# Xooplab — full corpus # generated 2026-05-27T06:40:36.887Z # Fields ## Fintech - url: https://xooplab.com/fields/fintech Money is software now. We work on the rails, risk models, and trust primitives that let value move as freely — and as accountably — as information does. --- ## Spatial intelligence - url: https://xooplab.com/fields/spatial-intelligence How built and natural space should be arranged is one of the oldest questions humans have asked. We formalise the priors — ancient and modern — into structured constraints a machine can reason over and test. --- ## Economic modeling - url: https://xooplab.com/fields/economic-modeling Markets and platforms publish more data than anyone audits. We build open, reproducible models of how value is created, captured, and distributed — instead of vibes and quarterly decks. --- ## Biological computation - url: https://xooplab.com/fields/biological-computation Living systems compute. We study the algorithms biology already runs and the tools that let us read, model, and design at the boundary of computation and life. --- ## Networking - url: https://xooplab.com/fields/networking Identity, provenance, and interoperability for a web whose primary reader is increasingly a machine. We build the protocols that let humans and autonomous agents share infrastructure safely. --- ## Insurance - url: https://xooplab.com/fields/insurance Risk is a coordination problem. We explore transparent, programmable insurance — pooling, pricing, and payout that anyone can inspect and that pays out without a fight. --- ## Blockchain - url: https://xooplab.com/fields/blockchain Shared state without a trusted intermediary remains genuinely useful for a narrow, important set of problems. We work on the cases where decentralisation earns its cost. --- ## Quantum communication - url: https://xooplab.com/fields/quantum-communication The next networking substrate carries information that cannot be copied or silently observed. We track what quantum links make newly possible — and what they break. --- ## Artificial intelligence - url: https://xooplab.com/fields/artificial-intelligence Artificial intelligence is the connective tissue of nearly everything we build. We work on the models, agents, and learning systems that let machines perceive, reason, and act — and, just as much, on keeping that intelligence open, legible, and accountable to the people it serves rather than sealed inside a black box. --- # Projects ## Wlcro - url: https://xooplab.com/projects/wlcro - status: active - fields: fintech, artificial-intelligence Wlcro brings consumer-grade fintech to the people and places incumbent banks treat as edge cases. We build payment rails, identity, and credit primitives that work first for the underserved — and turn out to work better for everyone. Roadmap: - Rails MVP (2025 Q4, done): Core ledger and single-currency payment primitives in production. - Cross-border settlement (2026 Q2, active): Multi-currency settlement with compliance and KYC tooling. - Credit primitives (2026 Q4, planned): Transparent, model-driven micro-credit with auditable scoring. --- ## Bio Net - url: https://xooplab.com/projects/bio-net - status: active - fields: biological-computation, networking, artificial-intelligence Bio Net treats biological data as a network problem. We build the pipelines and shared infrastructure that let researchers model living systems collaboratively, with provenance and reproducibility built in from the first byte. Roadmap: - Data fabric (2026 Q1, active): Provenance-tracked ingestion for heterogeneous biological datasets. - Model registry (2026 Q3, planned): Shared, versioned models with reproducible benchmarks. --- ## Mami - url: https://xooplab.com/projects/mami - status: active - fields: networking, artificial-intelligence Mami is media built for the edge — content that renders, adapts, and is served close to the people consuming it. We work on low-latency delivery and machine-legible media metadata so both readers, human and machine, get a first-class copy. Roadmap: - Edge delivery (2026 Q1, active): Latency-optimised media delivery at the network edge. - Structured media (2026 Q3, planned): Machine-legible captions, provenance, and rights metadata. --- ## Salus - url: https://xooplab.com/projects/salus - status: active - fields: insurance, blockchain Salus is programmable insurance: pooling, pricing, and payout that anyone can audit and that settles without a claims fight. We are proving the narrow, important cases where decentralised risk-sharing genuinely beats the incumbents. Roadmap: - Parametric pools (2026 Q1, active): On-chain parametric cover with transparent pricing. - Open underwriting (2026 Q3, planned): Community underwriting backed by auditable risk models. --- ## TURF - url: https://xooplab.com/projects/turf - status: active - fields: economic-modeling, artificial-intelligence TURF gives enterprises a defensible way to make hard decisions — structured, auditable models that turn scattered data and tacit judgment into choices a team can stand behind and revisit. Roadmap: - Decision graph (2026 Q1, active): Structured capture of options, evidence, and trade-offs. - Audit trail (2026 Q3, planned): Versioned, reviewable decision records. --- ## Auspicia - url: https://xooplab.com/projects/auspicia - status: deployed - fields: spatial-intelligence, artificial-intelligence Auspicia brings the world's oldest spatial sciences — Vastu Shastra and Feng Shui — into a form a machine can reason over. We pair certified practitioners with spatial diagnostics that turn directional and elemental priors into clear, testable guidance for the spaces people live in and build. Every room has a direction; every direction has meaning, and we make that knowledge legible to both the homeowner and the algorithm. Roadmap: - Spatial diagnostics (2025 Q3, done): Directional and elemental scan of a property delivered as a visual report. - Practitioner network (2026 Q1, active): Certified Vastu and Feng Shui consultants on-platform, translating the report into action. - Builder + realtor tools (2026 Q3, planned): Pre-construction alignment scoring so new builds start in balance. --- ## Human body as a machine - url: https://xooplab.com/projects/human-body-as-a-machine - status: active - fields: biological-computation, artificial-intelligence Human body as a machine treats physiology as an engineered system — inputs, control loops, failure modes, and maintenance schedules. We model the body's subsystems with the rigour engineers bring to machines: where the levers are that keep it running well, and what the earliest signals look like when something drifts out of spec. The aim is a shared, falsifiable model of the body that a clinician, an engineer, and an agent can all read from the same page. Roadmap: - Systems map (2026 Q2, active): Model the body's major subsystems as interacting control loops with measurable inputs and outputs. - Failure-mode catalogue (2026 Q4, planned): Catalogue the early, measurable signals that precede subsystem failure. --- # Publications ## Human Body: The Ultimate Hardware - url: https://xooplab.com/publications/human-body-the-ultimate-hardware - kind: Preprint - published: 2026-05-25 - author: Mohit Gulati (Xoop Innovation Labs) - cite_as: Gulati, M. (2026). Human Body: The Ultimate Hardware. Xooplab Working Paper XL–2026–001. xooplab.com/publications/human-body-the-ultimate-hardware. Abstract: A computational, bioelectric, and quantum architecture of the embodied self — and the Convergence Hypothesis for mind and substrate. The body understood not as a machine that resembles a network, but as a network in the fullest computational sense. Key claims: 1. Architectural — mapping a physiological control circuit (e.g., the HPA axis) onto a formal distributed-systems specification will reveal feedback-delay, redundancy, and failure-recovery mechanisms predicted by systems theory before they are found in the biology. 2. Bioelectric control — treating the bioelectric layer as a rewritable control plane will keep yielding morphological reprogramming (regeneration, tumor normalization, structural editing) unreachable by genetic means alone. 3. Scale-free inference — the Markov-blanket formalism will model the same inference dynamics at cell, tissue, organ, and organism scales with composable machinery, and one-scale interventions will have predictable effects at adjacent scales. 4. Two modes — there is a measurable line between transmitted bodily information (copyable) and information bound up in the unity of experience (if real, non-copyable); a high-fidelity mind-copy reproduces behavior but is a distinct subject from the instant of duplication. 5. Quantum substrate — if unity rests on non-classical shared state, manipulating quantum-coherence conditions in neural microtubules will modulate consciousness in dose-dependent, mechanism-specific ways no classical rate-based model predicts. 6. Co-evolution — progress toward strongly integrated machine intelligence will be gated by substrate transitions (neuromorphic, analog, biohybrid), not classical compute scaling alone. 7. Convergence — biohybrid systems will outperform pure-silicon and pure-biological systems on adaptive, low-power, self-repairing tasks, and the gap will widen as integration deepens. --- ## Encoding directional priors as model constraints - url: https://xooplab.com/publications/encoding-directional-priors-as-model-constraints - kind: Preprint - published: 2026-05-10 - board: Spatial harmony, computed (https://xooplab.com/boards/spatial-harmony-computed) - author: Mohit Gulati (Xoop Innovation Labs) - cite_as: Xooplab (2026). "Encoding directional priors as model constraints." xooplab.com/publications/encoding-directional-priors-as-model-constraints Abstract: A first pass at making spatial-harmony rules legible to a diffusion model. From the Spatial × ML board. Key claims: 1. Twelve directional/adjacency rules from Vastu and Feng Shui can be expressed as compatible JSON-LD constraints with measurable tolerance. 2. Layout outputs constrained on those rules were preferred 62%–38% in blind paired comparison (n=240). 3. Three of twelve rules account for most of the preference signal; the remaining nine are satisfied but invisible at this granularity. 4. Light-and-airflow proxies improved 9–14% per room under the constrained generator. 5. The result is a starting point, not a vindication of any specific tradition — null results on the other nine rules are also a contribution. --- ## What "login" means for an autonomous agent - url: https://xooplab.com/publications/what-login-means-for-an-autonomous-agent - kind: Note - published: 2026-04-18 - board: Identity for two readers (https://xooplab.com/boards/identity-for-two-readers) - author: Sahana Iyer (Engineer, Civic Labs) - cite_as: Xooplab (2026). "What 'login' means for an autonomous agent." xooplab.com/publications/what-login-means-for-an-autonomous-agent Abstract: Delegated, revocable agent credentials, and why session models break when the user is a process. Key claims: 1. OAuth scopes today are coarse enough that the user approves the agent's full surface area in one click. 2. Revocation needs to be immediate AND partial: kill one rogue agent without logging out the human or breaking unrelated agents. 3. Credentials must survive a clean agent-process restart without losing scope or audit lineage — implying issuer-side state, not agent-side memory. 4. Scope grammar should compose at request time ("this call, read only, $50 max") and remain legible to a non-technical user. --- ## Provenance as a first-class web primitive - url: https://xooplab.com/publications/provenance-first-class-primitive - kind: Essay - published: 2026-03-12 - board: The agent-legible web (https://xooplab.com/boards/agent-legible-web) - author: Aish Brown (Xoop Innovation Labs) - cite_as: Xooplab (2026). "Provenance as a first-class web primitive." xooplab.com/publications/provenance-first-class-primitive Abstract: If models are the second reader, citations can't be an afterthought bolted on at render time. Key claims: 1. Provenance is currently a presentation-layer feature; it should be a structural property of content. 2. Agents retrieve passages, not pages — so each atomic claim needs an attachable source, not a page-level citation. 3. A minimal primitive: stable claim IDs + source records + a verifiable hash, exposed in JSON-LD and an /llms.txt index. 4. Sites that ship provenance natively are disproportionately cited by retrieval systems, because they reduce hallucination risk for the consuming model. --- ## A reproducible measure of platform extraction - url: https://xooplab.com/publications/reproducible-measure-of-platform-extraction - kind: Method - published: 2026-02-20 - board: Platform capitalism, audited (https://xooplab.com/boards/platform-capitalism-audited) - author: Naoko Yamamoto (Postdoc, LSE) - cite_as: Xooplab (2026). "A reproducible measure of platform extraction." xooplab.com/publications/reproducible-measure-of-platform-extraction Abstract: Open method, open data, open to being wrong. A scoring function for value capture in two-sided markets. Key claims: 1. Extraction ratio = platform-side value capture / gross transaction value, sourced from filings + public datasets. 2. We publish the formula, inputs, and a cell-by-cell working notebook so anyone can fork and re-run. 3. Across three sample platforms the spread between highest and lowest extraction was ~6×, larger than expected. 4. Public filings undercount, so the score is a floor — not a verdict. --- ## Semantic diffing beats version pinning - url: https://xooplab.com/publications/semantic-diffing-beats-version-pinning - kind: Note - published: 2026-01-15 - board: Migration-resistant interoperability (https://xooplab.com/boards/migration-resistant-interoperability) - author: Diallo Okafor (Staff engineer, FinTech) - cite_as: Xooplab (2026). "Semantic diffing beats version pinning." xooplab.com/publications/semantic-diffing-beats-version-pinning Abstract: Why bidirectional adapters age better than frozen contracts. Early results from the Interop board. Key claims: 1. Version pinning postpones the cost of integration drift; it doesn't reduce it. 2. Semantic-aware diffing of API specs catches 73% of real breaking changes in our three-spec test, with 8% false-positive noise. 3. Bidirectional adapters generated from those diffs translate both ways, eating much of a typical migration. 4. Open question: how much of the residual 27% is automatable vs structurally judgement-bound. --- ## A field guide to the agent-legible web - url: https://xooplab.com/publications/field-guide-to-the-agent-legible-web - kind: Essay - published: 2025-12-08 - board: The agent-legible web (https://xooplab.com/boards/agent-legible-web) - author: Aish Brown (Xoop Innovation Labs) - cite_as: Xooplab (2025). "A field guide to the agent-legible web." xooplab.com/publications/field-guide-to-the-agent-legible-web Abstract: What changes when crawlers outnumber readers. Our founding argument, such as it is. Key claims: 1. Retrieval systems read passages, not pages; the trusted passages have structure. 2. Four compounding habits: answer-first sections, inline structured data, curated /llms.txt, claim-level provenance. 3. Agent legibility is not SEO — SEO optimises for ranking, agent legibility optimises for the evidence pipeline. 4. Sites that don't adapt fall into invisibility, where models paraphrase them without attribution. --- ## Self-hosted models on real planning tasks: a cost wall - url: https://xooplab.com/publications/self-hosted-models-cost-wall - kind: Method - published: 2025-11-22 - board: Reasoning on self-hosted models (https://xooplab.com/boards/reasoning-on-self-hosted-models) - author: Elena Volkova (Independent researcher) - cite_as: Xooplab (2025). "Self-hosted models on real planning tasks: a cost wall." xooplab.com/publications/self-hosted-models-cost-wall Abstract: Where open models stop being cheaper. Benchmarks and the spreadsheet behind them. Key claims: 1. On a 14-task financial-planning benchmark, the best open ≤14B model reaches 71% of frontier accuracy at ~6% of the cost. 2. Above ~78% accuracy target, open self-hosted models stop being the cost-effective choice on commodity GPUs. 3. Thin instruction-tuning on the benchmark's task family is dramatically cheaper than heavier prompting. 4. Inference cost dominates total cost only above ~10k requests/month; below that, dev + ops dominates and the calculus flips. --- ## Notes toward unified identity across devices - url: https://xooplab.com/publications/notes-toward-unified-identity-across-devices - kind: Note - published: 2025-10-04 - board: Identity for two readers (https://xooplab.com/boards/identity-for-two-readers) - author: Sahana Iyer (Engineer, Civic Labs) - cite_as: Xooplab (2025). "Notes toward unified identity across devices." xooplab.com/publications/notes-toward-unified-identity-across-devices Abstract: The problem statement that became the Identity board. Key claims: 1. Cross-device identity today is a UX trick over fundamentally fragmented credentials. 2. An agent acting across two of a user's devices is treated as two different agents by most identity providers. 3. A unified, revocable agent identity needs to be portable in scope, not just present on each device. --- ## Can a tradition be falsified, kindly? - url: https://xooplab.com/publications/can-a-tradition-be-falsified-kindly - kind: Essay - published: 2025-09-12 - board: Spatial harmony, computed (https://xooplab.com/boards/spatial-harmony-computed) - author: Rohan Kapoor (Independent practice) - cite_as: Xooplab (2025). "Can a tradition be falsified, kindly?" xooplab.com/publications/can-a-tradition-be-falsified-kindly Abstract: On studying belief systems without either mocking or endorsing them. Key claims: 1. Treating a tradition as a hypothesis preserves its dignity better than treating it as folklore. 2. Falsifiability is the kindest frame: it takes the tradition's claims seriously enough to test them. 3. Some parts of a tradition are testable; others are values judgements and should be named as such. ---