# Fathym

> The AI platform that keeps the intelligence you and your team create yours, and compounds it from your first workbench to your whole company.

## Pages

- [fAI — Your Personal AI Workbench](https://www.fathym.com/fai): fAI is a free, portable AI workbench that remembers your context across every session and tool, so your AI actually knows you and gets sharper as you work.
- [How fAI Works — Context That Compounds](https://www.fathym.com/fai/how-it-works): See how fAI captures your AI context, carries it across sessions and tools, and compounds it into knowledge you own. Set up in seconds, no lock-in.
- [OpenX — Build the Tools Your Team's AI Runs On](https://www.fathym.com/openx): Save a query over your live data, or point OpenX at a repo. Either becomes a tool every AI, app and teammate can call. When it matters, AI proposes; you decide.
- [OpenX Pricing — Standard, Pro & Sovereign](https://www.fathym.com/openx/pricing): Compare OpenX plans: start in our shared cloud, scale to a managed private runtime, or bring your own. Standard, Pro, and Sovereign. Self-serve, USD.
- [Holonic — An Open Framework for Composable AI](https://www.fathym.com/holonic): Holonic is an open framework for composable AI systems. Build capabilities, agents, and workflows from holons that compose at every scale.
- [Workbenches — Build Tools Once, Reach Them Every Way](https://www.fathym.com/workbenches): A workbench is a toolkit for one job, kept in one repo. Declare each tool once, reach it over MCP, an API and the web. Run it with fai run, or host it in OpenX.
- [Workshop — Where Teams Build AI Together](https://www.fathym.com/workshop): The Workshop is where teams compose workbenches into shared, governed AI, connecting data, tools, and people around one compounding context.
- [The Loop — The Compounding Way of Working](https://www.fathym.com/the-loop): The Loop is Fathym's compounding way of working with AI: intake, thoughts, workstreams, creation. Every pass makes the next one faster and sharper.
- [OpenIndustrial — Governed AI for Industry](https://www.fathym.com/open-industrial): OpenIndustrial connects manufacturing and lab data to any AI. Every query auditable, every answer attributed. Governed AI for regulated industry, no IT project.
- [IoT — AI for Your Operational Data](https://www.fathym.com/iot): Bring any AI to your operational data. Connect existing sensors and systems, save the queries that matter, and wire them to your dashboards, apps and AI. No rip-and-replace.
- [Mosaic — Ambient AI That Surfaces What Needs You](https://www.fathym.com/mosaic): Mosaic is ambient intelligence for your work. It reads the flow and surfaces what needs you, before you go looking. Coming soon.
- [How Mosaic Works — Ambient Intelligence](https://www.fathym.com/mosaic/how-it-works): See how Mosaic reads the signals across your work and surfaces what needs your attention, quietly and in context, before it becomes urgent.
- [Mosaic Marketplace — Surfaces & Skills](https://www.fathym.com/mosaic/marketplace): Coming soon: the Mosaic marketplace, with ready-made surfaces and skills that plug ambient intelligence into the tools and workflows you already use.
- [Investors](https://www.fathym.com/investors): Fathym is building the platform for intelligence sovereignty: own the AI intelligence you create, at every scale. Investor overview and materials.
- [The Fathym Platform — Own AI at Every Scale](https://www.fathym.com/platform): One platform, one shared context, from a personal workbench to your whole company. Each tier composes into the next, and the intelligence stays yours.
- [Contact](https://www.fathym.com/contact): Get in touch with Fathym about fAI, OpenX, a Sovereign deployment, or just to connect. We reply within one business day.
- [Privacy Policy](https://www.fathym.com/privacy): How Fathym collects, uses, and protects your data across fathym.com and our products.
- [Terms of Service](https://www.fathym.com/terms): The terms governing your use of Fathym's website and products.

## Documentation

- [With fai](https://www.fathym.com/docs/the-loop/running-it/with-fai-mcp): Run `fai` and the loop's skills are served to your AI tool - no new install.
- [Running It](https://www.fathym.com/docs/the-loop/running-it): Run the loop today: run `fai`, and use the skills from your AI tool.
- [From Your AI Tool](https://www.fathym.com/docs/the-loop/running-it/per-tool): The skills show up as fai:* prompts in your AI tool.
- [Compounding](https://www.fathym.com/docs/the-loop/concepts/compounding): Each pass starts where the last ended - the loop is a cycle with feedback, not a one-way line.
- [Concepts](https://www.fathym.com/docs/the-loop/concepts): Why the loop compounds, and what makes the compounding honest.
- [Honest Compounding](https://www.fathym.com/docs/the-loop/concepts/honest-compounding): You own and steward the loop; nothing is fabricated - only the work you actually did, carried forward.
- [The Loop](https://www.fathym.com/docs/the-loop): Run the Loop with the fai skills today - Intake, Thoughts, Workstreams, Creation, the compounding way of working.
- [Creation — execute + finalize](https://www.fathym.com/docs/the-loop/stages/creation): Do the creation - build from the spec, steward it, ship it; what emerges re-enters the loop.
- [Stages](https://www.fathym.com/docs/the-loop/stages): The four stages of the loop, and the skill that runs each.
- [Intake — intake-review](https://www.fathym.com/docs/the-loop/stages/intake): Bring signals in - transcripts, chats, files, GitHub issues - and route what matters into the workstream.
- [Workstreams — create + iterate](https://www.fathym.com/docs/the-loop/stages/workstreams): Shape understanding into an executable spec - tracks, phases, deliverables.
- [Thoughts — thought](https://www.fathym.com/docs/the-loop/stages/thoughts): Turn raw signal into structured understanding that compounds - research, synthesis, chapters.
- [Glossary](https://www.fathym.com/docs/the-loop/glossary): Plain definitions for the Loop and the skills that run it.
- [The Loop](https://www.fathym.com/docs/the-loop/overview): The four stages as one continuous, compounding cycle - and the skill that runs each.
- [Enterprise team journey](https://www.fathym.com/docs/openx/journeys/enterprise): Many teams, shared sources of truth, access rights - and bring your own agent or a local model, all running in your tenant. Scale without losing ownership.
- [Journeys](https://www.fathym.com/docs/openx/journeys): See OpenX through your team's eyes - a marketing team, an engineering team, an enterprise. Each walks the same arc: connect, crystallize, build, review, compound.
- [Engineering team journey](https://www.fathym.com/docs/openx/journeys/engineering): An engineering team wires up its delivery data, crystallizes the metrics that matter, and lets any agent build the next source of truth - reviewed before it ships.
- [Marketing team journey](https://www.fathym.com/docs/openx/journeys/marketing): A marketing team connects its signup and campaign data, crystallizes one shared source of truth, and keeps every teammate's AI on the same numbers.
- [What is OpenX?](https://www.fathym.com/docs/openx/what-is-openx): OpenX is the workshop where your team connects its information to AI, crystallizes repeatable work into shared sources of truth, and builds real things on it - owning every connection.
- [Connect Data](https://www.fathym.com/docs/openx/getting-started/connect-data): Bring a connection and a surface onto the canvas and wire them - your data's entry point and the place it becomes usable.
- [Get a Workspace](https://www.fathym.com/docs/openx/getting-started/get-a-workspace): Sign in, create your workshop, and build for free - choose a plan when you are ready to deploy and go live.
- [Build a Warm Query](https://www.fathym.com/docs/openx/getting-started/build-a-warm-query): Inside the surface, crystallize a warm query with Azi - a shared source of truth that becomes an API and an MCP tool.
- [Getting Started](https://www.fathym.com/docs/openx/getting-started): Connect data, crystallize a source of truth, then read and build it from any AI - in five steps, under ten minutes.
- [Read from Any AI](https://www.fathym.com/docs/openx/getting-started/read-from-any-ai): Point any MCP-capable AI at your workshop and read the same source of truth - Claude, ChatGPT, Cursor, or a local model. Bring any AI.
- [Get Data Flowing](https://www.fathym.com/docs/openx/getting-started/get-data-flowing): Ask Azi to propose a simulator, wire it up, accept the proposal, and deploy - your first taste of working with AI through proposals.
- [Build the Next](https://www.fathym.com/docs/openx/getting-started/build-the-next): Describe what you want; any AI proposes the whole shape - a new real feed, the wiring, and a combined query. You accept; the workshop compounds.
- [Warm Query](https://www.fathym.com/docs/openx/concepts/warm-query): A named, saved query that becomes an API endpoint and an MCP tool at once - your shared source of truth.
- [Proposals](https://www.fathym.com/docs/openx/concepts/proposals): How change happens in OpenX - anyone (or any AI) proposes, the right human accepts in the portal.
- [Workshop](https://www.fathym.com/docs/openx/concepts/workshop): Your team's OpenX space - the canvas where connections, surfaces, and warm queries come together.
- [Concepts](https://www.fathym.com/docs/openx/concepts): The mental models behind OpenX - workshop, surface, warm query, and the proposal model - in plain terms.
- [Surface](https://www.fathym.com/docs/openx/concepts/surface): A scoped boundary inside your workshop where data becomes queryable and your team's AI connects.
- [Changelog](https://www.fathym.com/docs/openx/changelog): What changed in the OpenX docs.
- [Meet Azi](https://www.fathym.com/docs/openx/meet-azi): Azi is your MCP-native teammate - it understands your workshop, proposes changes from that context, and never ships without you.
- [You stay in the loop](https://www.fathym.com/docs/openx/meet-azi/you-stay-in-the-loop): AI proposes; your team decides. The propose-review-accept contract that lets you build with any AI and still own what ships.
- [OpenX docs](https://www.fathym.com/docs/openx): Documentation for OpenX - the workshop where your team connects its information to AI, crystallizes the work that repeats into shared sources of truth, and builds on it. Owned by you.
- [Secrets](https://www.fathym.com/docs/openx/guides/security/secrets): Manage API keys and connection credentials - generate tokens with the right expiry, rotate without downtime, and revoke instantly.
- [Security](https://www.fathym.com/docs/openx/guides/security): How OpenX protects your data and how you manage tokens and secrets - all in your own tenant.
- [Ship with Confidence](https://www.fathym.com/docs/openx/guides/ship-with-confidence): Every decision in your workshop is remembered - who, what, when - so your team's judgment compounds and the history is always yours.
- [Host Your Own Workbench](https://www.fathym.com/docs/openx/guides/hosting-workbench): Build a workbench, host it as an MCP server, and share it - one command and your whole team has it, usable from any AI.
- [Working with Azi](https://www.fathym.com/docs/openx/guides/working-with-azi): Describe the answer you want; Azi proposes the query. You preview, refine, and approve - nothing is saved without you.
- [Surfaces](https://www.fathym.com/docs/openx/guides/surfaces): Surfaces are the scoped workspaces where your data becomes queryable and your team's AI connects.
- [Query Patterns](https://www.fathym.com/docs/openx/guides/query-patterns): Reusable shapes for warm queries - aggregation, threshold, comparison, and trend - with the plain-English ask for each.
- [Team Review & Access Rights](https://www.fathym.com/docs/openx/guides/team-review): When a change matters, AI proposes and the right teammate accepts - in the portal, per access rights, with the reason and approval recorded.
- [Guides](https://www.fathym.com/docs/openx/guides): Deep-dive guides for every built-in workbench in OpenX - drive each one your way: the canvas, any AI, or Azi.
- [Warm Queries](https://www.fathym.com/docs/openx/guides/warm-queries): A warm query is a saved query that automatically becomes an API endpoint and an MCP tool - the shared source of truth every teammate's AI reads from and can build on.
- [KQL Basics](https://www.fathym.com/docs/openx/guides/kql-basics): Enough KQL to read and tweak the queries Azi proposes - the language your warm queries are written in.
- [Activity Log](https://www.fathym.com/docs/openx/guides/change-history/activity-log): An immutable record of every change to your workshop - browse, filter, verify its integrity, and export.
- [Change History & Corrections](https://www.fathym.com/docs/openx/guides/change-history): OpenX records every change to your workshop and every correction to your data - an immutable history you can see, trust, verify, and export.
- [Corrections](https://www.fathym.com/docs/openx/guides/change-history/corrections): Fix recorded data with a documented reason - the original is never overwritten, and you can require a second approver.
- [Navigation](https://www.fathym.com/docs/openx/guides/workflow/navigation): Find your way around the workshop - the hierarchy, the main areas, the Inspector, keyboard shortcuts, search, and breadcrumbs.
- [Workflow](https://www.fathym.com/docs/openx/guides/workflow): Find your way around the canvas, and ship changes safely with the two-stage save-and-deploy flow.
- [Save vs Commit](https://www.fathym.com/docs/openx/guides/workflow/save-vs-commit): The two-stage change flow - save a private draft, then commit and deploy - plus what gets committed, history, and revert.
- [API Explorer](https://www.fathym.com/docs/openx/guides/api-explorer): Browse, test, and integrate your endpoints from one place - filtered to what your access rights allow.
- [MCP: Claude (Code & Desktop)](https://www.fathym.com/docs/openx/guides/mcp/claude-code-desktop): Connect Claude Code or the Claude Desktop app to your OpenX workshop over MCP - one fai command for Claude Code, an mcp-remote bridge for the Desktop app.
- [MCP: Cursor & VS Code](https://www.fathym.com/docs/openx/guides/mcp/cursor-vscode): Connect Cursor, VS Code (Copilot), or Cline to your OpenX workshop over MCP.
- [Bring Any AI (MCP)](https://www.fathym.com/docs/openx/guides/mcp): Connect Claude, Copilot, app-makers, a local model, or any MCP client to your workshop - to read your sources of truth and to build new ones. Fire us, keep running.
- [MCP: No-Code Apps](https://www.fathym.com/docs/openx/guides/mcp/app-makers): Use your OpenX sources of truth from no-code/low-code builders - Dify, Langflow, Flowise, n8n, Make.
- [MCP: Custom Clients](https://www.fathym.com/docs/openx/guides/mcp/custom-clients): Drive the OpenX MCP server from your own client or agent over JSON-RPC (Streamable HTTP).
- [APIs](https://www.fathym.com/docs/openx/guides/apis): Every warm query is automatically a REST API endpoint - call it from dashboards, scripts, Power BI, Grafana, or any HTTP client.
- [Permissions](https://www.fathym.com/docs/openx/guides/team/permissions): The full role matrix - what Viewer, Editor, and Admin can each do - and how every role change is recorded.
- [Team & Access Rights](https://www.fathym.com/docs/openx/guides/team): Invite teammates, set roles, and scope access - so the right people (and the right AI proposals) can build, review, and ship.
- [Inviting Members](https://www.fathym.com/docs/openx/guides/team/inviting-members): Add teammates to your workshop - send an invite, track pending ones, and revoke access when needed.
- [User Management](https://www.fathym.com/docs/openx/guides/team/user-management): The complete access model - where users are managed, how they sign in, workshop vs surface scope, SSO, and running in your own cloud.
- [Simulator](https://www.fathym.com/docs/openx/guides/connecting-data/simulator): Generate realistic sample data with the built-in simulator - no backend - to build and test surfaces and queries.
- [Live Devices](https://www.fathym.com/docs/openx/guides/connecting-data/live-devices): Connect real devices and streams - get connection credentials, point your device at them, and rotate safely.
- [Connecting Data](https://www.fathym.com/docs/openx/guides/connecting-data): Get your information flowing into the workshop - a built-in simulator, a live device, or a third-party API you poll.
- [API Source](https://www.fathym.com/docs/openx/guides/connecting-data/api-source): Poll a third-party REST API on a schedule - no custom code - and feed the responses onto your surface.
- [Database](https://www.fathym.com/docs/openx/guides/connecting-data/database): How to get data from an existing database into your workshop today - pull it through an API Source - and what native database support is coming.
- [Reference](https://www.fathym.com/docs/openx/reference): API reference for OpenX - REST endpoints, auth and tokens, response formats.
- [Auth & Tokens](https://www.fathym.com/docs/openx/reference/auth-and-tokens): How the /oi-api is authenticated - Bearer JWT, expiry, scopes, and what is enforced on every call.
- [REST API](https://www.fathym.com/docs/openx/reference/rest-api): Call your warm queries and workshop data over HTTP - the /oi-api endpoints, auth, responses, errors, and limits.
- [Query returns empty](https://www.fathym.com/docs/openx/troubleshooting/by-symptom/query-returns-empty): Debug an empty result - simplest query first, time window, field names, filters one at a time.
- [No data showing](https://www.fathym.com/docs/openx/troubleshooting/by-symptom/no-data-showing): Diagnose an empty surface - connection status, surface wiring, time window, simulator test.
- [By Symptom](https://www.fathym.com/docs/openx/troubleshooting/by-symptom): Find a fix by what you are seeing.
- [Can't connect](https://www.fathym.com/docs/openx/troubleshooting/by-symptom/cant-connect): Diagnose a failing connection - status, credentials, network, and testing the source.
- [Common Issues](https://www.fathym.com/docs/openx/troubleshooting/common-issues): Error messages, likely causes, and quick fixes - data, connections, queries, permissions, surfaces.
- [Troubleshooting](https://www.fathym.com/docs/openx/troubleshooting): Find your problem and get back to work - by symptom or by component, plus common errors and an FAQ.
- [FAQ](https://www.fathym.com/docs/openx/troubleshooting/faq): Common questions - warm queries, where your data lives, Azi, bringing any AI, access, and save vs commit.
- [Activity Log](https://www.fathym.com/docs/openx/troubleshooting/by-component/activity-log): Activity log issues - events not appearing, integrity check fails, empty export.
- [Queries](https://www.fathym.com/docs/openx/troubleshooting/by-component/queries): Warm-query issues - empty results, syntax errors, slow queries - with KQL help and the API error codes.
- [By Component](https://www.fathym.com/docs/openx/troubleshooting/by-component): Find a fix by the part you suspect - connections, surfaces, or queries.
- [Connections](https://www.fathym.com/docs/openx/troubleshooting/by-component/connections): Connection health - error vs connected-but-no-data vs delayed, per connector type, and a test flow.
- [Surfaces](https://www.fathym.com/docs/openx/troubleshooting/by-component/surfaces): Surface issues - not found, access, no data - plus linking connections and managing many surfaces.
- [Corrections](https://www.fathym.com/docs/openx/troubleshooting/by-component/corrections): Correction issues - stuck pending, can't approve your own, original seems changed.
- [Glossary](https://www.fathym.com/docs/openx/glossary): The OpenX vocabulary - workbench, workshop, warm query, shared source of truth, proposal, the three ways to drive it, and the MCP terms.
- [Vision](https://www.fathym.com/docs/openx/vision): OpenX is the workshop where teams connect their information to AI, crystallize shared sources of truth, and build on it - owning every connection.
- [Fathym Documentation](https://www.fathym.com/docs): Build a tool with your AI, run it yourself, and share it with your team. Docs for fAI, OpenX, Holonic, and The Loop.
- [Session Continuity](https://www.fathym.com/docs/fai/how-intelligence-grows/session-continuity): How fai carries everything you built in one session into the next.
- [Right AI for the Job](https://www.fathym.com/docs/fai/how-intelligence-grows/right-ai-for-the-job): fai works with 16 coding agents. Match the right one to your tool, your team, and your workflow.
- [Seeing What Your AI Knows](https://www.fathym.com/docs/fai/how-intelligence-grows/seeing-what-your-ai-knows): Use fai --preview to inspect the tools, vault entries, and project context your AI reads at session start.
- [Decisions and Patterns](https://www.fathym.com/docs/fai/how-intelligence-grows/decisions-and-patterns): How fai learns from your explicit choices and detects your implicit patterns.
- [Intelligence Compounds](https://www.fathym.com/docs/fai/how-intelligence-grows/intelligence-compounds): Why session 50 is dramatically more capable than session 1.
- [Team Coordination](https://www.fathym.com/docs/fai/how-intelligence-grows/team-coordination): How multiple sessions and team members share a project vault.
- [Never Stop Working](https://www.fathym.com/docs/fai/how-intelligence-grows/never-stop-working): What happens when an AI tool hits its limit, and why your context survives it.
- [How intelligence grows](https://www.fathym.com/docs/fai/how-intelligence-grows): What appears over time when you keep working with fai - from fewer questions to automatic workflows.
- [Automation Emerges](https://www.fathym.com/docs/fai/how-intelligence-grows/automation-emerges): How routine work disappears as fai learns your patterns.
- [Staying on Track](https://www.fathym.com/docs/fai/how-intelligence-grows/staying-on-track): How fai learns your direction and maintains consistency across sessions.
- [Capabilities Crystallize](https://www.fathym.com/docs/fai/how-intelligence-grows/capabilities-crystallize): How repeated workflows become permanent workbench intelligence.
- [fai + Your AI Tool](https://www.fathym.com/docs/fai/getting-started/fai-vs-others): How fai works alongside Cursor, Claude Code, Windsurf, Cline, Copilot, and others.
- [Privacy and Telemetry](https://www.fathym.com/docs/fai/getting-started/privacy): What fai sends as usage analytics, what it stores, where the data goes, and how to turn it off.
- [Getting Started](https://www.fathym.com/docs/fai/getting-started): Install fai and complete your first session in under 5 minutes.
- [Installation](https://www.fathym.com/docs/fai/getting-started/installation): Install fai on Windows, macOS, or Linux and verify it's working.
- [Local Models](https://www.fathym.com/docs/fai/getting-started/local-models): Use fai with Ollama, LM Studio, LocalAI, or cloud API keys. No Claude required.
- [Your Workbench](https://www.fathym.com/docs/fai/getting-started/your-workbench): What a fai workbench is, what it contains, and how it grows over time.
- [Your First Session](https://www.fathym.com/docs/fai/getting-started/first-session): Run fai, pick your providers, work with your AI - and see what carries over to session 2.
- [MCP server](https://www.fathym.com/docs/fai/mcp-server): fai's MCP server makes your workbench available to any MCP-compatible AI tool.
- [MCP server — governance](https://www.fathym.com/docs/fai/mcp-server/governance): Tool exposure, rate limiting, authentication, audit logging, and data filtering for fai's MCP server.
- [Your Workbench](https://www.fathym.com/docs/fai/concepts/workbench): The workbench as professional AI identity - portable, persistent, yours.
- [Tuning What fai Remembers](https://www.fathym.com/docs/fai/concepts/tuning): Choose an embedding model and keep capture focused so fai stays fast and remembers what matters.
- [Resilience](https://www.fathym.com/docs/fai/concepts/resilience): How fai ensures your AI context survives tool switches, model changes, and outages.
- [Concepts](https://www.fathym.com/docs/fai/concepts): The core ideas behind how fai works - optional depth for the curious.
- [The Architecture Boundary](https://www.fathym.com/docs/fai/concepts/architecture-boundary): The 10% you touch and the 90% that runs invisibly - how fai separates creation from infrastructure.
- [Progressive Growth](https://www.fathym.com/docs/fai/concepts/progressive-growth): How your framework emerges from how you work, not from configuration.
- [Context](https://www.fathym.com/docs/fai/concepts/context): What fai means by context - and why keeping everything matters.
- [What is fai?](https://www.fathym.com/docs/fai/what-is-fai): fai is Fathym's portable persistent AI workbench - the layer that makes every AI you use smarter over time.
- [Inside a knowledge vault](https://www.fathym.com/docs/fai/vaults/knowledge-vault-internals): The four sub-vaults - context, pattern, decision, and petnames - that make up project and personal vaults.
- [Personal vault](https://www.fathym.com/docs/fai/vaults/personal-vault): How you work, everywhere - your patterns, preferences, and instincts across every codebase.
- [Project vault](https://www.fathym.com/docs/fai/vaults/project-vault): What this project knows - architecture decisions, patterns, and conventions built from sealed sessions.
- [Vaults - how fai stores what you know](https://www.fathym.com/docs/fai/vaults): Three vault types store everything fai captures. Git-backed, version-controlled, readable by any AI.
- [Session captures](https://www.fathym.com/docs/fai/vaults/session-vault): The active session storage - everything captured while a session is open.
- [Sessions](https://www.fathym.com/docs/fai/sessions): How fai sessions open, capture, seal, and propagate knowledge to the next session.
- [How sessions work](https://www.fathym.com/docs/fai/sessions/how-sessions-work): The three git branches behind every fai session: journal, pending, and kept.
- [fai docs](https://www.fathym.com/docs/fai): Documentation for fai - Fathym's portable, persistent AI workbench. Find commands, concepts, examples, and guides.
- [Session Comparison](https://www.fathym.com/docs/fai/examples/session-comparison): Session 1 vs session 50 - the compounding effect made concrete.
- [Real-World Workflows - E-Commerce](https://www.fathym.com/docs/fai/examples/real-world-workflows): How an e-commerce solopreneur's AI context compounds across sessions.
- [Week 2 Experience - Hardware Engineer](https://www.fathym.com/docs/fai/examples/week-2-experience): A hardware engineer's compounding context after 8 sessions of PCB work.
- [Examples](https://www.fathym.com/docs/fai/examples): Real scenarios showing what fai changes over time.
- [Day 1 Experience - Music Producer](https://www.fathym.com/docs/fai/examples/day-1-experience): A music producer's first session with fai - what it captures, and what their AI already knows in session 2.
- [Operating modes](https://www.fathym.com/docs/fai/operating-modes): Three synthesis levels - from zero-cost capture to always-on intelligence.
- [Glossary](https://www.fathym.com/docs/fai/glossary): Every fai term, precisely defined.
- [Commands](https://www.fathym.com/docs/fai/commands): The complete fai command reference - everything you can type.
- [Agents](https://www.fathym.com/docs/fai/agents): How fai works with each coding agent — what gets configured, what gets captured, and how fai runs alongside it.
- [Knowledge flow](https://www.fathym.com/docs/fai/knowledge-flow): How raw session captures become the compound intelligence your AI uses on startup.
- [Synaptic-Oriented Programming](https://www.fathym.com/docs/holonic/concepts/synaptic-oriented-programming): Declare what things are and what they do; structure emerges from how they compose. Guardrails as part of an entity, not a gate in front of it.
- [Transcend and Include](https://www.fathym.com/docs/holonic/concepts/transcend-and-include): Why a larger whole never erases the smaller one, and how the 4-tier bound keeps the holarchy stable.
- [Builder, Runner, Engine](https://www.fathym.com/docs/holonic/concepts/builder-runner-engine): How a holon goes from definition to execution - define with the builder, run with the runner, never touch the engine.
- [Before and After](https://www.fathym.com/docs/holonic/concepts/before-and-after): Five contrasts between ordinary code and a holon - and what a holon carries that a function doesn't.
- [Concepts](https://www.fathym.com/docs/holonic/concepts): What a holon is, how it's defined and run, and why one shape composes at every scale.
- [What is a Holon](https://www.fathym.com/docs/holonic/concepts/holon): Every entity is whole and part at once - the two tendencies, and the no-aggregation rule that makes composition safe.
- [Holonic · Architecture](https://www.fathym.com/docs/holonic): The open-source architecture every Fathym product is built on - one holon shape from Workbench to Platform.
- [Creating Holons](https://www.fathym.com/docs/holonic/guides/creating-holons): The definitive guide - taxonomy, the universal contract, Builder/Runner/Engine, code style, sub-holon types, and file structure.
- [Testing with Intents (IDD)](https://www.fathym.com/docs/holonic/guides/testing): Intent-Driven Development - declare what a holon intends to do; the intent travels with it, version-controlled and machine-runnable.
- [Guides](https://www.fathym.com/docs/holonic/guides): How to build holons - the definitive creating-holons guide, testing with intents, and advanced builder patterns.
- [Advanced Patterns](https://www.fathym.com/docs/holonic/guides/advanced-holon-patterns): Beyond the base contract - builder strategy functions, the four-layer vocabulary rule, and orchestrating holons with an inner Steward.
- [Holonic MCP Prompts](https://www.fathym.com/docs/holonic/reference/holonic-mcp-prompts): The six holonic/* development prompts the fAI MCP server exposes for scaffolding, auditing, and explaining holons.
- [Reference](https://www.fathym.com/docs/holonic/reference): Look things up - the holonic glossary and the holonic MCP development prompts.
- [Glossary](https://www.fathym.com/docs/holonic/reference/glossary): The holonic architecture vocabulary, defined.
- [Capability Patterns](https://www.fathym.com/docs/holonic/patterns/capability): The deterministic behavioral holon - typed input and output, services, output hooks, composition, and agents.
- [Agent Patterns](https://www.fathym.com/docs/holonic/patterns/agent): The AI behavioral holon - a typed input and a message-based handle. Services, spawn, self-governance, composition.
- [Patterns](https://www.fathym.com/docs/holonic/patterns): The per-holon-type building blocks - Workbench, tool, capability, operation, vault, all the same holon shape.
- [Governance Patterns](https://www.fathym.com/docs/holonic/patterns/governance): The guardrail holon - trust levels, allowlists, custom logic, composites. A constraint that's part of an entity, not a gate in front of it.
- [Workflow Patterns](https://www.fathym.com/docs/holonic/patterns/workflow): The long-running holon - declarative .Flow() graphs or imperative .Execute(), approval gates, parallel steps, durable state.
- [Operation Patterns](https://www.fathym.com/docs/holonic/patterns/operation): The strategy holon - composes Capabilities, Agents, and sub-Operations into named, runnable actions.
- [Composition Patterns](https://www.fathym.com/docs/holonic/patterns/composition): How the holons fit together - the hierarchy, vertical vs horizontal composition, the governance cascade, and which entity to use.
- [Mosaic — coming soon](https://www.fathym.com/docs/mosaic): Mosaic is coming soon: a calmer way to work with AI, together.

## Blog

- [Your LangChain Agent Returns Nothing. Let Your Coding Agent Read Its LangSmith Trace.](https://www.fathym.com/blog/2026/10/13-read-langsmith-traces-from-your-coding-agent): When a LangChain agent returns nothing, the cause is in its LangSmith trace. One fAI command lets every coding agent you use read that trace itself.
- [How Does Compounding Intelligence Work in fAI?](https://www.fathym.com/blog/2026/10/02-how-compounding-intelligence-works): What actually happens between a capture landing on disk and a coherent FAI-PROJECT.md: the seven vaults, how a synthesis pass picks what to read, and how it distills more material than fits at once without dropping anything.
- [Your AI's Context Should Stop Growing](https://www.fathym.com/blog/2026/09/09-your-ais-context-should-stop-growing): Keeping a record of your work and giving AI agents context are different jobs. The record should grow forever. What you put in front of a model should not.
- [What Is Git-Backed Context in fAI (and What Does Owning It Mean)?](https://www.fathym.com/blog/2026/08/19-what-is-git-backed-context): In fAI, git-backed context keeps your AI's understanding of your work in a plain git repo on your machine: what it holds, and what owning it actually buys you.
- [Your AI's Memory Expires in 30 Days](https://www.fathym.com/blog/2026/08/18-your-ais-memory-expires-in-30-days): Your AI's memory finally has a version history, retained 30 days, then maybe deleted. History you don't own isn't history; it's a retention policy.
- [I Was Stuck in PU, and I Liked It.](https://www.fathym.com/blog/2026/06/14-stuck-in-pu): Every viral 'I used this one prompt' post sells you more, and then it evaporates. I went the other way and told my AI to do less. It's called PU.
- [AI Doesn't Write Bad Code. It Amplifies Yours.](https://www.fathym.com/blog/2026/05/21-ai-doesnt-write-bad-code): 900 engineers found AI amplifies whatever you give it. Governance alone doesn't fix that. Context is the missing layer.
- [Your Local AI Has Unlimited Tokens and Zero Memory](https://www.fathym.com/blog/2026/05/19-local-ai-unlimited-tokens-zero-memory): Local AI tools give you unlimited free tokens, but every session starts from scratch. Context Engineering turns free tokens into compounding intelligence.
- [The Most Expensive Word in AI Is Yes](https://www.fathym.com/blog/2026/05/15-the-most-expensive-word-in-ai-is-yes): Sycophancy isn't a bug you can see. It's a ratchet that atrophies the judgment you'll need when things break. The fix is productive underspecification.
- [Your AI Finally Remembers - But Does It Think?](https://www.fathym.com/blog/2026/05/07-your-ai-finally-remembers): Incomplete tasks persist in working memory until you commit a plan to a trusted system. Your vault closes the loop.
- [Stop Renting Your AI](https://www.fathym.com/blog/2026/04/27-stop-renting-your-ai): 70% of AI tokens are wasted context overhead. Every token should leave a deposit. Build persistent context that compounds, not windows that reset.
- [See What Your AI Sees](https://www.fathym.com/blog/2026/04/10-see-what-your-ai-sees): 66% of developers call 'almost right' output their top AI frustration. See every tool, vault entry, and context your AI has with fAI's inspector.
- [Google Antigravity Has a Context Problem. fAI Fixes It.](https://www.fathym.com/blog/2026/04/05-antigravity-context-engineering): Antigravity's 1M token window doesn't prevent context rot. Context Engineering keeps your AI sharp across sessions, not just inside one window.
- [You Weren't Tired. You Were Drawn Down.](https://www.fathym.com/blog/2026/04/02-drawn-down): AI brain fry is real. The bottleneck moved from output velocity to judgment quality. Rest isn't lost productivity. It's where your next decision comes from.
- [The Guardrails That Let You Share](https://www.fathym.com/blog/2026/03/30-guardrails-that-let-you-share): 5.5% of open-source MCP servers have tool poisoning vulnerabilities. Six guardrails: visibility, rate limiting, auth, scope, audit, redaction.
- [Your Sessions Don't Belong to Claude](https://www.fathym.com/blog/2026/03/18-your-sessions-dont-belong-to-claude): Break tool silos. One vault, every agent you use. Your context shouldn't be locked in Claude or Cursor. Configure all your coding tools from one source.
- [What Stays When the Session Ends](https://www.fathym.com/blog/2026/03/11-what-stays-when-the-session-ends): Stop re-explaining projects every session. Context vaults compress knowledge across sessions. Every unit of work makes the next unit easier.
- [Benchmarks Are for Labs. Evals Are for You.](https://www.fathym.com/blog/2026/03/10-benchmarks-are-for-labs): Benchmarks score models in general. You need to score yours on YOUR task. Your accuracy number determines your entire product architecture.
- [The Leaderboard Won't Tell You What to Build](https://www.fathym.com/blog/2026/03/09-the-leaderboard-wont-tell-you): Leaderboard winners change quarterly. Models show 13%+ accuracy drops on unseen benchmarks. Build a portable base that any AI amplifies.
- [From Developer to Steward](https://www.fathym.com/blog/2026/02/14-from-developer-to-steward): AI generates. Humans govern. The shift from execution to judgment makes architectural thinking the most valuable developer skill.
- [The Play Button for Enterprise](https://www.fathym.com/blog/2026/02/07-the-play-button-for-enterprise): Enterprise AI needs a play button: one click from idea to governed deployment. Not a dashboard. A creation surface with guardrails built in.
- [I Stopped Hiring Developers](https://www.fathym.com/blog/2026/02/06-hiring-stewards): AI writes code now. The job changed: from 'write this function' to 'should this function exist?' Stewardship is the new developer skill.
- [The Best Frameworks Are Discovered, Not Chosen](https://www.fathym.com/blog/2026/02/04-frameworks-discovered): The best frameworks aren't designed upfront. They're discovered through usage, extracted from patterns that survived real work.
- [Infrastructure Survives the Bubble](https://www.fathym.com/blog/2026/01/31-infrastructure-survives-the-bubble): Three AI bubbles: wrapper companies pop, foundation models consolidate, governance infrastructure survives. Like fiber optic after the dotcom bust.
- [You Hired Me. You Also Hired My Workbench.](https://www.fathym.com/blog/2026/01/28-the-portable-workbench): Your AI tools and methods are yours. The code you build is the company's. 75% of knowledge workers use AI. 52% hide it. The new employment contract.
- [Everyone's Talking About AI Power. They're Getting the Math Wrong.](https://www.fathym.com/blog/2026/01/27-whats-your-base): The equation isn't AI^capabilities. It's YOU^AI. A newcomer and veteran with identical AI see vastly different results. Your base is the multiplier.
- [FOMU Is the Real AI Fear](https://www.fathym.com/blog/2026/01/24-fomu-is-the-real-ai-fear): FOMU paralyzes enterprises more than competition. 40 to 60% of qualified B2B deals die to 'no decision.' Inaction costs more than intelligent risk.
- [The Governance Layer Nobody's Building](https://www.fathym.com/blog/2026/01/22-the-governed-ai-control-plane): AI governance separates proposal from execution. Build systems with auditability, reversibility, and the pattern: AI proposes, you decide, systems act.