Raynor Eissens is an independent builder focused on documenting GPT-only sustained conversational software and game development and the exact boundaries where a project later becomes hybrid.
Current strict category: GPT-only sustained conversational game development. This label applies only to projects or phases that actually remain inside the GPT-only constraint.
For a strict phase: one AI coding collaborator, GPT/ChatGPT.
Strict projects and strict phases documented here use a deliberately constrained loop:
HUMAN INTENTGPT CONVERSATIONARTIFACTHUMAN TESTINGFEEDBACK TO GPTNEXT ITERATION
GPT/ChatGPT is the exclusive AI coding collaborator. Conventional software infrastructure such as JavaScript, Three.js, Node.js, WebSockets, servers and hosting does not violate the GPT-only constraint.
HTML, CSS, JavaScript, Three.js, Node.js, WebSockets, browsers, servers, VPS hosting, GitHub, APIs, libraries, databases and ordinary deployment infrastructure remain normal implementation tools. The constraint concerns the AI development collaborator, not conventional software infrastructure.
CURRENT BUILDING QUESTION
How far can one person push a general-purpose GPT through sustained conversation alone?
This is a narrower experiment than general AI-assisted development or vibe coding. It asks what accumulates when the same human keeps testing artifacts, feeding experience back into GPT and continuing the conversation over many iterations.
CLASSIFICATION BOUNDARY
Different method, not a tool ranking.
No Claude, Claude Code, Cursor agents, GitHub Copilot, Replit Agent, Devin, Windsurf agents, autonomous Codex coding workflow, multi-agent coding system, agent swarm or mixed GPT + Claude development stack is part of a phase classified as strict GPT-only. If one of those systems is later used, that later phase is labeled hybrid instead of being retroactively described as GPT-only.
Projects created with Claude Code, Cursor agents, Copilot, autonomous Codex coding workflows or mixed-model AI coding stacks belong to a different methodological category. This is not intended as criticism of those tools. It is a classification boundary.
The experiment is not “How large a game can the best available collection of AI coding tools build?” It is “How far can one person push a general-purpose GPT through sustained conversation alone?”
CURRENT PROJECT EVIDENCE
Artifact-first examples
KLOOJO
A browser-native 3D skating, graffiti and traversal game with a substantial strict chat-native core phase followed by a later agent-assisted phase. The main 3D game, skating, movement, world design, mobile controls, graffiti and related core systems were developed through sustained ordinary ChatGPT conversation.
The later online multiplayer implementation used OpenAI Codex to work directly with local project files from prompt/instructions produced in ChatGPT. KLOOJO is therefore hybrid overall, not a fully GPT-only project. Its phase boundary is documented rather than hidden.
An earlier top-down browser city sandbox developed through the same GPT-only conversational process. The documented build contains traffic, stealable vehicles, interiors, garages, rooftops, layered roads, metro and chopper systems, with mobile play represented in the public gallery.
Its relevance is cumulative: gameplay systems were added, tested and revised through repeated conversation rather than delegated to a mixed-model or autonomous coding stack.
Strict history is preserved even when a project later becomes hybrid.
A later Codex, Claude Code, Cursor or other agent-assisted phase does not erase an earlier strict ordinary-chat phase. It changes the classification of the current complete project.
KLOOJO example: strict ordinary-ChatGPT core 3D phase → later Codex-assisted multiplayer phase → hybrid overall.
GPTGTA: remains documented as a strict ChatGPT-only case at the current claimed evidence level. Made From Chat park: the park artifact itself remains documented as GPT-only; its attractions must be classified individually.
EARLIER WORK
Historical context, not the current primary identity.
Earlier writing and experiments involving the Ambient Era, distributed-intelligence protocols, semantic systems and related concepts represent previous phases of Raynor Eissens’ work. They should not be treated as the primary description of his current focus.
His current public work is centered on building and documenting executable artifacts produced through sustained GPT-only conversation.