The direct answer
What moved? The published change log for site release 0.9.0 recorded directory v1.0.0 at 100 profiles: 26 linked to exact-model evidence records in the component catalog and 74 source-screened through four expansion batches of 19, 19, 18 and 18 additions. Directory v1.6.0 now records 111 profiles: 55 exact-model-linked and 56 source-screened. The gain of 29 linked profiles decomposes cleanly in the current data: 18 of the 74 batch-screened profiles crossed into exact-model linkage, and 4 profiles were added outside the batch structure—PJRC, ams OSRAM, TDK InvenSense, SLAMTEC—each arriving with one linked exact record.
What is a dossier? A sector intelligence dossier is a structured research layer on a profile. Its fields, verified from the directory data, are: a review date, an integration-relevance statement, official evidence routes (each a labeled official URL with an explicit scope), selection signals, procurement questions and evidence boundaries. A dossier tells an integrator where the official record lives and which questions remain open. It answers none of those questions itself.
What did not move? The screening standard. The 56 source-screened profiles establish organization identity, hardware-layer category fit, product focus and an official web route—nothing more. The directory's own boundary field states it directly: A source-screened profile establishes a public organization identity, hardware-layer fit, product focus and official web route. It does not establish exact-model specifications, availability, compatibility, performance, certification, partnership or independent validation. Exact-model evidence remains in the separate component catalog.
Exact-model linkage is not partnership, availability, compatibility or physical validation. A linked profile means the component catalog holds at least one order-code- or model-bounded evidence record for that manufacturer; the record's claims remain manufacturer-reported or explicitly unknown.
Five categories, one uneven conversion rate
Profile growth was small (+11 profiles, all in perception-and-sensing and compute-and-control). Linkage growth was where the work went. The split by hardware layer, recomputed from the v1.6.0 category summary against the v1.0.0 distribution in the 0.9.0 change log:
| Category | Profiles v1.0.0 → v1.6.0 | Exact-model linked | Source-screened |
|---|---|---|---|
| Motion & Actuation | 30 → 37 | 18 | 19 |
| Perception & Sensing | 25 → 28 | 14 | 14 |
| Compute & Control | 20 → 21 | 11 | 10 |
| Power & Energy | 15 → 15 | 7 | 8 |
| Interfaces & Integration | 10 → 10 | 5 | 5 |
Motion and actuation remains the largest layer by profile count and the least converted by share: 8 of 30 profiles link to exact records. Perception and sensing is half converted at 14 of 28. The change log did not publish per-category linkage for v1.0.0, so this table reports baseline profile counts only and does not estimate baseline linkage by layer.
The fifteen dossier manufacturers
All 15 dossiers now sit on exact-model-linked profiles; 15 of 15 link at least one exact record, and in the current data each links exactly one. The set spans all five layers: maxon, Harmonic Drive, Nabtesco, Kollmorgen, Teledyne FLIR, Zivid, AMD, Intel, Inventus Power, PEAK-System Technik, Vector Informatik, RoboSense, Vicor, Moxa, Beckhoff Automation.
| Manufacturer | Layer | Linked exact record |
|---|---|---|
| maxon | motion-actuation | cmp-maxon-ec20-flat-ie-688690 |
| Harmonic Drive | motion-actuation | cmp-harmonic-drive-csf-14-50-2uh |
| Nabtesco | motion-actuation | cmp-nabtesco-rv-20e-ratio-57 |
| Kollmorgen | motion-actuation | cmp-kollmorgen-tbm2g-07608-a-nnaa-00 |
| Teledyne FLIR | perception-sensing | cmp-teledyne-flir-bfs-pge-161s7c |
| Zivid | perception-sensing | cmp-zivid-2-plus-m60-zvd2p-m60 |
| AMD | compute-control | cmp-amd-kria-k26-commercial-sm-k26-xcl2gc |
| Intel | compute-control | cmp-intel-atom-x7425e-fj8071504822030 |
| Inventus Power | power-energy | cmp-inventus-s-2s1p-18650-57658-101 |
| PEAK-System Technik | interfaces-integration | cmp-peak-pcan-usb-fd-ipeh-004022 |
| Vector Informatik | interfaces-integration | cmp-vector-vn1630a |
| RoboSense | perception-sensing | cmp-robosense-m2 |
| Vicor | power-energy | cmp-vicor-dcm3623t50m31c2t00 |
| Moxa | interfaces-integration | cmp-moxa-eds-2008-el |
| Beckhoff Automation | compute-control | cmp-beckhoff-cx5130-0100 |
What the pairing establishes: an official research route with stated scope, a bounded set of open procurement questions, and one catalog record whose model identity or order code is pinned. What it does not establish: that the recorded model is available, in production, compatible with any robot, endorsed by the manufacturer, or performant as specified. Every one of the fifteen linked catalog records carries an explicit limitation that no RCI physical test of that unit exists—no specimen, no capture, no thermal run, no life test. The count of RCI physical tests across all fifteen dossier-linked models is zero.
Directory growth is not product verification
The version step from v1.0.0 to v1.6.0 changed how many profiles have a documentary bridge into the component catalog. It changed nothing about the evidence state of the claims on the other side of that bridge. Screening is not validation. A screened profile can belong to an excellent manufacturer or a mediocre one; the directory records only that the organization, category fit and official route were checked against public sources on a stated date.
The same discipline applies in the other direction: a manufacturer's absence from the exact-model-linked set is a statement about RCI's audit queue, not about the manufacturer. The 56 source-screened profiles are marked with exact-model audits queued, and no ordering within the directory implies a quality ranking.
Method and limits
All v1.6.0 figures on this page are recomputed at build time from the live directory release object that also generates the public JSON and CSV endpoints—profile counts, stage counts, category summary, dossier list and linked-record identifiers are read, not transcribed. Baseline v1.0.0 figures are taken from the published change log for site release 0.9.0, because no frozen v1.0.0 directory file is published; they are attributed, not recomputed. The four added profiles are identified from current data by their position outside the four expansion batches and their 2026-08-13 verification date, which matches the category deltas exactly.
- This is a documentary comparison of published directory states. It includes no hardware acquisition, measurement or teardown.
- Per-category linkage for v1.0.0 was not published, so category-level conversion rates cannot be computed for the baseline.
- Intermediate versions between v1.0.0 and v1.5.2 are summarized in the site change log and are not re-audited here.
- A dossier reflects official sources as reviewed on its stated date; official pages and routes can change after review.
Download the directory
The current release is published as machine-readable JSON and profile-level CSV, and browsable at the manufacturer directory. The v1.0.0 baseline facts cited here appear in the change log entry for site release 0.9.0.
For version-aware BibTeX and machine citation records, use Cite RCI.