Flapping Airplanes Application
Status: DRAFT - NOT FOR PUBLICATION Date: 2026-01-28 Context: Application to Flapping Airplanes (frontier data efficiency lab, $180M raised)
Why This Resonates
Their thesis: humans are 100,000x-1,000,000x more sample efficient than current models. They want “big ideas,” not incremental tweaks.
My angle: I’ve been building infrastructure for personal research that draws on millennium-old knowledge organization systems. Medieval indices, concordances, marginalia, the apparatus of scholarship - these are scaffoldings that made human learning efficient before we had compute. There’s something there.
What I Bring
- 10 years applied ML/NLP/LLM research, Q1 publications
- PhD candidate in CS (background: econ, policy, humanities)
- Systems thinker who builds infrastructure
- Comfortable with messy, ill-defined problems
- Historical perspective on knowledge organization
What I Don’t Bring
- Pure AI theory / math depth
- Publications in NeurIPS/ICML (applied focus)
- FAANG pedigree
The Pitch
Knowledge Engineer role. Document systems-of-systems (NASA-style). Study their work from the inside, in natural language. Help bridge research and infrastructure.
ATHENA System Index (ISO 999-inspired)
Personal research infrastructure. Named for goddess of wisdom, craft, and strategic warfare.
Index maintained in .index JSON files per project, queryable via kb.lookup().
Services (active)
| ID | Scope | See Also |
|---|---|---|
| eurostat-service | EU statistics API with PostgreSQL cache. 6000+ datasets. | oa-service, pt-demographics |
| trends-service | Google Trends data service with PostgreSQL caching. Evidence for forecasting. | sylva, eurostat-service, market |
| mario-service | Input-output analysis with MARIO + FIGARO. Multi-regional IO tables. | eurostat-service, pt-agriculture |
| mcbayes | Hierarchical Bayesian analysis of R&D expenditure effects on GDP growth. | stat-rethinking |
| backup-service | Restic backups to Hetzner Storage Box. Daily at 3 AM. Encrypted, deduplicated. | beszel |
| dbs-mcp | MCP server for querying multiple PostgreSQL databases via Claude. | studies, extraction |
| depa | Decentralized patronage API and market. Internal credits (OBOL). | market, ratio |
| calendar-service | CalDAV calendar with PostgreSQL visibility for Claude management. | studies-db |
Projects (active/foundation)
| ID | Scope | Status |
|---|---|---|
| drama | Dionysian experience generation using Greek dramatic principles. Ritual as container, drama as content. | active |
| mythos | Murder mystery dinner party generator. Every character is protagonist of their own arc. | foundation |
| futures | World-building system for modeling incomplete and branching futures (Apache AGE graph). | active |
| saddle | Bifurcation point detector for forecasting markets. Finds where present pivots into different futures. | active |
| synthetic-polls | Theory discovery through demographically-calibrated AI voter agents. | foundation |
| spectral-narratives | Market return decomposition with narrative generation. Find latent factors, generate economic stories. | foundation |
| beige-book | Fed Beige Book corpus download, ontology extraction, economic narrative analysis. | foundation |
| seshat | Quantitative historical dynamics via Seshat Global History Databank (Turchin/cliodynamics). | foundation |
| memetics | Cultural contagion tracking via 4chan archives and Google Trends. | foundation |
| streetview | Urban visual sampling and structured extraction from street imagery. | foundation |
| domus | AI-powered landlord onboarding and contract intelligence for rental platforms. | foundation |
| ciencia-pt | Automated Portuguese science paper analysis and reframing. | active |
| econ-paper | Automated economic research: Discretionary income analysis across life stages. | active |
| pt-agriculture | Portuguese agriculture research - farm relations, inputs, outputs, trends. | active |
Learning
| ID | Scope | Status |
|---|---|---|
| stat-rethinking | McElreath’s Bayesian statistics course - modular learning. | foundation |
| musics | 10-year piano and music theory journey, from novice to Bach. Casio CDP-S100. | active |
Infrastructure
| ID | Scope |
|---|---|
| red | Security testing project. Attack box red1 for validating infrastructure hardening. |
| hephaestus | Factory OS - digital twins, MCP control layer, robotics simulation. |
| themis | Legal structure analysis for AI/research organizations across jurisdictions. |
| dre-legal | Legal options for token economy and self-funding research. |
Archived
| ID | Reason |
|---|---|
| markitdown-db | Superseded by shelve + studies.library. Mistral OCR is better. |
| scriptorium | Dormant. Monastic text treatment using DSPy. |
Philosophy
Two branches reflecting different modes of truth-seeking:
Apollonian (order, reason, measure, proof) - mcbayes: Bayesian inference - saddle: Bifurcation detection - futures: Graph-based world modeling - stat-rethinking: Statistical foundations
Dionysian (theatre, ritual, transformation, play) - drama: Greek dramatic principles - mythos: Murder mystery generation - athena: Wikipedia exploration game
Economic Model (DEPA)
Three-tier, no quid pro quo:
- TANK - Open research, collective of curious minds, no money
- SCRIPTORIUM - Free tools, voluntary donations, data commons
- DEPA - Service futures, internal credits (OBOL), stake-based SLAs
Medieval Influences
- Concordance: Cross-reference systems for corpus navigation
- Index locorum: Place-based knowledge organization (the .index files)
- Marginalia: Layered annotation and commentary
- Florilegium: Curated excerpts and commonplace books
- Catena: Chained commentaries linking authorities (see_also fields)
- ISO 999: Modern indexing standard informing the schema
These aren’t nostalgic - they’re patterns for efficient knowledge transfer that evolved over centuries of human use.
Draft Message
I’m a 30yr applied ML/NLP researcher (10yrs, Q1 pubs) building personal research infrastructure. PhD candidate in CS, background in econ/policy.
Your sample efficiency thesis resonates. I’ve been thinking about medieval knowledge systems - indices, concordances, marginalia - as scaffolding that made human learning efficient. There’s something there.
I’d love to support your work as a Knowledge Engineer, documenting systems-of-systems (NASA-style). Or just learn from the inside.
garden.labs-lab.com
Contact
- Site: garden.labs-lab.com
- Email: [redacted]
- GPG: [attached]
This note is a private draft. Do not publish.