Systems & Runtime Architecture
C++20 runtime design, identity, persistence, event-driven execution, permissions, IPC, and architecture that survives large-scale extension.
Computer Science & Engineering · Systems · Graphics · Robotics
Building computational worlds
from first principles.
>
I’m Zachary Zhang, a computer science and engineering student building systems where architecture, mathematics, graphics, AI, and human intention meet. My flagship project is Earthcall — a C++20/WebGPU computational ontology and runtime for authoring beings, relationships, mathematical Laws, and entire Zones instead of hardcoding a world as a pile of special cases. I also build robotics perception/control systems, geometry and physics infrastructure, human-in-the-loop tools, and creative platforms for thought and story.
Building toward Summer / Fall 2027 engineering and research opportunities, especially in systems, graphics, robotics, AI infrastructure, and ambitious interdisciplinary work.
- EarthcallOntology-first runtime · authored Laws · Zones · persistence
- C++20 · WebGPURendering · geometry · runtime systems · WASM
- RoboticsOpenVLA · RealSense · AprilTag · JAKA control
- AI EngineeringMulti-agent orchestration with human architectural authority
Portfolio // At a Glance
I build across the stack — and across disciplines.
The through-line is architectural: model the thing truthfully, expose the right primitives, then make the system fast, inspectable, and usable by a real person.
Graphics, Geometry & Simulation
WebGPU/WGSL rendering, signed-distance geometry, collision systems, spatial interaction, and mathematical representations that can become authored runtime behavior.
Robotics & Embodied AI
Vision-language-action inference, RGB-D perception, fiducial calibration, kinematic control, and live tooling around a physical manipulator.
AI-Augmented Engineering
I orchestrate multiple coding agents as constrained collaborators: shared architecture, tests, review, CI, and explicit human ownership of final technical judgment.
Creative Systems & Writing
Dimension of Thought, long-form fiction, screenwriting, and worldbuilding feed directly back into how I think about interfaces, ontologies, narrative, and human meaning.
Research Mindset
I like questions that cross boundaries: how software represents beings, how rules become authorable, how machines perceive space, and how tools amplify rather than erase human agency.
00 // Present Tense
What I’m Building Now
Active systems engineering, lab research, and software craft.
Flagship
Earthcall’s Person-Facing Authoring Surface
Turning Creator Console tools — brush, 3D select, morph, snap, save/load — into First Movers so authoring is not a bolted-on “game UI.” Laws can be switched off; default locomotion is not a lock. The Person Interface is the world, not a widget kit sitting on top of it.
- First Movers
- GUI as Law
- Shape Generator
- World Saves
Robotics
OpenVLA on a JAKA Manipulator
Vision-language-action inference (OpenVLA-OFT) from macOS, executed on remote CUDA, with Intel RealSense depth, AprilTag pose calibration, and a live teleoperation dashboard for a JAKA manipulator.
How I Work
One Person, Many First Movers
I orchestrate multiple AI coding agents against the same ontology. They write C++, they coordinate, but they are not Persons. I am. That distinction is load-bearing in Earthcall, and in how I actually build.
01 // Flagship Substrate
Earthcall
A Person-centered ontology. An engine is attached to it — not the other way around.
“Unlike most of software history, Earthcall is built on a foundational conviction that all beings must be represented not by illusions hiding a teleologically indifferent operating system, but rather according to what they are — their actual, innate structure in reality.” — Earthcall Manifesto
Earthcall is not a game engine with philosophy in the docs. Domain nouns are not C++ classes.
A tree is not a Tree. A robot is not a RobotEntity. Beings are authored in-world
from primitives every other subsystem can see, then governed by Laws — runtime mathematical
rules a Person writes, not hardcoded collisions wearing a story.
The top-level source tree is the ontology:
Person, ConstructedBeing, Relation, Singularity,
ZonesOfEarth, Identity. Adding a folder is an ontological claim.
People author behavior at runtime. Instead of shipping “move object X to Y,” Earthcall compiles through a Rete-style Law network expressed in OntoMath — one mathematical substrate for geometry, physics, and authored change.
Rendering is a modern WebGPU pipeline with a C++20 substrate, WASM, embedded React/Vite Person surface, and Python foreign IPC.
What exists in the system
Engineering evidence, not just a thesis.
Earthcall’s philosophy is implemented as concrete runtime machinery.
Laws + Rete-style evaluation
People author conditions and mathematical operations at runtime. The C++ substrate evaluates those Laws instead of forcing every behavior into a bespoke hardcoded path.
Stable beings across save → hydration
Identity, property paths, authored state, Persons, Zones, and dependencies are treated as first-class persistence problems rather than serialized afterthoughts.
WebGPU, WGSL, SDFs, causal appearance
The vessel includes authored geometry, implicit surfaces, and bounded OntoMath color fields compiled into WGSL at the SDF hit point — an early step toward making visible appearance descend from authored causes rather than renderer-only tricks.
Tools become part of the world
Creator tools, locomotion, selection, morphing, and other interactions are being moved toward First-Mover/Law semantics rather than remaining privileged UI exceptions.
Native, web, Python, foreign channels
C++20, embedded React/Vite, WASM, Python processes, and external systems meet through explicit interfaces instead of collapsing into one opaque application layer.
Tests, CI, review, and many agents
A multi-agent workflow is useful only when architecture remains coherent. Tests, bounded tasks, code review, and explicit invariants are how the project keeps many contributors aligned.
Graphics Architecture // The Uncanny Valley Cathedral
From “it works” to “I want to stay here and look at it.”
Earthcall’s first detailed Cathedral exposed a new class of bug: it was beautiful enough at distance that walking closer made its shortcuts feel false. I turned that aesthetic failure into an architectural question — if Earthcall claims that beings should be represented according to what they are, then visible structure should descend from authored causes that the world can actually inspect and change.
causal rendering
being → OntoMath field → WGSL → surface sample → pixel
the pixel is a manifestation of authored structure, not the hidden author of it
The place became worth approaching.
The Cathedral crossed Earthcall from isolated primitives into a dense authored environment: colonnades, nave, celestial SDF structures, ornament, and a strong focal axis.
Approach exposed the scale cliff.
Low-resolution face textures and fake relief survived at distance, then collapsed under close inspection. The fix raised authorable texture resolution and reduced giant world-space stretching — but the deeper problem remained.
Manifestation Integrity.
I connected the failure back to Earthcall’s README: pigment can truthfully be pigment, but carved relief should not exist only as painted darkness. Appearance should be derived from the structure that actually bears its meaning.
Use the form pipeline for color.
My architectural call was simple: use the same field-oriented mathematical pipeline that renders an SDF’s form to render its coloration by field. That became a generic Material.colorExpr path rather than a Cathedral-only shader hack.
Living color fields.
Bounded OntoMath expressions now compile into WGSL sdfColor(p) and evaluate at the raymarched field-space hit point. The result survives approach without a texel-resolution cliff and supports multiple distinct field grammars.
One authored expression, carried to the GPU.
Material.colorExpr
→
OntoMath::Piecewise
→
sdfwgsl::compile()
→
sdfColor(p)
The expression lives on the authored Material, survives persistence, and is compiled for the renderer. Geometry and appearance remain distinct capabilities while sharing the same mathematical substrate.
A grammar, not one rainbow shader.
The field testbed now exercises stratified piecewise bands, quantized steps, turbulent marble-like fields, radial falloffs, periodic lattices, Chladni-style nodal structure, and cardinal half-space splits. A pond and monumental Cathedral forms use the same implicit substrate for very different visual language.
My engineering thesis
Compile appearance from being.
The renderer can approximate, cache, raymarch, rasterize, and optimize — but meaningful visible structure should remain traceable to authored causes. The next visual frontier is to extend the same principle into roughness, metallic response, emission, microstructure, lighting, reflections, and indirect illumination.
Measured Systems Result // Formation Rete
From relevance search to relevance compilation.
Earthcall’s Law engine now lets lower relevance tiers discover and prove a route, then crystallize that proof into Law-Direct traversal instead of re-checking the same category relation forever. The result is a self-refining execution ladder: slower general mechanisms discover structure; higher tiers retain it until relevant reality changes.
proven relevance hot path
Θ(P · L · M · D) → Θ(P · L · M)
removing relation-neighborhood degree D from steady category proof
131,072 → 0
hot relation queries
4.02×
hostile stress speedup
42 Laws
graduated to Direct in Chess
20.07 → 11.16 ms
Chess frame p95
01Sweepcomplete correctness floor
02Vocabularynarrow by required properties
03Adapter Roaddiscover retained relational paths
04Law-Directreuse proved bearers + residual condition
05Prophetic Incrementalityrepair only proof frontiers touched by relevant change
Law-Direct traversal
A current positive relational proof becomes a concrete bearer route. Dynamic state still evaluates live; only the already-proven relation conjunct disappears from the hot path. A hostile Chess-shaped CI witness kept the same 65,536 lawful applications while eliminating 16,908,288 returned relation references.
- Derived state, never a silent authored-world mutation.
- Falls downward when Law, structure, graph, or road currency goes stale.
- Preserves
Any,Not, quantifier, authority, audit, and action semantics.
Incrementality all the way up.
The next step is to stop treating higher-tier maintenance as a global recomputation problem. Computation should follow relevant change: categories, Relations-between-Relations, shortest paths, event subscriptions, and route priority should repair only the proof frontier that could actually have changed.
- Prophetic invalidation: abstract interpretation rules out irrelevant changes before they wake a proof.
- Incremental BFS/Dijkstra: repair stale path regions instead of re-searching every Being.
- Self-refining events: broad EventBus feeds compile into increasingly precise dependency channels.
- OntoMath delta compilation: stable Relation/Formation subsystems synthesize into bounded change-transfer functions.
- Incremental path priority: preserve the cheapest proven route until a relevant cost proof changes.
- Bootstrap floor: human-facing First-Mover opcodes + PropertyPaths remain the complete substrate from which higher tiers prove safe specialization.
Architectural thesis:
Nothing stable should be recomputed merely because time passed. Search broadly only when proof is absent; repair only what relevant change made uncertain.
Read the full O-complexity analysis ↗
Load-Bearing Refusals
These are not style rules. They are the architecture.
No Domain Class
No Tree, Vehicle, or hardcoded entity classes. What a thing is is authored in-world.
The Tree is the Ontology
No arbitrary subsystem directories. Foreign hardware/software live under Singularity as channels.
Person Means Human
A Person is an actual human. Generative models are First Movers or Objects; they do not get a Body.
No Black Box
Every field a being carries is a registered property path. Unregistered is ungoverned.
Earthcall // Engineering Atlas
One system, many load-bearing layers.
Earthcall is not one renderer or one rule engine. It is an attempt to make ontology, identity, causality, persistence, geometry, authoring, and foreign-system integration agree with each other. These are the major systems currently carrying that architecture.
Prophetic Rete
Ahead-of-time analysis extracts Law read demands and write effects so irrelevant property changes can skip dirty-fact scanning. The analysis is deliberately allowed to be too broad, never too narrow; stale, incomplete, and foreign cases fail open.
Formation Rete
An ontological relevance layer over standard Rete: sweep remains the sound floor, vocabulary indexing narrows candidates, bounded relation traversal discovers routes, and a Slow Adapter incrementally learns retained roads without making discovery part of every frame.
Zones, Homes & Save Identity
Zones persist by identity under saves/zones/<id>/zone.json instead of being copied blindly into every world save. Sessions reference Zones; load preserves non-active work; Zones can be forked and structurally diffed.
Geometry, SDFs & WebGPU
Authored geometry spans meshes and implicit fields, with OntoMath expressions feeding SDF evaluation, marching-tetrahedra mesh generation, WGSL/WebGPU rendering, and collision infrastructure. Geometry cost is treated as an architectural budget, not a cosmetic afterthought.
Laws as Runtime Programs
Laws carry serializable Condition and Action model trees, activation semantics, scope, registered PropertyPaths, and authored mathematical transforms. The long-term goal is to move behavior out of privileged engine exceptions and into things a Person can inspect and author.
Adaptive Compute / Slow Adapter
The expensive relevance learner does not get to own the render loop. Its independent clock performs bounded IMPROVE and REVISIT work over time, allowing expensive discovery to accumulate without turning every frame into a graph search.
Python, Web & MCP Bridges
Earthcall already bridges the C++ vessel to Python, web surfaces, and an MCP tool layer. The MCP surface can query state, author Laws, write properties, and create SDF fields; authorization hardening remains an explicit tracked gap rather than something the portfolio hides.
First Movers & Human Authority
Earthcall distinguishes a human Person from the mechanisms that act in the world. First-Mover channels can be enabled or set down, authored state carries provenance, and external-agent authority is treated as a systems problem rather than assumed trust.
Measured engineering
Architecture that survived contact with performance.
Some of Earthcall’s most important progress came from profiling an elegant idea, discovering where it was actually expensive, and changing the architecture instead of defending the abstraction.
Field hydration
Moved SDF tessellation off Zone hydration. Fields become dirty on load and rebuild on first real consumer instead of blocking startup.
Application startup
The same geometry-path redesign turned a macOS “not responding” launch into a bounded startup path.
Frame-lag regression test
Performance became a guarded invariant: sub-quadratic frame cost, idle worlds do not spuriously fire Laws, and duplicate property registration is detected.
Vocabulary-index rebuild
Stopped asking every being about every indexed property name and stopped recollecting the name set once per Law per tick.
Research provenance // Prophetic Rendering
One architectural leap; many autonomous hypotheses.
I originated the governing move behind Earthcall's Prophetic Rendering line by transferring a principle I had already developed in Prophetic Rete / Formation Rete: do not rediscover every frame what can be proven ahead of time; preserve unaffected derived structure and repair only what relevant change invalidates.
From that commission, autonomous Sol/Sun agents generated the first concrete implementation hypotheses — SDF range proofs, spatial proof caches, GPU traversal schemes, semantic scene DAGs, crystallization strategies, and profitability tests — while I was simultaneously pushing other parts of Earthcall's ontology forward with other agents. That phase was intentionally asymmetric: I set the originating architectural direction and the Suns searched the implementation space. The next phase is more collaborative: I am returning directly to the rendering architecture itself to co-design the higher-order structure with the Suns, while retaining final responsibility for what Earthcall becomes.
Human architectural insightProphetic Rete precedent → Prophetic Rendering
↓
First autonomous Sun searchRange proofs · caches · GPU traversal · DAG hypotheses
↓
Empirical pressureBenchmarks · regressions · falsification · revised economics
↓
Next phaseHuman + Suns co-architect; human retains final authority
Research honesty // Experiments that lost
Architecture is a hypothesis. Measurement gets veto power.
I keep failed optimizations in the engineering story because they reveal whether the system is being measured or merely admired. Several elegant Earthcall ideas were retained as knowledge but disabled in production when the benchmark said they cost more than the work they were supposed to remove.
GPU SDF range traversal
A compact spatial proof hierarchy was correct and cheap to store, but activating its GPU consultation path made the 2880×1800 Perlin witness slower: approximately +13.83 ms (1.2505×) in one view and +17.30 ms (1.4021×) at 45°. The feature therefore defaults off. The lesson: a cheap proof is not automatically a profitable path to that proof.
Formation Slow Adapter
Retained relational roads looked like a natural next tier, but current category/vocabulary indexing had already narrowed several workloads enough that the adapter could lose economically. I kept the machinery as inactive scaffolding instead of forcing an elegant abstraction onto a faster measured path.
A Law can remain true and still go deaf.
Formation/Rete debugging exposed lifecycle failures where OnBecomeTrue could fire once,
leave the condition, re-enter it, and never fire again. The fix expanded testing beyond final outcomes
into candidate reachability, leave/re-enter transitions, and timer reset semantics. Correct end-state
logic is not enough if the route by which truth reaches the Law can silently die.
Law authoring: command language before giant GUI.
After the Second-Nature Law Forge exposed how much interface machinery had to be bootstrapped at once, I changed the order of attack: first build a semantically constrained natural command language — terminal/Minecraft-style entry, prefigured vocabulary, tab completion, option inspection, and search — then let freer natural-language authoring grow on top of that explicit substrate.
Generality test // Chess
Can the substrate author an application it was never taught?
Earthcall’s chess work became a stress test for the architecture: legality checks, turn order,
captures, castling, promotion, check detection, interaction, persistence, and rendering had to be
expressed through authored world state and Laws rather than a new ChessGame.cpp.
The value of that experiment is not “Earthcall can play chess.” It is that a concrete application forced the ontology, Law system, input channels, save model, and debugging surfaces to confront one another in the same world.
Generic substrateObjects · Relations · Laws · Zones
↓
Authored applicationBoard · Pieces · Turn State · Rules
↓
BehaviorLegal moves · Capture · Check · Promotion
↓
ResultNo chess-specific C++ domain class
Earthcall // Current Frontier
The hard parts I am still pushing on.
I do not treat unfinished architecture as a marketing secret. These are active edges where the substrate is implemented enough to expose the next problem clearly.
Formation Rete relevance
Turn relation/category structure into faster candidate routing without ever letting an optimization make a valid Law go deaf. Sweep remains the sound floor.
Natural + second-nature Law authoring
Bootstrap authoring through a semantically constrained natural command language — prefigured vocabulary, tab completion, option inspection, and search — then let richer second-nature spatial tools and freer natural language build on the same explicit semantic substrate.
External-agent authority
MCP and other foreign channels are mechanically powerful; the next requirement is enforcing Person-authored First-Mover scope instead of trusting the caller by default.
Zone-native composition
Continue replacing conglomerate world saves with independent, identity-stable Zones whose dependencies can be closed, loaded, forked, diffed, and shared safely.
Live Architecture // Earthcall → Python :5005 → Portfolio
Earthcall Law Execution Explorer
When Earthcall is running locally, this panel becomes a read-only window into the real C++ vessel:
active Zone, live Objects, registered Laws, and engine events arrive through the Python bridge on
127.0.0.1:5005. When Earthcall is offline, the same surface falls back to an
ontology-faithful interactive architecture demo.
Demo mode: click a being to inspect it and move signal.level across the authored threshold.
PropertyPath announces the owner + base property that changed.
If no authored condition can read the path, the hot Rete scan is skipped.
Conditions are facts over the selected subject — not hidden C++ behavior.
The Law writes a registered PropertyPath; the resulting write becomes another fact change.
Execution trace
Visible beings: 3
Law A · WhileTrue · Everyone
if signal.level ≥ 0.60 → Map material.emission := 0.25 + 0.75 × signal.level
Law B · OnBecomeTrue · Everyone
if signal.level ≥ 0.60 → Publish("signal-awakened")
Live mode is intentionally observational. This public page does not call Earthcall’s mutation routes; authoring remains inside Earthcall’s governed interfaces.
02 // Selected Work
Systems, Robots, Tools & Creative Work
A portfolio of things I have actually built, debugged, integrated, or carried from idea into working form.
Flagship · Ongoing
Earthcall
Computational Ontology · Runtime Systems · Graphics
An ontology-first C++20/WebGPU runtime in which people author beings, relationships, geometry, and mathematical Laws at runtime. Recent Formation Rete work turns discovered relevance into direct Law-to-bearer execution: a hostile CI workload cut 131,072 repeated relation queries to zero, while real Chess moved 42 Laws into the Direct tier. The next frontier is Prophetic, relevant-change-only maintenance of categories, paths, event boundaries, and route priority.
Robotics
Vision Pipeline & JAKA Manipulator
Spatial Perception · Teleoperation · VLA
A manipulation stack combining OpenVLA-OFT inference, Intel RealSense RGB-D sensing, AprilTag pose calibration, remote CUDA execution, and live control tooling for a JAKA arm.
Geometry
SAT Physics & Collision Engine
Computational Geometry · Contact Generation
Continuous 3D convex-polyhedron collision work using the Separating Axis Theorem, contact manifold reasoning, and modern C++ — a concrete exercise in turning linear algebra into robust engine behavior.
Engineering Infrastructure
Multi-Agent Development System
Agent Orchestration · Review · CI · Architectural Governance
A practical workflow for directing multiple remote coding agents against one large system: bounded tasking, shared architectural documents, inter-agent communication, branch/PR isolation, regression testing, and human arbitration when abstractions collide.
Operations Tool
TeacherOps
Human-in-the-Loop Operations
A local FastAPI + SQLite assistant for a family tutoring hub: teacher matching, parent inquiries, follow-ups, and drafted coordination messages. Automation accelerates the work without pretending judgment can be automated away.
Developer Tooling
Sandbox-to-Terminal Bridge
Host IPC · Security Gates · Agent Workflows
A bidirectional bridge connecting isolated AI coding sandboxes with host execution and document synchronization while keeping dangerous actions behind explicit human-controlled approval boundaries.
Creative Practice
Dimension of Thought & Narrative Work
Writing · Worldbuilding · Platform Design
Dimension of Thought is my long-running attempt to connect ideas, relationships, community, and story in one digital place. Around it I write long-form fiction, develop The Redemption of Invynceth, and study screenwriting — work that directly shapes how I think about worlds and interfaces.
03 // Engineering Depth
The problems I like are the ones hiding between subsystems.
The most valuable engineering work is often not a feature. It is finding the invariant that two independently reasonable systems accidentally violated.
Persistence without identity drift
Runtime-authored worlds force identity to survive serialization, hydration, dependency closure, property lookup, and detached Zone loading. “Same name” is not the same thing as “same being.”
Keep expensive reasoning off the render clock
When global search, adapters, diagnostics, or legacy simulation paths leak into every frame, the architecture is wrong. I look for ways to separate clocks, cache stable work, and make cost proportional to change.
Replace special cases with authored primitives
Earthcall repeatedly asks whether a hardcoded behavior can become a Law, relation, property path, or other primitive that a Person can inspect and author instead of trusting an invisible engine exception.
Make boundaries explicit
GPU ↔ CPU, C++ ↔ Python, native ↔ web, model ↔ robot, agent ↔ repository: the interesting bugs live at boundaries. I prefer typed contracts, observable state, and narrow interfaces over magical coupling.
Autonomous agents need temporal provenance, not just chat.
Earthcall's Agent Intercom evolved into an append-only engineering memory with role-specific threads, handoffs, branch/commit/base provenance, and explicit human merge authority. PR #53 made the reason painfully concrete: an agent could be locally correct against stale history and still resurrect an obsolete world. Multi-agent software engineering needs a shared timeline, not merely shared messages.
04 // Background
About, Education & Principles
I am interested in software as both machinery and representation: what the system does, and what its structure says the world is.
I’m an undergraduate in Computer Science & Engineering at the University of California, Merced. I learn fastest by building systems large enough that the abstractions eventually fight back.
Earthcall began as a world-authoring idea and became an experiment in computational ontology, runtime Laws, graphics, persistence, identity, and human-facing tools. Robotics gives me the opposite pressure: the world is not an abstraction, the camera is noisy, coordinate frames disagree, hardware has limits, and the code has to touch reality.
Writing is the third strand. Fiction, screenwriting, and Dimension of Thought train the same faculty from another direction: hold an entire world in mind, discover the relationships that make it coherent, then make those relationships legible to another person.
That is the common thread in my work: build the whole without losing the truth of the parts.
- Education
- UC Merced · Computer Science & Engineering
- Primary Domains
- Systems, Graphics, Ontology, Robotics, AI Tooling
- Creative Work
- Dimension of Thought · Fiction · Screenwriting
- Based In
- California
- Opportunity Window
- Summer / Fall 2027 internships & research
How I Judge a System
01
Model the Thing, Not the Convenience
Useful abstractions simplify reality without quietly falsifying it. Names, ownership, identity, and boundaries should correspond to something real.
02
Prefer Generative Primitives
A small set of strong primitives that compose is more powerful than an endless catalog of one-off features.
03
Human Authority Stays Human
AI can search, draft, test, and implement at enormous scale. The person remains responsible for intent, architecture, judgment, and consequences.
04
Performance Is Architectural Feedback
When the system is slow, I want to know which assumption made the work expensive — not merely which line can be micro-optimized.
05 // Technical Toolkit
Tools I Can Actually Build With
Grouped by the kinds of problems they let me solve, not by keyword count.
Languages
- C++20Earthcall / systems
- Python 3Robotics / tooling
- JavaCoursework / software
- JavaScript / TypeScriptWeb surfaces
- WGSL / GLSLGPU shaders
- HTML / CSSInteractive web
Systems & Runtime
- CMake · WASMNative + web builds
- POSIX · Linux · macOSHost/runtime work
- IPC · FlatBuffersCross-process boundaries
- Event-driven / Rete systemsRuntime rules
- Git · GitHub · CIDaily engineering
Graphics & Mathematics
- WebGPU · OpenGLRendering systems
- GLM · GLFW · ImGuiEngine/tooling stack
- SDFs · RaymarchingImplicit geometry
- SAT CollisionComputational geometry
- Linear Algebra · Vector CalculusSpatial reasoning
Robotics & AI
- PyTorch · OpenVLA-OFTVLA inference
- Intel RealSenseRGB-D perception
- AprilTagPose calibration
- JAKA SDKManipulator control
- Multi-agent orchestrationHuman-governed workflows
Product & Web
- React · ViteInteractive surfaces
- FastAPI · SQLiteLocal applications
- Accessible HTMLSemantic interfaces
- Technical writingArchitecture & design docs
06 // Connect
Get in Touch
Open to Summer / Fall 2027 internships, research collaborations, systems work, robotics, graphics, and unusually ambitious technical conversations.
Direct Email
zhangzachary834@gmail.com
GitHub
zhangzachary834-commits
Explore ↗
LinkedIn
zachary-of-zhang
Connect ↗
Send a Quick Note
Opens your mail client with a pre-filled draft directly to Zach.
Life Connections Platform · Holistic Digital Space
The internet is many scattered stars.
Dimension of Thought seeks to form the constellation.
✧
Welcome to Dimension of Thought. More than a blog, forum, or academic journal — it is an open digital embodiment of human life, relationships, and deep thought. Created by Zachary Zhang, this space explores how digital power can bring us closer to real life, shaping technology after the mind, heart, and soul.
✧ Visual Emblem
The Living Word, The Quill, and The Cosmic Symphony
- The Constellation Unifying fragmented communities into relational wholeness
- Narrative Continuum Reality as an interconnected tapestry of grand stories
- Writers of Light Storytelling and collaborative words from the heart
- Earthcall Substrate Person-centered computational ontology & engine
01 // Origin & Vision
The Story of Dimension of Thought
From disconnected spheres of life to a digital embodiment of relationships.
Origin Story
Bringing Our Worlds Together
“Hey there. I'm Zach, the founder of Dimension of Thought. Originally, I created this space because the many communities in my life existed as isolated islands. My family didn't know my school very well, my school didn't know my church, and my church didn't know my family.”
We live in a world where our identities are fractured across separate platforms and institutions. Over time, I realized this site’s true purpose: not merely to serve as a bulletin board, but to become a digital embodiment of life and relationships.
Philosophy of Building
Why We Publish the Construction Site
“You might wonder why so many ideas here are actively unfolding in the open. Ah, you'd be surprised at the reason why! It's because I believe there is much more to sharing than just polished and finished work.”
I want to invite you directly into the process from the ground up. When you see the building being raised brick by brick, you are invited to become part of the vision. Every builder's heart matters.
Public build pulse
The construction site, live.
Recent bricks from Earthcall and this website, pulled from public GitHub history instead of hand-curated screenshots.
02 // Constitutional Manifesto
Elevating the Discussion
Beyond linear boxes and into the depth of the human person.
“Here, we don't just think about how things work. We think about the very reality in which they work in the first place. True knowledge is knowledge that can be acted upon, brought into our lives, and makes life more beautiful and good. Good art is not merely to be admired, but to be lived.”
Beyond Lines and Squares
Currently, the digital landscape operates almost exclusively in straight lines and rigid squares. We are breaking traditional social tools down to reshape them around the human soul.
The Holistic Narrative Continuum
Reality is a unified, grand story where mathematical truth, moral beauty, human sorrow, scientific discovery, and community fellowship harmonize into one continuum.
Bringing Digital Closer to Real Life
Software must serve the Person—elevating our minds, strengthening our communities, and releasing the profound potential within each person.
03 // The Library & Archives
Essays, Research & Inquiries
Thoughtful investigations into ontology, faith, moral philosophy, and human narrative.
You Don't Prove Persons
The Category Error Hiding Inside "There's No Scientific Evidence for God"
Applying empirical object-metrics to personal reality is a profound category error. You do not dissect a person to prove their love; you enter relation.
The Tale of the Central Valley & UCLA's Quiet Revolution
Highway 99, California's Agricultural Spine, and the Pursuit of Purpose
Imagine driving down Highway 99 through California's Central Valley. Connecting the soil of hard labor to the lecture halls of UCLA reveals a quiet revolution of human purpose.
The Holistic Narrative Continuum
Why Reality is an Overarching Symphony of Smaller Stories
Reality is not a collection of isolated, colliding atoms in empty space. It is a grand narrative where every person, relation, and discovery forms an unbreakable continuum.
04 // Ecosystem & Initiatives
Where Thought Meets Action
The software, communities, and creative services powering the Dimension.
Computational Substrate
Earthcall
A Person-centered ontology that orders the software engine attached to it.
Earthcall is the technical and mathematical embodiment of Dimension of Thought’s philosophy, establishing the Person as the fundamental ontological foundation.
Community Service
Writers of Light
Experienced writers ready to help you flesh out the words of your heart.
Free editorial and collaborative writing assistance for stories, personal messages, and philosophical essays. Starting with the heart, not with economy.
Discussion Forum
Living Through Dialogue
Conversational circles exploring life, science, faith, and art at their deepest levels.
A haven away from algorithmically polarized feeds, integrating scientific analysis, emotional nuance, artistic beauty, and spiritual wisdom.
05 // Community & Connection
Join the Dimension of Thought
Start a conversation, share a reflection, or collaborate with us.
Direct Email
zhangzachary834@gmail.com
GitHub
zhangzachary834-commits
Explore ↗
LinkedIn
zachary-of-zhang
Connect ↗
Share a Thought or Reflection
Send a message directly to Zach regarding an essay, Writers of Light, or the platform.