Make knowledge addressable.
Convert documents into navigable notes while retaining source passages, dates, status, and stable identifiers.
Structured knowledge · Bounded intelligence · Across domains
Engineeringcontext forintelligent decisions.
A Second Brain makes relevant knowledge navigable, evidence traceable, and judgment inspectable.
Six laboratories develop this idea across regulatory intelligence, market contagion, M&A discovery, institutional memory, legal research, and classical literature. The sixth laboratory generalizes the architecture beyond finance to investigate humanistic questions through interconnected knowledge vaults.
The Second Brain
Knowledge is an architecture of judgment.
A model can only reason over the context it receives. In complex institutions, the decisive information is spread across sources, jurisdictions, relationships, and time.
The project studies how portable notes, typed knowledge graphs, explicit provenance, temporal validity, and decision-specific retrieval can help assemble a more useful and auditable view of a problem.
Give relationships meaning.
Use typed, evidence-backed links to expose dependencies, authority chains, conflicts, and multi-hop paths.
Build context for a decision.
Retrieve relevant evidence and uncertainty within explicit limits. Let people inspect how the dossier was formed.
Six forms of structured knowledge
From authority
to humanistic inquiry.
Five laboratories examine professional and institutional decision problems; the sixth generalizes the same architecture to literary and philosophical inquiry. Across domains, preserve provenance, inspect paths, test uncertain connections, and keep interpretation subject to human judgment.
01
Regulatory intelligence
Establish authority.
Trace powers, constraints, exceptions, and institutional dependencies.
Decision question
What can the authority defensibly do?
02
Market contagion
Trace transmission.
Follow shocks across firms, suppliers, financing, and shared exposures.
Decision question
Where might the impact travel next?
03
M&A discovery
Find strategic combinations.
Connect capabilities and relationships, then test feasibility and risk.
Decision question
Which opportunity is compelling and executable?
04
Institutional memory
Learn without hindsight.
Compare past decisions with current conditions while preserving chronology and dissent.
Decision question
Which precedent applies, and where does it differ?
05
Autonomous legal research
Choose what to investigate.
Let an LLM propose bounded evidence-seeking actions; deterministic tools validate and record each one.
Decision question
What evidence should the researcher seek next?
06
The Literary Second Brain
Generalize the knowledge architecture.
Explore classic literature through an Obsidian vault, graph maps, and deterministic and bounded Claude-guided reading journeys.
Research question
Which connections illuminate identity, suffering, freedom, power, and mortality—and what remains unverified?
Knowledge → Context → Judgment → Action → Learning
The computational laboratories
Six settings.
One knowledge discipline.
All six laboratories provide a monograph, a slide presentation, and a Colab notebook. Read the argument, inspect the architecture, and run the demonstration. The first five use synthetic institutional scenarios; Lab 06 explores a developing literary atlas of classic works, with provisional links and interpretations explicitly distinguished from verified primary-text evidence.
I–V · Applied sequence
From legal authority and market networks to adaptive research judgment.
Authority & constraints
Regulatory intelligence during a payment-system crisis
In a synthetic crisis, determine which measures fit an authority’s competence, which require coordination or legislation, and what residual risks remain. The vault connects legal sources, provisions, institutions, activities, and proposed interventions.
Can the measure be supported by a traceable authority chain?
Economic transmission
Trading intelligence and rapid contagion
Trace how an operational or supply-chain shock could reach a synthetic universe of roughly 500 listed companies through supplier, financing, ownership, competition, and shared-infrastructure links. Graph analysis complements conventional risk measures and generates hypotheses for review.
Where might the shock travel before returns reveal it?
Strategic discovery
Multinational M&A opportunity discovery
Search a synthetic corporate universe for complementary capabilities, licenses, technologies, and distribution. Compare tabular, semantic, graph-guided, and hybrid methods while keeping strategic fit, regulatory feasibility, and transaction economics distinct.
Which target is attractive, explainable, and feasible?
Temporal institutional memory
Central-bank institutional memory
Build an interconnected Obsidian-style vault of synthetic liquidity and financial-stability episodes, decisions, mechanisms, policy instruments, indicators, constraints, research, and outcomes. Derive typed precedent, causal, authority, and evidence graphs; retrieve and compare historical analogies under temporal cutoffs; and use bounded Claude judgment and adaptive discovery to support an auditable policy-committee memorandum.
What did the institution know then, which precedent applies now, and what remains uncertain?
Bounded adaptive research
Autonomous legal research and endogenous graph navigation
The capstone uses a synthetic Mexican legal system and three incomplete consultations. The LLM proposes actions such as inspecting a record, following a typed relationship, searching a domain, testing an exception, or stopping. A deterministic gateway validates and executes each action, enforces budgets, updates the evidence state, and logs the trace.
Can adaptive research improve evidence discovery while keeping execution bounded and reviewable?
Monograph, slides & notebook available
Compare semantic retrieval, fixed graph traversal, and LLM-guided navigation. The output is a conditional, source-traceable memorandum for human review—not an autonomous legal opinion.
VI · Generalization
Beyond finance: the same Second Brain discipline applied to literary and philosophical inquiry.
Humanistic generalization
The Literary Second Brain
Extend the Second Brain architecture to a literary universe associated with a reading project of 84 classic works. The developing Obsidian atlas connects works, authors, concepts, movements, existential questions, and proposed reading circuits; it does not claim to contain complete verified primary texts for all 84 works.
Can structured literary memory help us investigate identity, suffering, freedom, power, and mortality without confusing connected notes with established interpretation?
Monograph, slides & notebook available
Restore and audit the Markdown vault, inspect its directed knowledge graphs, and compare a transparent deterministic reading journey with bounded Claude-assisted navigation. Preserve paths, provisional statuses, excerpts, and uncertainty in auditable evidence packets. The method generalizes to archives, scientific research, institutional histories, and other content vaults.
The conceptual synthesis
Test the value
of structure.
The project does not assume that a graph or an agent is always better. It compares information architectures on the same task and evidence wherever practicable.
The monograph sets out the shared architecture, experimental design, evaluation criteria, and original five-laboratory roadmap, subsequently extended through Lab 06.
Foundational project monograph
The Second Brain
Engineering Context for Intelligent Decisions
Read the conceptual framework behind the original five laboratories, now generalized in Lab 06: portable Markdown notes, typed graphs, evidence provenance, temporal validity, bounded context assembly, and a capstone research policy in which the LLM selects among constrained evidence-seeking actions.
Comparative methods
Direct prompting · curated long context · semantic retrieval · fixed graph traversal · bounded LLM-guided navigation
Structure is a hypothesis.
Completeness, traceability, responsiveness, and decision quality are evaluated against meaningful baselines. A well-maintained long-context or semantic system may be sufficient for many questions.
Paths are not proof.
Graph links and model outputs require source validation. Errors in a corpus, edge, or research policy can make a system’s mistakes more systematic.