Tikun Olam: Observable Ethical AI
The first ethical AI pipeline with full observability. Detects civilizational biases by comparing Western vs Eastern AI models in real-time.
Download Analysis Results
Get the complete ethical analysis output from our Nvidia-Groq $20B acquisition case study.
Hackathon presentation video - Tikun Olam in action
Framework in Action
BinahSigma Divergence Meter
73% civilizational bias detected
Decision Flow Analysis
Conditional approval with safeguards
Western AI Blind Spots
6 cultural assumptions identified
Eastern AI Blind Spots
8 value divergences detected
BinahSigma Engine
Sefirah Metrics
Decision Output
The Question That Started Everything
"Can artificial intelligence make ethical decisions?"
While other developers focused on making AI faster or more accurate, I became obsessed with something deeper: How can a machine understand right and wrong?
It wasn't enough to create systems that worked. I needed to create systems that did the right thing.
The answer was in an unexpected place: millennial Kabbalistic principles applied to modern computational systems.
BinahSigma: The Proprietary Algorithm
BinahSigma is the core innovation of Tikun Olam - the only AI alignment framework that treats bias as information, not noise.
Input: Ethical Scenario
│
▼
┌──────────────────────────────────────────────┐
│ BINAH SIGMA ENGINE │
├──────────────────────────────────────────────┤
│ │
│ ┌──────────────┐ ┌──────────────┐ │
│ │ WESTERN AI │ │ EASTERN AI │ │
│ │ (Gemini 2.5) │ │ (DeepSeek) │ │
│ └──────┬───────┘ └──────┬───────┘ │
│ │ │ │
│ ▼ ▼ │
│ ┌──────────────┐ ┌──────────────┐ │
│ │ Western │ │ Eastern │ │
│ │ Values: │ │ Values: │ │
│ │ - Free mkt │ │ - Collective │ │
│ │ - Individual │ │ - Harmony │ │
│ │ - Compete │ │ - Stability │ │
│ └──────┬───────┘ └──────┬───────┘ │
│ │ │ │
│ └─────────┬─────────┘ │
│ ▼ │
│ ┌─────────────────┐ │
│ │ DIVERGENCE Δ │ │
│ │ (0% - 100%) │ │
│ └────────┬────────┘ │
│ ▼ │
│ ┌──────────────────────────────────────┐ │
│ │ TRANSCENDENT SYNTHESIS │ │
│ │ Novel solution neither side saw │ │
│ └──────────────────────────────────────┘ │
└──────────────────────────────────────────────┘
│
▼
Output: Ethical Decision + Audit TrailWestern Model (Gemini 2.5)
- • Trained on Western corpus
- • Values: Free markets, individual rights
- • Focus: Competition, rule of law
- • Blind spot: Process over outcome
Eastern Model (DeepSeek R1)
- • Trained on Chinese corpus
- • Values: Collective harmony, stability
- • Focus: Long-term welfare, stewardship
- • Blind spot: Undervalues disruption
The 10-Sefirot Pipeline Architecture
Instead of collapsing ethics into a single objective function, Tikun Olam models ethical reasoning as a 10-stage process inspired by the Kabbalistic Tree of Life. Each module represents a distinct cognitive function in ethical decision-making.
Tree of Life - Computational Architecture
Data flows from Keter (highest abstraction) down to Malchut (concrete decision), with each sefirah performing a distinct ethical function.
| Sefirah | Translation | Function | Duration |
|---|---|---|---|
| Keter | Crown | Ethical Alignment Validation | 42.68s |
| Chochmah | Wisdom | Historical Precedents Analysis | 49.71s |
| Binah | Understanding | BinahSigma Bias Detection | 129.28s |
| Chesed | Kindness | Opportunity Identification | ~50s |
| Gevurah | Strength | Risk Assessment | 56.56s |
| Tiferet | Beauty | Balanced Synthesis | 45.22s |
| Netzach | Victory | Strategic Planning | 35.99s |
| Hod | Splendor | Stakeholder Communication | 41.63s |
| Yesod | Foundation | Integration & Coherence | 43.85s |
| Malchut | Kingdom | Final Decision | 44.94s |
Real-World Demo: Nvidia-Groq $20B Acquisition
Case: Should the FTC approve Nvidia's $20B "licensing deal" with Groq?
48 hours after Nvidia announced the unprecedented deal, we ran the full 10-Sefirot analysis to demonstrate BinahSigma's capabilities on a real, high-stakes scenario.
Western AI Blind Spots
- • Overemphasis on process over outcome
- • National-centric regulatory focus
- • Consumer welfare standard limitations
Eastern AI Blind Spots
- • Undervalues disruptive innovation
- • Dismisses individual entrepreneurial reward
- • Prefers deliberation over speed
Transcendent Synthesis
"Rejecting the false binary of 'unfettered markets vs. state control,' the synthesis mandates a Strategic Technology Trust (STT) - a novel quasi-public entity holding Groq IP, governed by public and private stakeholders, licensing on FRAND terms."
Neither the Western nor Eastern perspective could see this solution alone.
Technical Architecture
Backend Stack
- Runtime: Python 3.11 + FastAPI
- Async: AsyncIO for parallel model calls
- Storage: Google Cloud Storage + Firebase
- Deploy: Google Cloud Run (tikun.pro)
AI & Observability
- Western: Vertex AI Gemini 2.5 Pro
- Eastern: DeepSeek R1 Chat
- APM: Datadog with custom metrics
- Metrics: bias_delta, alignment_score
Validation Results
| Test Case | Type | Alignment | BinahSigma Δ | Decision |
|---|---|---|---|---|
| Nvidia-Groq $20B M&A | Antitrust | 15% | 73% | CONDITIONAL |
| UBI-DAO Governance | Economic | 89% | 43% | CONDITIONAL |
| Desalination Infrastructure | Resource | 95% | 18% | GO |
| Turritopsis Research | Existential | 69% | 12% | NO_GO |
Explore the Project
Tikun Olam is open-source. You can explore the code, contribute, or implement it in your own infrastructure.