A reasoning and research assistant in private development, built by Viorel Cacau. Explore what has been implemented, what has been tested, and what remains unresolved.
Status reviewed 29 October 2026 · No public app-store release announced
CPR-IntelAI is being developed and evaluated. Public and contract-service readiness has not been established. This website is a project status page, not access to a released service.
Watch how CPR-IntelAI works — a development demonstration · Visit the main project website
Backend, browser/mobile work and audit evidence were committed on platform4-development. Private browser checks cover chat, photos, voice and source-linked research; results vary by task and provider.
Checkpoint: 9da6244 · 29 October 2026
The checkpoint offline baseline recorded 1,713 passed and 2 failed. Critical lint and mobile TypeScript checks passed; the selected client suite recorded 41 passed. A separate, narrower venue/routing suite recorded 440 passed on 28 October.
These runs have different scopes. They are not overall accuracy scores, independent validation or certification.
Two puzzle-format tests remain failing. Live retrieval can be incomplete, stale or unavailable. Search excerpts do not independently verify claims. Photo interpretation can be wrong; some photo exchanges are not retained in history. Always-available public hosting is not established.
Feature implementation and limited demonstrations do not establish reliable performance for every task.
Voice input and spoken replies have been exercised in the private browser. Providers and device permissions affect availability.
Research replies can show retrieved source links. A source link or search snippet is not independent verification of a claim.
Experimental checks help surface inconsistencies. Contradictions and factual errors can still be missed.
Explores Physical, Symbolic and Social, Psychological, Meaning, and Coherence perspectives. This framework is not empirical validation.
Photo attachment, camera capture and image questions have been tested privately. Interpretations remain fallible; appearance cannot authenticate objects.
The private test setup uses authenticated access. Configured model, search and voice providers may process submitted data; all-on-device processing is not claimed.
Use voice or text to ask CPR anything. Routing and context handling are tested but can fail.
CPR can organise exploration through Five-Realm perspectives when appropriate.
For live questions, configured search adapters attempt retrieval and expose source links when available.
Checks aim to surface inconsistencies; their presence does not guarantee detection.
Review the answer, its sources, missing evidence and uncertainty before relying on it.
Development boundary: response time depends on the task and providers. No independently measured speed advantage or all-on-device processing claim is established.
Read the current status and retained limitations before considering a future pilot or service.
Private development · No public trial or app-store release announced
Not yet. The browser is being tested privately. Public and contract-service readiness require further task-specific validation and operational work.
The 29 October checkpoint recorded 1,713 passed and 2 failed in an offline baseline. Other suites have different scopes. Passing tests do not establish overall accuracy, independent validation or decision-grade certification.
No. Retrieval and source links are implemented, but sources can be missing, stale or incorrect. Search excerpts are attributed publisher material, not independent corroboration. CPR can make factual and routing errors.
No. The private setup includes a backend and configured external model, voice and search providers. Do not infer end-to-end encryption or all-local processing from the website.
No public app-store release is announced here. The current private browser has been exercised on desktop and iPhone; this is not a general compatibility guarantee.
A way to organise questions across physical, symbolic/social, psychological, meaning and coherence perspectives. It is not a scientific validation method by itself and does not prove a claim.