From v1 to v2.
Not a patch. A rebuild. What the engine became between version one and version two — in capability, posture, and reach.
One generation jump.
v1 → v2
Single tool runner → Engine + Lab + Agent Mode
One discipline at a time → Three pillars (Human Intelligence + AI), one engine
Capability upgrades.
- ▸ 01The Engine, unlockedv1 · Ran one specialist at a time.v2 → A single engine spins up specialist tools and agents on demand, on any subject.
v2 is no longer a fixed assistant. It is an engine: you describe the brief, it generates the working tool, then runs it.
- ▸ 02Agent Modev1 · No chaining. One prompt, one answer.v2 → Chained, parallel agents that plan, hand off, and converge in seconds.
Multi-stage reasoning shipped as a first-class capability — the engine deploys a small team of specialists against your brief and returns a single, reconciled output.
- ▸ 03The Labv1 · A chat box.v2 → A working surface — Build, Chat, Scan, Speak, Agent — five capabilities, one room.
v2 collapses every mode of working with the engine into one cohesive console, instead of scattered prompts.
- ▸ 04Speak — audio inv1 · Keyboard only.v2 → Live dictation or uploaded audio — transcribed and audited in one pass.
Talk to the engine the way you'd brief a colleague. The full unlocked NeuroCortex listens, transcribes, and reasons over what you said.
- ▸ 05Scan — visionv1 · Text only.v2 → Audits any image — brand, document, scene, chart — through a neuroscience lens.
v2 sees. Drop in a packshot, an X-ray, a slide, a courtroom photo — the engine returns what the human mind would notice, and what it would miss.
- ▸ 06Three pillars, one enginev1 · Generalist prose.v2 → Native reasoning across Human Intelligence (NeuroMarketing — digital + traditional marketing; Neuro-Forensics — traditional + cyber forensics; Neuro-Analytics — all brain & behavioural data → decision) and AI (Audit, Governance).
Each module carries its own posture, vocabulary and audit checks — composed inside one engine instead of bolted on as separate apps.
- ▸ 07Audited trust posturev1 · No published posture.v2 → Ships under ISO/IEC 42001 · 27001 · 27701 — Lead Auditor: Dr. Nitnem Singh Sodhi.
v2 is the first version of the engine you can deploy with a paper trail. AI management, information security and privacy — audited end-to-end.
Where we started.
120B-parameter omni-model.
505 × 30B specialist variants — one per Lab tool.
Text, image, audio.
Psychology, neuroscience, psychiatry, neuropsychology, psychotherapy, neurofeedback.
What v2 actually is.
Proprietary 200B-parameter sparse Mixture-of-Experts (MoE) omni-model.
~12B parameters active per token — real-time reasoning at MoE efficiency.
Text, vision, voice, structured data.
February 2026 — current to today.
Tool-use, multi-step planning, autonomous workflow execution.
Structured, evidence-led, transparent.
Multi-layered: toxicity, self-harm, jailbreak, PII, hallucination.
PII detection + redaction. Zero data retention.
Polished, decision-ready narrative.
Served via the Bharat Neurotech inference stack.
NeuroCortex is Bharat NeuroTech's proprietary base model. V1 was customised by Dr. Nitnem Singh Sodhi on the neuropsychological, neuroscientific, psychiatric and behavioural frameworks he uses with enterprise clients. V2's protocols, prompts and reasoning chains are authored by him directly — the world's first engine built to decode intelligence, human, artificial and business.
v3, in the workshop.
v2 is live and worked daily inside Bharat NeuroTech. v3 is being architected by Dr. Nitnem Singh Sodhi.
