Every decision has a shadow. ORACLE shows you what is hiding in it.
First-order thinking is fast, easy, and focuses on the immediate, visible effects of a decision. Second-order thinking is deliberate and looks at the consequences of those consequences. In complex systems, the first-order consequence is often a false signal—what feels good in the short term may create disastrous long-term effects. ORACLE forces you to step outside the immediate moment, mapping out the cascading effects of your choices across time, domains, and hidden assumptions so you can navigate uncertainty with clarity.
[ DECISION ]
│
▼
[ CARTOGRAPHER ] ─────► Maps the decision landscape (Domains, Stakeholders)
│
▼
[ HISTORIAN ] ────────► Extracts precedents from History & ArXiv
│
▼
[ SIMULATOR ] ────────► Projects consequences across 4 time horizons
│
▼
[ DEVIL'S ADVOCATE ] ─► Audits for blind spots, biases, and systemic risks
│
▼
[ REFRAMER ] ─────────► Generates lateral alternatives and paradigm shifts
│
▼
[ SYNTHESIZER ] ──────► Compiles the final oracle verdict and D3 tree
- 6 Specialized Agents: A LangGraph orchestrated pipeline of specialized intelligences working sequentially.
- Consequence Mapping: Projections across 4 distinct time horizons (0-3 months, 3-12 months, 1-3 years, 3-10 years).
- Historical Precedents: Contextual grounding via ArXiv research and web intelligence.
- Adversarial Blind Spot Detection: A dedicated "Devil's Advocate" agent designed to stress-test your assumptions.
- Accordion UI: Clean, immersive intelligence briefing interface that shows the Synthesis Verdict first, allowing deep dives into individual agent findings.
- D3 Consequence Tree: Interactive visual representation of cascading effects.
- Clone the repository
- Install dependencies:
pip install -r requirements.txt
- Configure Environment: Create a
.envfile in the root directory and add:GROQ_API_KEY=your_groq_api_key TAVILY_API_KEY=your_tavily_api_key
- Launch ORACLE:
python server.py
- Access the Interface: Open your browser and navigate to
http://localhost:8000.
oracle-consequences-engine/
├── agents/ # Specialized LangGraph LLM nodes
├── frontend/ # HTML interface
├── graph/ # Workflow orchestration (oracle_graph.py)
├── memory/ # Memory storage
├── static/ # CSS, JS, and D3 visualizers
├── tools/ # Research and utility tools
├── config.py # Global parameters
├── server.py # FastAPI backend with SSE streaming
└── requirements.txt # Python dependencies
Second-Order Thinking
Coined by Howard Marks and championed by decision theorists, second-order thinking recognizes that interventions in complex systems always have unintended consequences. ORACLE automates the cognitive friction required to look past the immediate reward.
Systems Theory
A decision in one domain (e.g., career) inevitably perturbs adjacent domains (e.g., relationships, health). ORACLE’s Cartographer maps these interconnected nodes so the Simulator can accurately predict cascading ripple effects.
Red Teaming
Humans are wired for confirmation bias—we seek evidence that supports our desired outcome. The Devil's Advocate agent serves as an automated "Red Team," structurally incentivized to destroy your assumptions and identify existential blind spots before they occur in reality.
-
"Drop out of college to start an AI startup."
- First order: Freedom, excitement, potential wealth.
- Second order (ORACLE): High probability of burnout, loss of credential safety net, strain on social networks outside the tech bubble.
-
"Move my engineering team to a fully remote, asynchronous workflow."
- First order: Reduced overhead, happy employees, broader hiring pool.
- Second order (ORACLE): Degradation of tacit knowledge transfer, increased isolation, communication bottlenecks hiding behind text, potential divergence of company culture.
-
"Introduce a strict performance ranking system in my company."
- First order: Clear metrics, weed out low performers.
- Second order (ORACLE): Hyper-competition, destruction of team cohesion, gaming the metrics, exodus of top talent who dislike toxic environments.
