This document contains proprietary technology and trade secret information belonging to PhilSocial LLC and PHL Foundation Inc. Please provide your information and acknowledge the confidentiality agreement below to proceed.
GIVET is a blockchain powered all in one App: one trusted entry point for community, payments, media, and generosity. Underneath it is a five year, 120 repository engineering estate built by a distributed team across mobile, web, backend, AI, blockchain, and infrastructure. This dossier shows how it is built, what it runs on, how it scales, where it is honest about gaps, and what the technology is worth on a replacement cost basis.
A modern, conventional, cloud native stack with no exotic choices: React Native and Next.js on the front, Python FastAPI services in Docker on AWS ECS Fargate, Go for Cognito triggers and media processing, Aurora PostgreSQL plus DynamoDB, Redis and OpenSearch for data, SQS and Celery for async work, Terraform and Helm for infrastructure, and a separate AI service (GlorAI) that routes across frontier models through OpenRouter with Qdrant vector memory. The code is organized as hexagonal, per feature DDD slices with migrations, typed schemas, linting, tests, and AI assisted engineering standards checked into every core repo.
One identity, one wallet, one API surface, many services. Every pillar of GIVET (PhilSocial, GlorAI, PhilRadio, Gaming, People Wallet, Philcoin, Education, PhilMarket Place, PRO) is a service behind the same edge, the same Cognito identity, and the same data plane. That is the super app pattern: value comes from orchestration and reduced friction, not from feature count.
Sheet A-1 reads like a house. Users come through the front door on the left, pass the foyer (edge and identity), and enter the great room where the core services live. The pantry holds data. The vault holds the token. The foundation holds infrastructure. Utility lines from external providers enter through meters on the right wall. Every room is a repository the auditor can open on GitHub.
| Step | What happens | Why it scales |
|---|---|---|
| 1 | Client (React Native, Next.js) calls the API over HTTPS through CloudFront and WAF. | Static and media traffic never touches the origin; WAF drops abuse at the edge. |
| 2 | Cognito issues JWTs; Go Lambda triggers run pre and post sign up logic. | Managed identity scales to millions without a single owned auth server. |
| 3 | philsocial-gateway and API Gateway apply Redis backed rate limits and log the request. | Per user and per route limits protect the core from bursts. |
| 4 | philsocial-api (stateless FastAPI containers on ECS Fargate) serves the feature slice: feed, wallet, donation, chat, reels. | Stateless containers scale horizontally on CPU and request count. |
| 5 | Reads hit Redis first, then Aurora read replicas; writes go to the Aurora primary or DynamoDB. | Cache offloads 60 to 80 percent of database reads; Aurora supports up to 15 read replicas. |
| 6 | Heavy work (moderation, enrichment, transcoding, rewards) is published to SQS and processed by workers with retry and dead letter queues. | Async decoupling keeps API latency flat under load. |
| 7 | Value events (rewards, donations) settle on Polygon through the rewards server and audited contracts. | Polygon handles thousands of TPS at sub cent fees; batching keeps cost per reward near zero. |
Every item below was read from a package manifest, README, or infrastructure file in the philcoin-phl organization on October 5, 2026, or from the April 2026 Architecture Reference and Security Architecture Audit.
| Layer | Technology | Where | Evidence |
|---|---|---|---|
| Mobile | React Native 0.77 · Expo SDK 52 · TypeScript 5 · Tamagui · Zustand · TanStack Query · ethers 6 · MMKV · Vision Camera · FFmpeg Kit · i18next · reCAPTCHA Enterprise · jail-monkey | philsocial-app | package.json, Maestro e2e flows, Husky, Jest, CLAUDE.md standards; 12,506 commits |
| Web | Next.js 15 · React 19 · TypeScript · Tailwind · Redux Toolkit | philsocial-web, backoffice-web, referral-web | 2,089 commits on web; 32 in last 30 days |
| Core API | Python 3.11 · FastAPI 0.116 · Pydantic 2 · SQLAlchemy 2 + asyncpg · Alembic · PynamoDB · Redis 6 · web3 7 · Stripe 15 · LiveKit · Agora tokens · uvloop | philsocial-api | requirements.txt; hexagonal per feature slices; schema.graphql; pytest, ruff, mypy, pre-commit, GitHub Actions |
| AI service | FastAPI 0.128 · Tortoise ORM + asyncpg · Celery · Redis · OpenAI SDK · tiktoken · Qdrant client · Replicate · boto3 · websockets | philglorai-api | pyproject.toml and README: OpenRouter routing, vector memory, compression, function calling, credit wallet |
| AI worker | SQS consumer · CLIP · sentence-transformers 3.3 · langdetect · OpenAI omni-moderation · gpt-4.1-mini copyright check | philsocial-ai-worker | README: ECS service philsocial-ai-worker-prod, DLQ retry contract shared with the API |
| Go services | Go · chi router · Cognito Lambda triggers (pre/post signup, pre signin, custom message, social auth) · media processor · push notifications | backoffice-api, philsocial-media-processor, philsocial-*-cognito | 7 Go repositories; backoffice-api 375 commits |
| Streaming | FastAPI · PostgreSQL · Alembic · analytics module · docker-compose-prod | philradio-api | README: stations, shows, podcasts, play events, aggregates |
| Data | Aurora PostgreSQL · DynamoDB · ElastiCache Redis · OpenSearch · S3 · Qdrant (vectors) | AWS us-east-1 | Architecture Reference Apr 2026; Terraform (ECS Fargate, ALB, ElastiCache, Route 53) |
| Async | SQS with DLQ · Celery + RabbitMQ · WebSockets | api, ai-worker, glorai | Audit Apr 2026; ai-worker README |
| Blockchain | Solidity 0.8.25 · Foundry · UUPS proxies · Chainlink oracles · ECDSA KYC · Ape (Python) · Brownie | phil-contracts, rewards-server, 20+ contract repos | Audit Apr 2026: 145 unit, 31 PoC, 11 fuzz, 5 invariant tests; two audit passes Feb 2026 |
| Infra | Terraform (HCL) · Terragrunt · Helm · Kubernetes agent · ECR · ECS Fargate · EKS · Lambda · CloudFront · WAF · GuardDuty · CloudTrail · Secrets Manager · GitHub Actions | philcoin-infra, helm-templates, Terragrunt-PHL, kubernetes-agent | deploy-to-ecr.sh, Dockerfiles in every service, workflows in ai-worker and api |
| Observability | CloudWatch · Grafana · Sentry (org phl-foundation-inc) · Firebase Crashlytics and Performance | all services | gateway README; app.json; package.json |
Roughly 20 repositories are archived predecessors (philapi, blockphil, philsocial-mobile, PHILSOCIALAPP2.0, early crowdfunding contracts). They document five years of iteration; the live estate is the ~100 active repositories above.
GlorAI is not a chatbot bolted on the side. It is a separate, metered AI service with its own identity, memory, and billing, already integrated with PhilSocial PRO authentication and already able to call into the social graph. It is the orchestration brain of the super app: the layer where a user delegates a goal (give to this cause, find me a mentor, launch my station) instead of tapping through menus.
Two agents are specified and checked into the organization, ready for the dev team to give them a body inside GlorAI. The Onboarding Teaching Agent (free) teaches the GIVET philosophy and how earning works in the first sixty seconds, then invites the member to PRO. The PRO Personal Agent is one agent per PRO member, configured through four simple tabs (focus, tone, action level, skills), metered by the credit wallet, with a phased skill library (assist, grow, act with approval) and a belief core loaded above every system prompt that no customization can override. This is the moat: not a feature, a relationship.
| AI capability | Status | What it unlocks |
|---|---|---|
| Multi model chat with memory and function calling | Live in GlorAI | Personal guidance grounded in the member's own activity |
| Embeddings for posts, causes, airdrops | Live in AI worker | Semantic feed ranking, cause discovery, GIVET MATCH |
| Automated moderation and copyright detection | Live in AI worker | Faith and family safe platform at scale without a moderation army |
| Credit metered AI monetization | Live in GlorAI | PRO upsell, agent usage billing, remote pricing |
| Onboarding Teaching Agent | Spec and prompts in repo; dev integration next | Activation, generosity profile, first action, PRO conversion |
| PRO Personal Agent | Spec and prompts in repo; dev integration next | One agent per member; the retention and ARPU engine |
| GIVET MATCH (has meets needs) | Design complete; builds on embeddings plus function calling | Where generosity creates opportunity, as a product |
| Rewards integrity (identity and geography gating, incentive cohort quarantine) | Rule defined; implementation in the rewards path | Protects the token economy from farming; keeps engagement metrics honest |
The architecture is horizontally scalable at every tier by design: stateless containers, managed identity, managed databases with read replicas, caching, CDN, and queue based decoupling. The table shows what runs today and exactly what changes at each order of magnitude. Nothing in the path to 10 million users requires a rewrite; it requires configuration, capacity, and the operational items listed.
| Tier | Today | To 1M users | To 10M users |
|---|---|---|---|
| Edge | CloudFront, WAF 17 rules, Route 53 | Same; tune cache policies, add signed URLs for media | Multi region CloudFront origins; Shield Advanced |
| Identity | Cognito, 110K+ identities, Go triggers | Same; Cognito scales to tens of millions | Same; add adaptive auth and device fingerprinting |
| API | FastAPI on ECS Fargate behind ALB; gateway with Redis rate limits | Target tracking autoscaling on CPU and RPS; 10 to 40 tasks; separate read and write services | Split hot slices (feed, wallet) into their own services (the philsocial-ms-* repos already exist); EKS for scheduling density |
| Database | Aurora PostgreSQL primary; DynamoDB for high write paths | 2 to 4 Aurora read replicas; connection pooling (RDS Proxy); partition activity tables by month | Up to 15 replicas, Aurora Global Database, shard social graph by user id; DynamoDB on demand for feeds and counters |
| Cache | ElastiCache Redis | Redis cluster mode; cache feed pages, profiles, balances; 60 to 80 percent DB offload | Multi AZ Redis cluster; fan out on write for feeds |
| Async | SQS with DLQ (AI worker), Celery | Scale workers by queue depth; separate queues per job class | Introduce an event streaming backbone (Kafka or Kinesis) for verified listen, view, and reward events; TimescaleDB for time series analytics |
| Media | S3, Go media processor, CloudFront | MediaConvert for transcoding; HLS for radio relays | Edge transcoding, multi CDN, DRM where licensing requires |
| AI | GlorAI on ECS, OpenRouter, Qdrant, Celery | Qdrant cluster; response caching; model routing by cost tier; batch embeddings in the worker | Dedicated inference tier for small models (moderation, embeddings) on GPU spot; agents as isolated per user contexts |
| Blockchain | Polygon, batched rewards | Same; daily Merkle batch payouts; indexer (backend-blockchain-indexer) | Same chain; off chain ledger with periodic on chain settlement for micro rewards |
A k6 load test suite was authored for the core API targeting 145,000 concurrent users and roughly 19,300 requests per second at peak, covering sign in, feed read, post write, wallet balance, and donation paths. It has not yet been executed against a production sized environment because a dedicated staging environment is still being provisioned. We propose running the suite with the auditor present during the dev hour, against a scaled copy of the API, and publishing the latency and error curves in this dossier.
| Repository | Total commits | Last 30 days | Default branch | Notes |
|---|---|---|---|---|
| philsocial-app | 12,506 | 74 commits, 6 authors | develop | React Native, Maestro e2e, Husky, Jest, CLAUDE.md standards |
| philsocial-api | 4,859 | 93 commits, 6 authors | develop | FastAPI hexagonal; alembic, ci, docs, tests, .claude rules |
| philsocial-web | 2,089 | 32 commits, 2 authors | develop | Next.js 15 / React 19 |
| backoffice-api + web | 591 | 24 + commits, 4 authors | main | Go chi API and Next.js UI for support |
| philsocial-media-processor | 291 | 1 commit | develop | Go; stable |
| philglorai-api | 163 | 13 commits, 3 authors | dev | Event pipeline docs, migrations, integration tests |
| philradio-api | 114 | 0 (feature complete for v1; station self service next) | develop | FastAPI; docker-compose-prod |
| rewards-server | 114 | stable | main | Python / Ape; 10 open PRs |
| philsocial-ai-worker | 32 | new service | develop | GitHub Actions to ECR and ECS |
| phil-contracts | 16 | stable | main | Foundry; 192 tests |
Engineering standards are written down and enforced in the repos themselves: CLAUDE.md and .claude rules route code standards by path, pre-commit hooks block unformatted or untyped code, and every service ships with Docker, migrations, OpenAPI docs, and tests. The team practices AI assisted engineering as a discipline, which is why a distributed team of this size sustains the velocity above.
This section values the technology estate only, not the company. It answers one question: what would it cost a well funded team to rebuild what exists in the philcoin-phl organization today, to the same scope, with the same integrations, data, and audits. Three methods converge on the same range.
PHL Foundation's engineering records show 509,672 technical hours invested since 2021 across mobile, web, backend, blockchain, AI, QA, DevOps, and product management. Applying blended, fully loaded US market rates for the same mix of seniority:
| Scenario | Blended loaded rate | Hours | Replacement cost |
|---|---|---|---|
| Low (nearshore heavy mix) | $140 / hr | 509,672 | $71.4M |
| Mid (US and nearshore blend) | $165 / hr | 509,672 | $84.1M |
| High (US senior, fintech and AI premium) | $195 / hr | 509,672 | $99.4M |
| Component | Rebuild team | Duration | Low | High |
|---|---|---|---|---|
| Mobile super app (feed, reels, chat, go live, wallet, donations, onboarding, i18n, 2 platforms) | 7 engineers + QA | 30 months | $9.5M | $13.0M |
| Web app, PRO, backoffice, referral and founders sites | 4 engineers | 24 months | $3.8M | $5.5M |
| Core API, gateway, microservices, media pipeline, notifications, Cognito triggers | 6 engineers | 30 months | $8.5M | $12.0M |
| AI platform: GlorAI, worker, preference ML, agents, moderation, evaluation tooling | 4 AI engineers | 18 months | $4.0M | $6.5M |
| Blockchain: 20 contracts, rewards server, indexer, wallets (RN, Kotlin, Swift), bridge, staking and swap apps, audits | 5 engineers + auditors | 30 months | $7.5M | $11.0M |
| PhilRadio streaming platform and analytics | 3 engineers | 12 months | $1.6M | $2.4M |
| Infrastructure, SecOps, IaC, CI/CD, observability, compliance | 3 engineers | 36 months | $3.6M | $5.2M |
| Product, design, QA, PM across 9 pillars | 8 people | 36 months | $9.0M | $12.5M |
| Integration tax: 30 plus third party integrations, app store, exchange listings, KYC, payments | $4.0M | $6.0M | ||
| Five years of iteration, failed paths, and rework (20 archived repos) at 35 percent of build | $18.0M | $25.9M | ||
| Total rebuild cost | $69.5M | $100.0M |
Scope is deliberately narrow. Each requirement is tied to a revenue engine or to a scale prerequisite, has an acceptance criterion the auditor can test, and builds on a service that already exists.
| # | Requirement | Builds on | Acceptance criteria | Quarter |
|---|---|---|---|---|
| P0 | Scale readiness: staging environment, k6 suite executed, autoscaling policies, RDS Proxy, Redis cluster, SAST/DAST in CI, secrets hardening, Terraform state verified | philcoin-infra, philsocial-api | p95 API latency under 300 ms at 10,000 RPS with error rate under 0.1 percent; all April audit items closed | Q4 2026 |
| P0 | Rewards integrity: verified identity and geography gating before Philcoin payout; incentive cohorts quarantined; passive impressions earn zero | rewards-server, philsocial-api, Didit | No payout without KYC tier and allowed geography; farming cohort excluded from all reported engagement | Q4 2026 |
| P1 | Onboarding Teaching Agent live inside GlorAI: three question generosity profile, first action, PRO invitation; user.activated event | philglorai-api, philsocial-agents | 60 percent of new users complete the profile; 25 percent complete a first generosity action in session one | Q4 2026 |
| P1 | PhilRadio station self service: launch flow, owner dashboard, Claim button, Icecast/Shoutcast/HLS relay, verified listen events | philradio-api, philsocial-app | A station goes live in under 15 minutes without staff; 50 stations per month by month 6 | Q1 2027 |
| P1 | Billboard Economy: tracked GIVET share links, verified view, click, and conversion attribution, Philcoin payout waterfall (in app first, USDT last), self service Billboard dashboard | philsocial-api, rewards-server, ai-worker | End to end attribution with fraud scoring; payout within 7 days; advertiser report matches first party counts | Q1 2027 |
| P2 | PRO Personal Agent: four tab setup, phase 1 skills, credit metering, confirm before irreversible actions | philglorai-api, credit wallet | PRO conversion lift of 3 points; agent sessions per PRO member per week of 3 or more | Q1 2027 |
| P2 | GIVET MATCH v1: has meets needs matching from generosity profiles and embeddings | ai-worker embeddings, GlorAI function calling | Match acceptance rate above 20 percent; 1,000 opportunities created in the first 90 days | Q2 2027 |
| P2 | Event streaming backbone and time series analytics for verified listen, view, reward, and generosity events | SQS today; Kafka or Kinesis; TimescaleDB | Single source of truth for every metric reported to advertisers and investors | Q2 2027 |
Six screens, one journey: a person is taught to give, sees a feed where sharing earns, holds value in one wallet, delegates to an agent, listens to a community station, and gives to a verified cause. Under each screen is the service that powers it, so the auditor can trace every pixel to a repository.
The GIVET flywheel is in the architecture, not on a poster: the Generosity Engine (daily prompt and streak), the Billboard Economy (attention converted to verified value), GIVET MATCH (has meets needs), and the Impact ledger (outcomes, not vanity) are each a service, a queue, and a table.
Most technology companies build a product and then spend years searching for a community. We did it the other way around. For thirty years I have stood in front of congregations on six continents, and for five of those years I have been building, with my own money, the technology that community deserved. What you have read in these pages is not a plan. It is finished work: 120 repositories, 20 contracts on chain, an AI that already knows how to listen, and a platform that has already moved more than 90 million Philcoin into the hands of ordinary people for doing good.
I did not build this because I love technology. I built it because I believe generosity is the most powerful force for good in the world, and the networks that shape daily life were built to extract attention, not to multiply it into good. Facebook taught the world to connect. TikTok taught it to create. GIVET teaches the world to give: time, knowledge, mentorship, access, encouragement, and yes, money. Everyone has something to give. Someone else needs exactly that. The technology in this dossier exists to close that distance.
Our mission is to build the world's leading technology powered ecosystem where generosity creates connection, opportunity, and economic flourishing. Our purpose underneath it is to advance God's Kingdom. I say that plainly because I will not hide it, and I will never require anyone to share it in order to belong. The Kingdom is in the architecture: the economics reward giving instead of attention, the AI serves the person instead of the advertiser, and the token pays the people who do the good, not the people who own the platform.
Technology does not carry a vision. People do. Every line of this platform was built under a founder who paid for it when no one else would, who put the real numbers in Section 07 instead of the flattering ones, and who will still be here when the next model, the next chain, and the next trend have come and gone. What backs this technology is not a roadmap. It is thirty years of trust, a community of millions who are waiting for it, and a founder who will not stop until the practice of generosity is as natural online as scrolling is today.
When you understand the principle of the GIVET App, you will realize that you are not contributing to just any organization. You are aligning with an innovative movement that fuels generosity and teaches giving. Welcome to the way giving happens.
Enter your authorized email to access the TSNDA Records Dashboard.
All access acknowledgments on this device are captured automatically and listed here. Submissions are also sent to the site's Netlify Forms inbox (form name: tsnda), which is the complete record across every viewer.
| # | Full Name | Phone | Date and Time | Status |
|---|