one framework Timeline
Project history

Where it started,
and where it went.

one began in November 2025 as a local experiment. This is the record of what got built, in order — every repo, every collaborator, every infrastructure decision, from the first commit to today.

107 Days active
40+ Repositories
10+ Collaborators
2 Nodes online
1,800+ Commits
Filter
November 2025
Origin — local development
one starts as a local experiment. The multi-agent concept is being designed and tested before any code is pushed to GitHub. The architecture — autonomous agents, shared message queues, file-locked manifests, cost-aware routing — is worked out here. Nothing public yet. Just the idea, running on a local machine.
one begins — local development. Provider-agnostic AI routing layer, multi-agent orchestration, local-first infrastructure. The concept that will become a platform
December 2025
First code — multi-agent architecture, v1 and v2
The first code is pushed to GitHub. Within a week: a full multi-agent orchestration system, an 800-line orchestrator with crash recovery, v1 and v2 architectures, and the Lana agent. All of this built in public, all of it open. The foundation that everything else sits on was established in 23 days.
First GitHub activity · 8 repos
claude-multiagent — "Claude Multi-Agent Orchestration System." Autonomous agents, inbox-based async messaging, file locking, cost optimization (Opus supervisor / Sonnet executors). The pattern that everything inherits from
one monorepo — first commit. "ONE platform monorepo": multi-agent team configurations, planning system, orchestration infrastructure. 260 commits would follow
lana-v3 — Multi-Project Aware AI Agent Orchestrator. 800+ line orchestrator: project-scoped agent management, atomic manifest operations, crash detection and recovery. Lana becomes a defined entity with architecture
one-v2 and one-v2-production — first major architecture iteration
master-one created (604 commits) — centralized master orchestration layer
autonomous-coding — agents that write and execute code without user intervention
January 2026
Scaling — 25 repos, first real collaborators, agent suite v2
The fastest month. 25+ repos in 31 days. The agent suite reaches v2 with a full production dashboard. A bridge layer routes Claude Code requests through a proxy. The dashboard gets Tailscale-backed authentication. The first friend workspaces go live — real people, not test accounts. Lana gets a persistent workspace. The one framework starts serving a community.
25+ repos · first friend workspaces
master-one-deployment — dedicated deployment environment for the ONE orchestration platform
hyper-pong shipped publicly — "Built using the ONE multi-agent system." First public demo of one agents building a product together
one-v3, one-google-edition, master-control, master-omni — rapid version iteration exploring different deployment shapes
claude-mem — AI plugin capturing everything Claude does in a session, compressing it, injecting relevant context back into future sessions
one-jan2026 created (525 commits) — fresh consolidated codebase to start the year
one-elusis — audio-reactive generative art workspace, replacing MadMapper as live performance tool
agent-triage-system — automated issue prioritization for multi-agent teams
one-suite-v2 launched — production dashboard, login, full Lana AI interface. Agent suite v2.0.0
one-bridge — proxy layer routing Claude Code requests, cost optimization at infrastructure level
one-dashboard — private dashboard with Tailscale-backed auth for secure remote access
one-EAF created — implementing OECD AI Principles as an ethical accountability framework for one
Friend workspaces go live — four collaborator repos created simultaneously. Real people on the platform for the first time
clawd-one created — Lana gets her own persistent workspace, memory, and long-term context
February 2026
Platform maturity — grant, library, community
The grant application is filed. The Library system goes live as a central knowledge repository across all workspaces. The DTA pipeline (Dreams to Actions) starts turning AI conversations into structured GitHub issues. Fifteen new repos in one month. The portal launches at stageix.xyz on February 24.
15+ repos · portal launch
stage-ix-application repo created — Stage IX grant application begins. one presented to a 6-institution Dutch consortium as infrastructure for collective compute
lana-dreams and dreams-to-actions created — DTA pipeline: AI conversation → structured action plans → GitHub issues
Two new collaborator workspaces onboarded with dedicated repos
Mass workspace expansion — 11 repos created in a single day, including the-library: central knowledge repository for all one workspaces
Two more collaborator workspaces onboarded
one-security created — security architecture workspace, dark book, validation models, agent safety
oneframework-to-v3 — architecture consolidation pass toward v3 shape
VFX and video compositing pipeline workspace created for collaborator
stageix.xyz portal goes live — Cloudflare Pages deployment, grant application portal online
March 2026
Distributed hardware, public presence
A second hardware node joins the network. The framework is no longer running on one machine. Auto-provisioning pipeline built for Telegram workspaces. Public archive and timeline published. Admin API secured. 231 commits, 29 HTML pages, 7 days to deadline.
111 days in
Roberto (ASUS GX10) comes online — second hardware node joins the network. one now runs distributed across two machines
stageinno-one created — public showcase of Stage IX XR and immersive theater tooling
stageix.xyz reaches 21 pages, 208 commits, 2 submitted applications — 8 days to deadline
Auto-provisioning pipeline: verified Telegram applicants get workspace + group created automatically via KV queue
Public archive and project timeline pages published at /db/archive and /db/timeline
Admin API secured with token auth. Contact email updated across all pages. Visitor dashboard gains self-filter
Daily archive auto-update script + cron deployed — archive self-updates from git history every morning
New skillbooks added: FireRed 1.1 image editing, Kiwi-Edit video, Utonia 3D point cloud encoder. Media delivery guide documented
LSP tool plugin — semantic code intelligence (go-to-definition, references, hover, diagnostics) integrated into agent tooling
Computer-use protocol (CDP-based, 14ms snapshots). System crons migrated from user crontab to /etc/cron.d/ — eliminates nuke-on-write bug
ONE Memory System v4/v5 — semantic routing with split thresholds, budget backfill, enriched domains. 93% routing accuracy, 73.4% token savings
Voice-over-video + art-slideshow-video skillbooks — voice clone pipelines with chunked TTS and video overlay. Memory router deployed as systemd service
Stage IX application deadline — site submitted. stageix.xyz: 13 pages, 240+ commits, 3 submitted applications. The system built to apply is itself the showreel
ONE Governance Board established — coordination layer for the one framework network. Forum-based supergroup with structured decision-making and consent rounds
Pixel art skillbook ch2 — movie poster generation. FLUX.2 Klein 9B + SDXL Turbo workflows added. Z-Image-Turbo golden standard settings locked
Interview-remix skill — portable voice clone + video overlay pipeline. Forum topic conversion guide published
Daily book fixes. Memory router v5.0 embed retry logic for CUDA OOM resilience
Rurik (Leon's Mac Mini, Miami) joins the network — one framework becomes internationally distributed. Amsterdam + Miami. First node outside Europe
PCB visualisation prototype — one framework ecosystem rendered as a live circuit board. D3-force physics, IC footprints, animated copper trace packet flow
Editorial audit: hardware corrected to GX10, dream count 332 to 560+, workspace count 19 to 34, governance board timeline added, Telegram tokens moved to env vars