OB Labz OSRelease Updates

What's new

Latest updates and improvements to the AI Operating System.

v3.10.0Master OSAll OSMarch 29, 2026

→Voice assistant: Hey Jarvis wake word activates Claude Code with Ember voice (Emma Multilingual, Edge TTS, 41 languages)|Infrastructure control plane: DigitalOcean connected, Namecheap API (17 domains), 1Password integrated, unified infra map|Security hardening: git history cleaned, secret scanner installed, OB Net sovereign blueprint|Code v6 protocol: 15 agents, 3 human gates, scaffold-first architecture, 93%+ target success rate|WP Engine security plan: 9-phase hardening across 7 WordPress sites|Haiku-powered ClickUp ops agent for cost-optimized task management

v3.9.0Master OSAll OSMarch 20, 2026

→/scout skill: autonomous Knowledge Scout — monitors YouTube channels, transcribes videos, scores relevance, and ingests insights into the OS knowledge base daily
→/self-optimize skill: autonomous prompt optimizer based on Karpathy's autoresearch pattern — runs skills N times, scores against binary evals, mutates prompts, keeps winners
→YouTube transcription pipeline — any video can now be transcribed and analyzed for OS-relevant insights
→First scout scan: 16 videos analyzed across 2 channels (Nick Saraev, Nate Herk), 4 insights ingested
→New KB: agent-architecture-patterns.md — 13 documented patterns including verification loops, stochastic consensus, self-annealing, model tiering
→New KB: claude-code-skills.md — advanced skills reference with YAML options, eval patterns, /btw technique, skill type classification
→ClickUp workspace IDs hardcoded to eliminate MCP discovery overhead on every skill invocation
→All 22 OS skills classified as capability uplift vs encoded preference — 21/22 confirmed durable
→Global skills directory (~/.claude/skills/) created for identity-level skills across all projects

v3.8.0Master OSAll OSMarch 20, 2026

→/consensus skill: multi-agent stochastic consensus for strategic decisions
→/deep-ask skill: reverse prompting — 5 clarifying questions before major tasks
→/review skill: sub-agent verification loop for completed work
→4 email accounts integrated (Gmail x2, Zoho Mail, MS365) with unified /email inbox
→Email triage learning loop — auto-improves routing accuracy over time
→Knowledge base expansion: CCS v3 specs, Zoho Mail integration, agent architecture patterns
→Calibration + counterfactual learning loops for decision quality tracking
→Inbox snooze system for deferred message handling

v3.7.0Master OSCTO OSMarch 19, 2026

→Hourly bidirectional knowledge sync between Master OS and CTO OS
→CTO OS skills and agents auto-pulled to Master OS on each sync cycle
→Self-learning decisions and loop state sync bidirectionally across all OS instances
→Sync scripts on both sides running hourly via cron

v3.6.0Master OSMarch 19, 2026

→Deploy verification step — all deployments now verified live before marked as done
→Deployment tracking loop — self-learning system monitors deploy outcomes and failure patterns
→Auto-changelog update script — releases page updates automatically when features ship
→/release skill: one command to publish a new release to this page

v3.5.0Master OSMarch 19, 2026

v3.4.0Master OSPartner OSCTO OSDev OSMarch 19, 2026

→Project Agent System (PAS) v1.1 — Backend + QA agents running in parallel
→/orchestrate skill: plan and execute multi-agent pipelines with human approval gates
→Project context layer with schemas and role-aware agent configs
→Child OS instances can now run their own scoped agent pipelines

v3.3.0Master OSMarch 19, 2026

→Email actions routable directly to child OS instances
→/calibrate skill: process calibration proposals for skills, rules, and knowledge changes
→Pipeline/opportunity tracking learning loop — measures decision acceptance patterns
→Agent performance metrics loop — tracks output quality over time
→/sync-projects skill: auto-syncs project files with actual repo state

v3.2.0Master OSPartner OSCTO OSDev OSMarch 18, 2026

→Bidirectional messaging system — master OS ↔ child OS communication via Supabase inbox
→Zero-trust security architecture for all child OS instances
→/inbox skill: message dashboard — view, triage, and act on messages from all instances
→/notify, /reply, /recall skills for cross-instance communication
→/audit-child-os skill: detect configuration drift, content leaks, and security issues
→VS Code extension: forgot password flow for login

v3.1.0Master OSMarch 18, 2026

→Action Pattern Automation Engine — detects recurring command patterns and proposes automation
→Git auto-push agent running on crontab — OS auto-commits and pushes changes on schedule
→/learning-stats skill: evolution metrics dashboard, gap analysis, learning velocity
→/sync-learning skill: sync operational data into OS YAML files
→/briefing skill: daily intelligence brief — project status, blockers, action items at session start
→/pulse skill: weekly scorecard with project health, velocity, and quarterly rock progress
→Decision logging pipeline with outcome tracking

v3.0.1All OSMarch 16, 2026

→VS Code extension: setup wizard, Supabase auth, session management, update checker
→Voice foundation: STT (speech-to-text) + TTS (text-to-speech) pipeline, cross-platform
→Messaging sync via Supabase: outbox/inbox watcher with archive and flush
→Gmail integration: OAuth setup, onboarding wizard, AI-powered email triage
→/email skill: classifies as urgent / important / actionable / noise / spam
→/email-snooze skill: postpone emails to resurface later via file-based queue
→ClickUp MCP: full task, doc, and chat integration

v3.0.0Master OSPartner OSCTO OSDev OSMarch 15, 2026

→OB Labz AI OS v3 initialized — clean architecture, complete rebuild
→8-domain knowledge base: tech, business, revenue, finance-legal, operations, ai-data, risk, founder
→LEVEL 2 autonomy system: pre-approved actions execute freely, approval log tracked
→Core skills operational: /health, /improve, /update-kb, /task, /doc
→Multi-instance architecture: partner and company branches with isolated access
→Child OS distribution via GitHub fork model
→Self-learning system initialized: Stage 1 with decision log structure in place