications; Nutanix partner on a documented ISV catalogue; VAST partner on a focused ecosystem; Dell and HPE partner on formal programs. The first draft read moderate on the Nutanix reading; the ChatGPT pass showed the documented catalogue is one data pipeline. Gap, authority Absent.
Nothing offered, nothing inherited. Absent.
Moved from moderate to gap on the GA-gate: the App Store's documentation covers its installation and one blueprint. Named, not scored: the App Store (scored at 2A as deployment tooling), the software partners page. Public evidence that moves the cell: documented deployable applications in the catalogue would argue partner on the Nutanix reading; a first-party application would argue strong.
NeuralMesh documentation at docs.weka.io (version 5.1: S3 protocol, limitations, versioning, object lock, lifecycle rules, bucket notifications, audit webhook, users and authentication with LDAP and OIDC, performance buckets; CSI plugin and topology-aware provisioning; prerequisites and compatibility; AWS, Azure, GCP, and OCI installation and Terraform modules; Cloud Deployment Manager; KMS management; release support and commitments; the WEKA App Store and the WEKA AI Data Platform installation guides; the llms.txt documentation index; version 5.0: NeuralMesh Axon overview, deployment, and maintenance; WEKA Operator deployments and day-2 operations; Composable Clusters; Organizations; licensing overview and classic license; bare-metal installation paths; SMB-W; monitoring with Prometheus and Grafana; Local WEKA Home security and compliance; stateless client best practices); weka.io product pages (NeuralMesh, NeuralMesh Axon, Augmented Memory Grid, WEKApod); WEKA press releases (WEKA Breaks the AI Memory Barrier with Augmented Memory Grid on NeuralMesh, November 18, 2025; WEKA Unveils WEKApod 3, July 21, 2026; WEKA Debuts NeuralMesh 6, July 22, 2026; WEKA and Andromeda Partner, July 30, 2026) and the OCI article; GitHub license API for weka/csi-wekafs. Peer-reviewed cell by cell through the labs claims ledger (weka-<layer>-chatgpt, ChatGPT gpt-5.5) and as a whole row by Antigravity (weka-row-agy); totals and escalated items in reviews/weka-judgment.md. /reconcile (October 5, 2026): Layer 3 calibration citations updated for the Layer 3 re-reads (Nutanix, VAST, VMware to partner; CoreWeave, NVIDIA, AMD to gap) and the MongoDB BI ruling; no grade or authority change on this row. /reconcile (October 5, 2026): settled escalations cleared from the published text and ruling attributions neutralized; open escalations stay named as pending a ruling. No grade, component, or authority change. /reconcile (October 5, 2026): calibration text citing NetApp's AI Data Engine updated after its documentation was withdrawn and NetApp's 1A moved to moderate. Text only; grades and readings unchanged. v1.4 (Layer 0 sub-layers, October 16, 2026): Layer 0 split into 0A Compute, 0B Network Fabric and 0C Storage Substrate by runtime decision (Keith, October 16, 2026); storage left 1A for 0C, and 1A, now Data Foundation & Governance, is graded on the catalog that classifies the data. 0A gap, 0B gap, 0C strong; summary edits (summary worksheet, Keith, October 16, 2026). v1.5 (October 19, 2026, /reconcile): plain-language pass: methodology rule numbers in the prose and audit notes replaced with the reasoning they stood for; no grade, chip or authority reading changed. v1.6 (October 23, 2026, Layer 1 managed-services pass): WEKA AI Data Platform gets a door. Grade unchanged. v1.7 (October 24, 2026, /reconcile): 1B moves from moderate to gap: the WEKA AI Data Platform has no documented support level and packages NVIDIA's RAG pipeline, whose ingestion library NVIDIA AI Enterprise doesn't support. Found in the 1B dependency audit under the headend standard. Grade changed: 1B moderate to gap. v1.8 (October 24, 2026, /reconcile): 2B moves from moderate to gap, and the AIDP ingest facet leaves 1C: the AI Data Platform's RAG server is NVIDIA's reference blueprint, and the application has no documented support level. Found in the 1B dependency audit follow-ups. Grade changed: 2B moderate to gap. v1.9 (October 24, 2026, /reconcile): 1B calibration updated: VAST reads moderate at 1B, InsightEngine unscored without a product guide. Found in the 1B dependency audit follow-ups. Grade unchanged. v1.10 (October 24, 2026, /reconcile): 1B calibration restored: VAST reads strong at 1B again, InsightEngine GA on VAST's own announcement. Grade unchanged. v1.11 (October 24, 2026, /reconcile): 1C scores the Augmented Memory Grid: WEKA's March 16, 2026 release announces it generally available, which passes rule 4 of the GA evidence table. Grade unchanged. v1.12 (October 24, 2026, /reconcile): GA evidence recorded on 4 of 4 Layer 1A chips under the GA evidence table version 2. Grade unchanged. v1.13 (October 24, 2026, /reconcile): GA evidence recorded on 2 of 2 Layer 1C chips under the GA evidence table version 2. Grade unchanged. v1.14 (October 24, 2026, /reconcile): GA evidence recorded on 9 of 10 Layer 0C chips under the GA evidence table version 2. Grade unchanged. v1.15 (October 24, 2026, /reconcile): GA evidence recorded on 3 of 3 Layer 2A chips under the GA evidence table version 2. Grade unchanged. v1.16 (October 24, 2026, /reconcile): GA evidence table version 2.1 applied: 1 chips recorded or narrowed. Grade unchanged. v1.17 (October 24, 2026, /reconcile): tenancy chips re-read on native multi-tenancy in the 5.1 documentation (tenants replace organizations; per-tenant QoS caps on throughput and IOPS; tenant network spaces and tenant-scoped S3). The earlier "No QoS Isolation" reading came from the 4.4 Organizations model. Grade and DAPM unchanged. v1.18 (October 27, 2026): each component now carries a durable offering id; nothing else changed.