cbrowser 18.37.0 → 18.37.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/docs/COGNITIVE-OPTIMAL-TRANSPORT-RESEARCH.md +63 -24
  2. package/package.json +1 -1
package/docs/COGNITIVE-OPTIMAL-TRANSPORT-RESEARCH.md CHANGED
@@ -1,8 +1,8 @@
1
1
  # Cognitive Optimal Transport — Research Synthesis
2
2
 
3
- **Date:** 2026-04-10
4
- **Issue:** #159
5
- **Sources:** 40+ papers across neuroscience, HCI, and mathematics
3
+ **Date:** 2026-04-11 (updated)
4
+ **Issues:** #159, #160, #161, #162, #163
5
+ **Sources:** 50+ papers across neuroscience, HCI, accessibility, and mathematics
6
6
  **Research agents:** 3 parallel, 80+ targeted queries
7
7
 
8
8
  ---
@@ -131,39 +131,56 @@ This is a single, principled, theoretically grounded number. No competitor has a
131
131
 
132
132
  ## Mathematical Foundations (Implementability)
133
133
 
134
- | Component | Complexity (d=25) | GPU | TypeScript Feasible |
135
- |---|---|---|---|
136
- | Sliced Wasserstein distance | O(L×n×d) ~500K ops | No | Yes, sub-ms |
137
- | Gaussian W₂ + geodesic | O(d³) ~15K ops | No | Yes, sub-ms |
138
- | Gaussian barycenter | O(K×d³) per iter | No | Yes, sub-ms |
139
- | DRO adversarial personas | O(N×d) per LP | No | Yes, sub-ms |
140
- | Sinkhorn discrete OT | O(n²/ε²) per iter | No | Yes, <10ms |
141
- | Normalizing flow (RealNVP) | O(K×d²) per sample | No | Yes, <1ms |
134
+ | Component | Complexity (d=26) | GPU | TypeScript Feasible | Status |
135
+ |---|---|---|---|---|
136
+ | Sliced Wasserstein distance | O(L×n×d) ~520K ops | No | Yes, sub-ms | **Shipped v18.26** |
137
+ | Gaussian W₂ + geodesic | O(d³) ~17.5K ops | No | Yes, sub-ms | **Shipped v18.27** |
138
+ | Gaussian barycenter | O(K×d³) per iter | No | Yes, sub-ms | **Shipped v18.26** |
139
+ | DRO adversarial personas | O(N×d) per LP | No | Yes, sub-ms | **Shipped v18.27** |
140
+ | Sinkhorn discrete OT | O(n²/ε²) per iter | No | Yes, <10ms | **Shipped v18.26** |
141
+ | Page Understanding (DOM) | O(n) elements | No | Yes, <500ms | **Shipped v18.35** |
142
+ | Site Model Learning | O(1) amortized writes | No | Yes, <50ms | **Shipped v18.35** |
143
+ | Goal Decomposition | O(k×s) subgoals×strategies | No | Yes, <100ms | **Shipped v18.35** |
144
+ | siteFamiliarity gating | O(1) lookup | No | Yes, <1ms | **Shipped v18.35** |
142
145
 
143
- **Key insight:** For d=25 traits, the Gaussian assumption gives closed-form solutions for everything. No GPU needed. The entire framework runs in pure TypeScript at sub-millisecond latency.
146
+ **Key insight:** For d=26 traits, the Gaussian assumption gives closed-form solutions for everything. No GPU needed. The entire framework runs in pure TypeScript at sub-millisecond latency.
144
147
 
145
148
  ---
146
149
 
147
150
  ## Novel Contributions (What Nobody Has Done)
148
151
 
149
- 1. **First persona system with mathematically grounded cognitive distance** — W₁(personaA, personaB)
150
- 2. **First accessibility tool measuring transport-cost information loss** — already shipped in v18.26.0
151
- 3. **First adversarial UX testing via distributionally robust optimization** — Wasserstein balls around known personas
152
- 4. **First persona interpolation using displacement geodesics** — McCann interpolation preserves trait coupling
153
- 5. **First unified multi-layer OT accessibility score** — sum of transport costs across 6 cognitive layers
154
- 6. **First attention-as-transport model for web UX** — persona saliency via filtered W₂
152
+ 1. **First persona system with mathematically grounded cognitive distance** — W₁(personaA, personaB) — *shipped v18.27*
153
+ 2. **First accessibility tool measuring transport-cost information loss** — *shipped v18.26*
154
+ 3. **First adversarial UX testing via distributionally robust optimization** — Wasserstein balls around known personas — *shipped v18.27*
155
+ 4. **First persona interpolation using displacement geodesics** — McCann interpolation preserves trait coupling — *shipped v18.27*
156
+ 5. **First unified multi-layer OT accessibility score** — sum of transport costs across 6 cognitive layers — *shipped v18.27*
157
+ 6. **First attention-as-transport model for web UX** — persona saliency via filtered W₂ — *shipped v18.28*
158
+ 7. **First persistent site knowledge graph for browser automation** — navigation maps, element reliability, goal paths that accumulate across sessions — *shipped v18.35*
159
+ 8. **First real-time page understanding via DOM analysis** — type classification, affordance mapping, form detection in <500ms — *shipped v18.35*
160
+ 9. **First siteFamiliarity-gated cognitive simulation** — persona site knowledge auto-scales to actual data; no simulating expertise on unknown sites — *shipped v18.35*
161
+ 10. **First research-backed cognitive disability personas** — autism spectrum (Yaneva 2018), intellectual disability (Karreman 2007), aphasia (W3C COGA), dyscalculia (UK Gov) — *shipped v18.35*
155
162
 
156
163
  **Publishable gap identified:** No existing work computes W(expected_experience, actual_experience) for UX abandonment prediction. The neuroscience (Dabney), behavioral signals (Yamauchi), and frustration data (Ceaparu) exist separately but nobody has unified them under optimal transport.
157
164
 
158
165
  ---
159
166
 
160
- ## Implementation Priority
167
+ ## Implementation Status
161
168
 
162
- 1. **Phase 1 (immediate):** Trait space as probability measure + cognitive distance. Pure math, no browser needed.
163
- 2. **Phase 2:** Adversarial persona generation via DRO. Solves the "what cognitive profile breaks this interface?" question.
164
- 3. **Phase 3:** Persona geodesic interpolation. Enables custom persona blending and sensitivity analysis.
165
- 4. **Phase 4:** Six-layer cognitive transport scoring. Requires integrating with page analysis pipeline.
166
- 5. **Phase 5:** Attention-as-transport saliency modeling. Most complex, most differentiated.
169
+ | Phase | Feature | Status | Version |
170
+ |-------|---------|--------|---------|
171
+ | 1 | Trait space as probability measure + cognitive distance | **Shipped** | v18.27 |
172
+ | 2 | Adversarial persona generation via DRO | **Shipped** | v18.27 |
173
+ | 3 | Persona geodesic interpolation (McCann) | **Shipped** | v18.27 |
174
+ | 4 | Six-layer cognitive transport scoring | **Shipped** | v18.27 |
175
+ | 5 | Attention-as-transport saliency modeling | **Shipped** | v18.28 |
176
+ | 6 | Real-time page understanding (DOM analysis) | **Shipped** | v18.35 |
177
+ | 7 | Persistent site knowledge graph | **Shipped** | v18.35 |
178
+ | 8 | Cross-session browser profiles (AES-256-GCM) | **Shipped** | v18.35 |
179
+ | 9 | Autonomous goal decomposition | **Shipped** | v18.35 |
180
+ | 10 | siteFamiliarity trait + knowledge gating | **Shipped** | v18.35 |
181
+ | 11 | Research-backed disability personas (4 new) | **Shipped** | v18.35 |
182
+
183
+ All phases complete. 104 MCP tools, 26 cognitive traits, 21 personas (11 accessibility).
167
184
 
168
185
  ---
169
186
 
@@ -189,6 +206,28 @@ This is a single, principled, theoretically grounded number. No competitor has a
189
206
  - Perry, Zorzi, Ziegler (2019) "Personalized Dyslexia Models" — Psych. Science
190
207
  - Yamauchi & Xiao (2018) "Cursor Emotion Reading" — Cognitive Science
191
208
 
209
+ ### Disability / Accessibility HCI (v18.35)
210
+ - Yaneva et al. (2018) "Web Users with Autism: Eye Tracking Evidence" — Behaviour & IT
211
+ - Yaneva et al. (2020) "Keep It Simple: Eye-Tracking for ASD" — W4A
212
+ - Raymaker & Nicolaidis (2019) "AASPIRE Web Accessibility Guidelines" — PMC6485264
213
+ - Britto & Pizzolato (2016) "Interaction Design for ASD" — Semantic Scholar
214
+ - Karreman et al. (2007) "Accessible Website Content for ID" — J. Applied Research in ID
215
+ - Rocha et al. (2015) "Optimising Web Designs for Learning Disabilities" — PMC4467236
216
+ - W3C COGA (2024) "Cognitive Accessibility User Research" — W3C
217
+ - W3C COGA (2024) "Aphasia Research Module" — W3C
218
+ - W3C COGA (2024) "Dyscalculia Research Module" — W3C
219
+ - UK Gov Design System (2022) "Designing for Dyscalculia and Low Numeracy"
220
+ - Butterworth (2005) "Developmental Dyscalculia" — Science
221
+ - Cockburn et al. (2007) "Familiar Interfaces" — HCI
222
+ - Tauscher & Greenberg (1997) "How People Revisit Web Pages" — CHI
223
+
224
+ ### Browser Agent Architecture (v18.35)
225
+ - AgentQ / MultiOn (2024) "MCTS-guided Browser Agent" — arXiv 2408.07199
226
+ - rtrvr.ai (2025) "DOM Intelligence Architecture" — 81.39% WebBench
227
+ - SeeAct (2024) "GPT-4V Grounding on Web Pages" — ICML
228
+ - Stagehand v3 (2025) "Action Caching with DOM Hash Validation" — Browserbase
229
+ - WebArena (2024) "Contextual Experience Replay" — 51% improvement with prior knowledge
230
+
192
231
  ### Mathematics
193
232
  - Agueh & Carlier (2011) "Barycenters in Wasserstein Space" — SIAM
194
233
  - Altschuler & Boix-Adsera (2022) "Barycenters are NP-Hard" — SIAM
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "cbrowser",
3
- "version": "18.37.0",
3
+ "version": "18.37.1",
4
4
  "type": "module",
5
5
  "description": "Cognitive browser automation that thinks like your users—and helps AI agents navigate too. Simulate real user cognition with abandonment detection, constitutional safety, chaos engineering, and UX friction discovery. Sites that pass CBrowser's cognitive tests are easier for both humans and AI agents to navigate.",
6
6
  "main": "dist/index.js",