Changelog

"AI assistants that remember your team's preferences across sessions and projects."

"Organic standard discovery: Patterns that emerge across projects become team standards."

"Block deprecated libraries, enforce naming conventions — all through natural language decisions."

"Code Discovery turns merged PRs into documented decisions — no manual tracking needed."

"Color-coded glows mean you can spot insight types at a glance from the taskbar."

"Dismiss forever keeps your notification center clean without losing data."

"Perfect for onboarding: Document years of decisions in minutes, not days."

"Evidence-backed: Each discovered decision includes proof from your codebase."

"Works with your Git provider: GitHub, GitLab, or Bitbucket."

"Zero configuration: Write decisions in natural language, AI infers what to enforce."

"Prevents architectural drift before it merges — not after."

"Links violations back to decisions so developers understand the "why"."

"Provenance answers "which branch introduced this decision?" — critical for governance audits."

"Archive reasons capture institutional knowledge: why was this path abandoned?"

"Double-click positioning means the graph layout matches your mental model from the start."

"Review Changes button provides instant access to diff view when working on a branch."

"Shootout mode lets you compare two alternative strategies side-by-side before deciding."

"Inspired by Git diff and PR review interfaces — familiar mental model for developers."

"Branches are additive layers on top of the live graph, not clones. The live graph stays read-only."

"Use "supersedes" to replace live decisions — the relationship IS the diff."

"Inspired by Git branches and Terraform plans: explore safely, commit deliberately."

"Start with Read-Only to get security approval, upgrade to Full Access when ready to publish."

"Docs now scale for future integrations (Linear, Slack, etc.)."

"Filter on the list, switch to graph — your filters follow you."

""By Status" sorts by urgency: decisions needing attention bubble to the top."

"Delegation solves the "vacation problem" — decisions don't stall when owners are away."

"Ratification adds a governance checkpoint without slowing down day-to-day decisions."

"Ask Claude "What did we decide about authentication?" and it queries Arbtr directly."

"MCP is the emerging standard for AI tool integration — we're early adopters."

"Git is the archive, Arbtr is the editor. One-way sync keeps things simple."

"GitLab users finally get first-class support — MRs, not PRs."

"The dashboard answers two questions: "Am I blocking anyone?" and "Is the system healthy?""

"Health score formula: 100 base, -15 per conflict, -5 per stale decision, +10 for active teams, +5 for work in progress."

"No more floating panel in the corner — graph issues join insights in the unified taskbar."

"Filter chips use semantic colors: violet for tags, sky blue for statuses, red/orange for graph issues."

"Bezier curves are the default for clean automatic routing. Switch to orthogonal when you need precise control."

"Inspired by professional diagramming tools — no Pro subscription required."

"The magic border uses a 1000% oversized spinning gradient clipped by overflow:hidden. Pure CSS, GPU-accelerated."

"Scroll-and-show means you never lose context when exploring insights across decisions."

"The taskbar is inspired by VS Code's status bar — always visible, never intrusive."

"Ripple effects show the "Chain Reaction": Source decision → Impacts → Downstream dependency."

"Arbiter runs in the background after every action — adding arguments, creating decisions, editing content."

"The panel docks at the bottom like VS Code's terminal. Expand to review, collapse to focus."

"Similar decisions at 70%+ similarity? Arbiter will tell you before you create a duplicate."

"A decision without links to the real world is just a diary entry. Now you can connect them."

"Context links answer "why did we decide this?" Implementation links answer "where is it being built?""

"Spark asks "What problem are you trying to solve?" — much friendlier than a blank form."

"The audit log captures 17 different event types. We track everything."

"Board meeting in an hour? Generate an executive briefing in 30 seconds."

"The chatbot requires 5+ decisions to activate. We need context to give good answers."

"Semantic search means "auth" finds "login" and "authentication". No more keyword gymnastics."

"We ditched auto-calculated "solidity scores" for explicit weights. You know better than an algorithm."

"Engineers hate forms. Now they can just paste their Slack thread and get a decision record."

"The "It's Me" banner appears when AI is confident about who you are. Pretty creepy, pretty cool."

"Our target users live in VS Code. They expect Cmd+K. We deliver."

"The graph is the money shot. It shows what lists never can."

"Domains were inspired by how architects actually think about systems."

"We built 6 relationship types because architecture is nuanced. "depends_on" and "enables" mean very different things."

"Day one decision: PostgreSQL over NoSQL. We never looked back."

"RLS means even our own bugs can't leak your data to other teams."