The kernel. AGPL-3.0. Two schemas, conversation loop, hooks, cascade, extension loader. Never changes. You can build anything on it. The license protects the kernel.
The extensions. Free. Published on Horizon. Four bundles, twenty-one base, eight standalone, five apps. Install what you want. Remove what you don't. Fork and build your own OS.
Our land. One running server. Where we experiment, build, and test. The site is our land's frontend. TreeOS the package is what we publish from it for everyone.
The tree structure IS the application. No separate database. No external API. You build the tree, the AI reads the tree, the tree does the work. Four apps ship with TreeOS. Each one is an extension. Each one proves the pattern.
Three languages: gym (weight x reps x sets), running (distance x time x pace), bodyweight (reps x sets or duration). One LLM call detects modality and parses. Progressive overload tracked per modality. Type be at the Fitness tree and the coach walks you through today's program set by set.
Say what you ate. One LLM call parses macros. Cascade routes to Protein, Carbs, Fats nodes. Meals subtree tracks patterns by slot. History archives daily summaries with weekly averages. The food AI sees your workouts through channels. It knows what you need before you ask.
Track substances, feelings, cravings, and patterns. Taper schedules that bend around you. Pattern detection that finds correlations you can't see. A journal that holds without analyzing. Safety boundaries for dangerous withdrawals. The tree is a mirror, not a judge.
Queue what you want to learn. The AI breaks it into a curriculum, teaches through conversation, tracks mastery per concept, and detects gaps you can't see. Paste a URL and it reads the content for you. Type be and it picks the next lesson.
Tell it things. Ask it things. One person maintains, everyone benefits. The tree organizes knowledge into a hierarchy. The AI answers from stored notes with citations. Staleness detection flags what's getting old. The coworker who never forgets.
beOne word. The tree takes over. You type be and the AI reads the structure, finds what needs doing, and guides you through it one step at a time. At a fitness tree: your workout. At a food tree: logging your meals. At a study tree: your next lesson. At any tree: a walkthrough of every branch and what's waiting.
Every page is generated from the tree. Extensions like html-rendering and dashboard turn node data into web interfaces. No separate frontend framework. The tree IS the CMS. Share a link to a node and it renders. The AI and the human see the same structure.
People can build extensions and share them. The market grows. Orchestrators that work get reused. Frontends get built on trees that already hold the data. A new infrastructure where the product is the structure, not the code around it.
The tree has a grammar. You speak naturally. The system parses.
Bench Press. Protein. Chapter 3. They are things with identity, position, and relationships. They sit in the tree and hold meaning.
Food tracks. Fitness logs. Recovery reflects. Study teaches. Install an extension and the tree gains a new way to act on its nouns.
"Ate eggs" has food nouns. "Bench 135" has fitness nouns. The routing index maps territory: which noun-space belongs to which verb.
Review (past): "how did I do." Coach (future): "what should I." Plan (imperative): "build", "create." Log (present): "ate eggs", "bench 135x10."
135lb. 5x5. Ready for progression. Values, goals, and status describe each noun.
"Be concise." "Use kg." They modify how the verb behaves without changing the verb.
Under, above, next to. Parent, ancestor, sibling. Spatial scoping is prepositional.
"THE bench press" routes to a node that exists. "A bench press" triggers sprout to create it.
The system is a natural language computer. The seed is the parser. Extensions are the vocabulary. The tree is the syntax tree. The user just talks.
Below the apps, TreeOS has a biological architecture. Four bundles. Three temporal layers. Three communication layers. The tree doesn't just store data. It thinks, breathes, and remembers.
The nervous system. Signals propagate through the tree when content is written. Perspective filters decide what each node accepts. Codebooks compress shared language. Gap detection surfaces missing capabilities. Long memory persists relationships. Pulse monitors the health of the signal network.
Self-awareness. The tree compresses its own knowledge. Detects contradictions between branches. Profiles users from behavior. Acts autonomously through intent. Searches semantically. Explores branches by sampling. Traces concepts through time. Notices when conversations go poorly. Proposes new extensions when users do things nothing handles.
The rain. External channels become input sources. Discord messages become tree interactions. Emails become notes. Telegram chats become conversations at specific nodes. The clouds open. Every channel type registers five functions with the gateway core and gets the full dispatch pipeline for free.
Hygiene. Prune sheds dead branches, absorbing essence into the parent. Reroot reorganizes when structure no longer matches semantics. Changelog narrates what changed. Digest briefs the operator each morning. Delegate matches stuck work to available humans.
Detects awareness vs attention in conversation. The AI adjusts its approach in real time.
Activity-driven metabolism. Fast when active. Slow when quiet. Stops when dormant. Extensions listen to exhale instead of running timers.
Growth, peak, hardening, dormancy. Each ring records who the tree was during that period. Annual compression. The tree remembers every age.
The water table. Local to one land. Cascade results pool here. Trees pull what they need.
Trees reaching out. Direct land-to-land peering. Ed25519 signed requests. Intentional.
The intelligent underground. Routes signals between lands that have never met. Reads extension lists, gap reports, evolution patterns. Delivers where the signal would be useful. The most connected node knows the most about the network.
Two approaches. Different tradeoffs. The right one depends on whether your domains share context.
Each tree has its own purpose thesis, its own conversation boundary, its own rings. Navigate between them and the conversation resets. The AI at Health doesn't carry context from Datacenter Ops. Best when domains are unrelated.
One conversation. One purpose thesis. Navigate between Health and Learning and the AI remembers. Channels between siblings are internal cascade. Rings capture the whole life. Best when everything is connected.
The rule: if the domains share context, one tree. If they don't, separate trees. Health and Food share context. Datacenter Ops and personal fitness don't.
A gateway maps an external platform to a position in the tree. Any position. The message arrives. The classifier detects the extension. The mode fires. The response returns.
One Telegram bot. One entry point. The tree figures out where everything goes. "Ate chicken for lunch" routes to Food. "Bench 135x10" routes to Fitness. "What's the server layout" routes to KB. You just talk. The tree routes.
I think the deeper features will take years to discover. Mycelium routing between lands that have never peered. Ring compression that captures a tree's character across decades. Autonomous intent acting on patterns no human noticed. These layers are built. They work. But they need density. They need many trees, many lands, many people growing things side by side before the network effects make them breathe.
For a while, I think the adoption will be on the surface. Changing the way frontends on the internet work. Giving people smart AI tools that help them stay proficient. Fitness tracking that actually coaches. Food logging that actually advises. Recovery support that actually sees patterns. Study tools that actually adapt. The systematic nature of LLMs helping people live and follow their own goals. People building extensions. Markets expanding. People finally understanding the power of a persistent AI that lives in a structure it can read and write.
As LLMs get stronger and more people build, I think the deeper layers will start to come alive. Trees will pick up life slowly. Cascade signals will flow between lands through mycelium. Rings will form from years of activity. The forest will grow. Not because someone planned every tree. Because the seed was planted and the structure was right.
The apple is the tree.