You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
feat(brand): framework-agnostic messaging, logo in README, remove scorer from GitHub Packages, remove em dashes from prose
- Replace arrow icon with logo SVG in README header
- Retitle as "The Open Standard for Human-AI Collaboration: AI Fluency"
- Remove direct 4D Framework coupling from platform-level descriptions
- Remove scorer from GitHub Packages release step (fluently-scorer is on npm)
- Replace all em dashes in user-visible prose across site HTML files
- Make site copy framework-agnostic at platform level; keep 4D refs where content is specifically about 4D
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
AI is powerful but not magic. The gap between good and poor AI-assisted work is not the model — it is the collaboration. Fluently is an open platform that operationalizes **human-AI collaboration frameworks** as a shared knowledge commons, a scoring engine, a CLI, and an MCP server.
14
20
15
-
AI is powerful—but not magic. Teams shipping with AI today face a hard truth: **output quality depends entirely on clarity**. You need to know *what to delegate*, *how to describe it*, *how to verify it works*, and *who's accountable*. That's the **4D Framework**: Delegation, Description, Discernment, Diligence.
21
+
Fluently ships with the **AI Fluency 4D Framework** by Dakan & Feller (Delegation, Description, Discernment, Diligence) as its bundled standard. Any framework with named dimensions can be registered alongside it, enabling teams to benchmark the same task across multiple frameworks and choose the one that fits their context.
16
22
17
-
**Fluently operationalizes the 4D Framework.** It gives you a CLI, MCP server, and a growing knowledge base of best practices—built on Anthropic's AI Fluency research by Dakan & Feller. Score your team's AI workflows. Learn from the community's Fluently 4D cycles. Ship with confidence.
23
+
**Agent-agnostic.** Works with Claude, GPT, Gemini, Mistral, Copilot, and any MCP-compatible client.
24
+
25
+
**Framework-agnostic.** Start with 4D. Bring your own framework. Compare them side by side.
18
26
19
27
---
20
28
@@ -26,31 +34,31 @@ AI is powerful—but not magic. Teams shipping with AI today face a hard truth:
26
34
$ fluent compare --task "Automated code review with human sign-off"
Each suggestion links to the full Fluently 4D cycle—delegation guidelines, framing prompts, hallucination patterns to watch for, and approval workflows.
53
+
Each result links to the full cycle: delegation guidelines, framing prompts, hallucination patterns to watch for, and approval workflows.
46
54
47
55
---
48
56
49
57
## Install in 30 Seconds
50
58
51
59
**Option 1 — Zero install (browser):** Use the [live scorer](https://fluently-org.github.io/fluently/) right now. No setup required.
52
60
53
-
**Option 2 — Zero install (terminal):** Run instantly with npx — no PATH changes needed:
61
+
**Option 2 — Zero install (terminal):** Run instantly with npx, no PATH changes needed:
54
62
```bash
55
63
npx fluently-cli score "Automated bug triage with human review"
56
64
```
@@ -61,27 +69,35 @@ npm install -g fluently-cli
61
69
fluent --help
62
70
```
63
71
64
-
Requires Node.js 20+. The CLI is self-contained — no configuration files needed.
72
+
Requires Node.js 20+. The CLI is self-contained, no configuration files needed.
65
73
66
74
---
67
75
68
76
## What's Inside
69
77
70
-
**CLI** — Run `fluent` in your terminal to score 4D workflows, compare against community patterns, and contribute new Fluently 4D cycles.
78
+
**CLI** — Run `fluent` in your terminal to score workflows, compare against community cycles, and contribute new patterns across any registered framework.
79
+
80
+
**MCP Server** — Embed collaboration scoring into any AI agent or MCP-compatible client (VS Code Copilot, Cursor, Continue, Claude Desktop, Claude Code, and more). Tools like `find_relevant_cycles()`, `get_dimension_guidance()`, `evaluate_compliance()`, and `compare_frameworks()` are available to your agent.
71
81
72
-
**MCP Server** — Embed 4D scoring into any AI agent or MCP-compatible client (VS Code Copilot, Cursor, Continue, Claude Desktop, Claude Code, and more). Expose tools like `find_relevant_cycles()`, `get_dimension_guidance()`, and `contribute_cycle()` to your agent.
82
+
**Knowledge Base** — Community-contributed collaboration cycles organized by domain (coding, writing, research, education, legal, healthcare, general). Each cycle belongs to a framework, carries per-dimension guidance, and is validated before merge.
73
83
74
-
**Knowledge Base** — 50+ community-contributed Fluently 4D cycles organized by domain (coding, writing, research, management, product, etc.). Each cycle is scored on the 4D dimensions and peer-reviewed before merge.
84
+
**Scorer** — The shared engine (`fluently-scorer` on npm) that validates schemas, computes similarity, scores collaboration quality, and evaluates framework compliance. Used internally by the CLI and MCP Server. Import it directly if you need programmatic access:
This isn't just a framework—it's a **commons for AI fluency**. Every Fluently 4D cycle you contribute teaches thousands of teams how to collaborate smarter with AI.
96
+
This is not just a framework. It is a **commons for AI collaboration literacy**. Every cycle you contribute teaches teams how to collaborate smarter with AI, regardless of which model or tool they use.
81
97
82
-
Built on the **AI Fluency 4D Framework** by Dakan & Feller. Everything — code, knowledge cycles, and tooling — is released under the **MIT License**.
98
+
Built on the **AI Fluency 4D Framework** by Dakan & Feller as the founding standard. Everything (code, cycles, and tooling) is released under the **MIT License**.
83
99
84
-
[Browse the Knowledge Base →](https://fluently-org.github.io/fluently/)
100
+
[Browse the Knowledge Base](https://fluently-org.github.io/fluently/)
85
101
86
102
---
87
103
@@ -90,45 +106,55 @@ Built on the **AI Fluency 4D Framework** by Dakan & Feller. Everything — code,
90
106
### For Teams Using AI
91
107
92
108
1.**Install the CLI** and run `fluent compare --task "your workflow"` to see how others solve similar problems.
93
-
2.**Check the Fluently 4D cycles**— find your use case in the knowledge base above.
94
-
3.**Score your process**— run `fluent score` with your delegation intent to identify gaps.
109
+
2.**Check the cycles** in the knowledge base for your use case.
110
+
3.**Score your process**with `fluent score` to identify gaps.
95
111
96
112
### For Teams Building AI Tools
97
113
98
114
1.**Integrate the MCP server** into your AI agent or IDE plugin (any MCP-compatible client).
99
-
2.**Expose 4D scoring** to your users — let them verify AI output quality before shipping.
2.**Expose compliance scoring** to your users with `evaluate_compliance()` so they can verify collaboration quality before shipping.
116
+
3.**Use the scorer engine** directly with `npm install fluently-scorer`.
101
117
102
118
### For Contributors
103
119
104
-
**Share your Fluently 4D cycles and help the community.** We're looking for patterns, anti-patterns, and real-world lessons from teams shipping AI features.
120
+
**Share collaboration cycles and help the community.** We are looking for patterns, antipatterns, and real-world lessons from teams shipping AI features.
105
121
106
-
→ [**Contributing Guide**](CONTRIBUTING.md) — Detailed walkthrough for submitting a new Fluently 4D cycle.
122
+
[Contributing Guide](CONTRIBUTING.md) — Detailed walkthrough for submitting a new cycle.
107
123
108
124
---
109
125
110
-
## How 4D Scoring Works
126
+
## How Compliance Scoring Works
111
127
112
-
Every task, workflow, or prompt gets scored across four dimensions:
128
+
Every collaboration session can be evaluated against a framework's named dimensions. The 4D Framework uses four:
113
129
114
130
| Dimension | Question | What We Measure |
115
131
|-----------|----------|-----------------|
116
-
|**Delegation**|*Who decides?*| Clarity on AI autonomy vs. human oversight. Escalation triggers matter. |
|**Discernment**|*Is it right?*| Your ability to spot hallucinations and overconfidence. Red flags & green signals. |
119
-
|**Diligence**|*Who's accountable?*| Governance, review workflows, audit trails. Who approves before shipping. |
132
+
|**Delegation**| Who decides? | Clarity on AI autonomy vs. human oversight. Escalation triggers matter. |
133
+
|**Description**| What context? | Quality of framing, examples, and constraints. Reduces ambiguity. |
134
+
|**Discernment**| Is it right? | Your ability to spot hallucinations and overconfidence. Red flags and green signals. |
135
+
|**Diligence**| Who's accountable? | Governance, review workflows, audit trails. Who approves before shipping. |
136
+
137
+
Each dimension scores 0 to 100. The framework does not replace human judgment. It sharpens it.
120
138
121
-
Each scores 0–100. The framework doesn't replace human judgment—it sharpens it.
139
+
Any framework registered in Fluently can define its own dimensions, evaluation criteria, best practices, and dimension combinations. The scorer and MCP server adapt automatically.
140
+
141
+
---
142
+
143
+
## Cross-Framework Benchmarking
144
+
145
+
Fluently's `compare_frameworks` MCP tool and `compare` CLI command let you run the same task through every registered framework simultaneously. This surfaces which framework's vocabulary and structure best fits your collaboration context, without having to commit to one upfront.
146
+
147
+
Agents can also call `evaluate_compliance(text, framework_id)` mid-session to check whether the current conversation adheres to the chosen framework, and receive actionable guidance on which dimensions need attention.
122
148
123
149
---
124
150
125
151
## Roadmap
126
152
127
-
**Embeddings-based similarity** — Replace keyword matching with semantic search so `find_similar_cycles()` surfaces truly relevant patterns regardless of terminology.
153
+
**Embeddings-based similarity** — Replace keyword matching with semantic search so `find_relevant_cycles()` surfaces truly relevant patterns regardless of terminology.
128
154
129
-
**VS Code extension** — Inline 4D scoring in your editor. @fluently in comments to get suggestions while you write, review, or debug.
155
+
**VS Code extension** — Inline scoring in your editor. Mention `@fluently` in comments to get suggestions while you write, review, or debug.
130
156
131
-
**Web playground** — Try 4D scoring live without installing. Experiment with Fluently 4D cycles. Generate your own.
157
+
**Web playground** — Try scoring live without installing. Experiment with cycles. Generate your own.
0 commit comments