Hybrid Cognitive Intelligence · Pre-Release v3.4.0 · July 2026
A cognitive intelligence platform that understands language instead of predicting it. No large language models. No neural networks. No black boxes. A deterministic 18-stage pipeline that reasons, retrieves, and responds, with every decision traceable and every claim verified by automated tests. Currently in pre-release; metrics shown describe the target state at full public publication.
Release status: READY · Generated by the Evidence Engine from the live system
Condrox AI is documented for several audiences. Pick the entry point that matches what you need.
Condrox AI is a hybrid cognitive intelligence platform. It understands and responds to natural language without using large language models such as GPT-4, Claude, or Gemini.
Instead of predicting the next token in a sequence, Condrox AI understands language by looking up what each word means, retrieving relevant knowledge, reasoning about the question, and synthesizing a response through a structured 18-stage pipeline. Every step is deterministic. Every decision is traceable. Every claim is backed by a test.
Large language models are autocomplete on a massive scale. They predict what word should come next based on patterns learned from billions of texts. They are fluent, but they do not actually understand anything, and they cannot tell you why they said what they said.
Condrox AI works more like a careful librarian. It looks up what each word means, finds the relevant knowledge, reasons through the question step by step, and writes a response it can explain. It is less fluent than an LLM. It is honest, traceable, and verifiable.
The v3.4.0 Fiksfasen patch series is complete. 11 targeted patches addressed every weakness identified by the Evidence Engine in the v3.3.0 baseline. Evidence score rose from 80.84% to 93.87%, and the full-system release status moved from NOT_READY to READY. No regressions: 938 tests still pass.
The Condrox Evidence Engine is now fully operational across all seven phases. The engine generates academic-grade verification reports covering conversation quality, reasoning, agent usage, pipeline flow, and full-system integration. Three academic editions (Lab, Academic, Peer Review) are produced on every run, alongside a static Evidence Portal with score-history chart and report library. Test suite grew from 855 to 938.
Introduced a new Word Identity System with 5 modular JSON stores (approximately 185 words) and an O(1) lookup engine with softmax intent classification. The conversation suite went from 32 failures to 51 of 51 (100 percent) passing.
Delivered a full AI Operating System with 8 subsystems: EventBus, ServiceManager, BootManager, MonitorLoop, StateManager, TaskQueue, SelfRepairManager, and AIOSKernel. 51 new tests.
User Input → API → BrainController → 18-stage Pipeline → Final Output
↓
WordIdentity • Memory • Knowledge • Reasoning • Agents
↓
26 Conversational Intelligence Components • AI Operating System
↓
5 Agents: Code • Analyze • Architect • Explain • DefaultCondrox AI is a from-scratch cognitive architecture that uses symbolic reasoning, semantic word identity, BM25 knowledge retrieval, rule-based inference, and a custom AI operating system. All of it is implemented in pure Python with zero ML framework dependencies.
| Property | Value |
|---|---|
| System Name | Condrox AI |
| System Type | Hybrid Cognitive Intelligence |
| Version | 3.4.0 |
| Architecture | 18-stage cognitive pipeline + AI-OS |
| Knowledge Base | 21,917 items across 15 domains |
| Word Identities | ~185 words (expanding to 500 to 1,000) |
| Test Coverage | 938 pytest tests (all passing) |
| Conversation Suite | 51 of 51 passing (100 percent) |
| Evidence Engine | 7 phases: system, conversation, reasoning, agent and pipeline, full-system, academic, portal |
| Deployment | Docker, Kubernetes, CI/CD (GitHub Actions) |
| API | FastAPI, 17 endpoints, auth + rate limiting |
| Dashboard | React 19 + TypeScript + Vite (planned) |
| Monitoring | Prometheus + Grafana |
| Architect | Tobias Østen (sole developer) |
| License | Copyright © 2026 Tobias Østen |
Condrox AI is developed by one person. I work alone on the entire system — architecture, development, testing, operations, documentation, and continuous improvement.
All information on this site is updated regularly, but as a solo developer there is a limit to how much I can keep live at any given time. I do my absolute best to ensure that everything presented here is correct, up to date, and technically verified.
The Condrox system is built from scratch with a focus on:
I am one person who has spent 8 months of focused work building this, and I want to show that a solo developer can actually deliver a complete, stable, and verified AI system. Everything you see here is the result of hard work, iteration, and a goal of doing things properly — even without a team.
Founder & Sole Architect, Condrox AI
Contact: kontakt@pcnorge.no
Tobias Østen is the founder and sole architect behind Condrox AI. He designed and built the entire system alone, from the first line of code to the final deployment manifest. His work sits at the intersection of cognitive architecture, symbolic reasoning, semantic identity modeling, multi-agent coordination, and operating system design for AI workloads.
He builds AI systems that do not rely on probabilistic language models or black-box neural networks. Instead, he engineers deterministic cognitive pipelines: systems that reason through structured logic, rule-based inference, and semantic understanding. Every decision the system makes can be traced, explained, and verified.
For more, see About Condrox.