Research Observatory
Curated research on AI architecture, cybersecurity, financial services, and enterprise technology — organized for both human review and AI-assisted exploration. 20 source files, each credibility-rated and sourced, covering landscape data, architecture trends, security threat surfaces, regulatory compliance, financial services deployment, consolidation economics, and enterprise deep dives.
01 Landscape & Market Data
Landscape & Tool Counts
Comprehensive mapping of the AI tool ecosystem from startup counts to market sizing across commercial and open-source segments.
- 12K–41K commercial AI tools across categories
- $22–92B GenAI market sizing with three-tier pyramid
- Startup counts, category breakdowns, fragmentation analysis
Usage Patterns & Adoption
Cross-referenced adoption data from the most rigorous surveys available, segmented by user type and usage frequency.
- Gallup: 46% of US workers using AI at work
- Pew: 52% weekly AI interaction, McKinsey: 88% of orgs
- Shadow AI prevalence, developer stats, tool sprawl dynamics
Platforms: Consumer & Developer
Registry of 14 consumer and developer AI platforms with financials, user counts, and competitive positioning.
- ChatGPT 900M WAU, Claude $14B ARR, Cursor $2B ARR
- Gemini 750M MAU, Grok 78M MAU (17.8% US share)
- Perplexity $200M ARR, DeepSeek, Meta AI coverage
Platforms: Enterprise
Deep analysis of 18 enterprise AI platforms covering Microsoft, Salesforce, ServiceNow, Oracle, SAP, and infrastructure providers.
- Microsoft Copilot: 15M paid users, 3.3% conversion rate
- Salesforce Agentforce: 29K deals, $800M ARR
- Oracle 1000+ agents, ServiceNow 80B workflows/yr
Global Non-US Landscape
Analysis of non-US AI ecosystems including China's domestic players, EU regulation, and emerging market dynamics.
- DeepSeek R1: $0.55/M input, $294K training cost
- Qwen 600M+ downloads, Baidu 200M MAU, ByteDance Doubao 155M WAU
- EU AI Act impact, Mistral €11.7B valuation, Middle East HUMAIN $1.2B
02 Architecture & Technical
Architecture Trends
Mapping the evolution from simple chat interfaces to orchestration platforms, agentic frameworks, and OS-level integration.
- Orchestrators: Perplexity Computer, ChatGPT Agent, Claude/MCP, Gemini 3
- Reasoning models: o3, Opus 4.6, Gemini 3
- Agentic frameworks: LangChain 1.0, CrewAI, AutoGen, OpenAI Agents SDK
Technical Foundations
Seven real technical shifts underlying the current wave, with hype cycle positions and analyst predictions.
- Function calling, reasoning models, context scaling (1M+ tokens)
- Distillation, inference optimization (KV cache, speculative decoding)
- Analyst predictions from Gartner, Forrester, IDC, Deloitte
03 Cybersecurity & AI Security
AI-specific threat surfaces, enterprise security transformation, and regulatory compliance frameworks. Emphasis on demonstrated attacks over theoretical risks, with vendor claims scrutinized against independent evidence.
AI-Specific Threat Surface
Deep analysis of attack vectors unique to LLMs and agentic systems, with demonstrated vs. theoretical risk grading for each category.
- EchoLeak CVE-2025-32711 (CVSS 9.3): first zero-click production prompt injection exploit
- MCP agent escalation CVE-2025-6514 (CVSS 9.6), Cisco OpenClaw ClawHavoc findings
- Model poisoning: Anthropic 250-document backdoor proof, HuggingFace namespace hijack
AI & Enterprise Security Ops
How AI is transforming SOC operations, offensive security, shadow AI governance, and financial services fraud — with vendor claims tested against independent data.
- $244B security spend 2026, 4.8M workforce gap (ISC2), 94% SOC noise rates
- Deepfake wire fraud: $410M losses H1 2025, $193M Hong Kong ring
- Shadow AI: 80% workers using unapproved tools, $670K breach cost premium (IBM)
Regulatory & Compliance
Framework-level detail on AI security regulation: NIST, OWASP, EU AI Act, SEC/FINRA, MITRE ATLAS, and emerging agentic AI standards.
- NIST AI RMF + IR 8596 Cyber AI Profile + CAISI Agent Standards Initiative
- OWASP LLM Top 10 2025 + Agentic Top 10 (ASI01–10), full diff tables
- SEC CETU, FINRA RN 24-09, NYDFS 23 NYCRR 500, EU AI Act Article 15
File: research/sources/cybersecurity-regulatory-compliance.md
Cybersecurity Implications
Cross-domain synthesis connecting threat data, adoption patterns, architecture trends, and regulatory frameworks into practitioner-focused cybersecurity implications.
- Convergence thesis: MCP universality + 57% agent deployment + 98% shadow AI = systemic attack surface
- OWASP Agentic Top 10 mapped to SOC operations with detection difficulty ratings
- Regulatory trilemma: SR 11-7 gap + EU AI Act enforcement Aug 2026 + NIST standards pending
04 Financial Services
AI Deployment Reality
Production AI deployments across banking, capital markets, insurance, wealth management, and RegTech — with verified ROI data and infrastructure reality checks.
- JPMorgan LLM Suite (230K+ users), Morgan Stanley (98% FA adoption), Capital One (100% cloud-native)
- 42% of FS AI proofs-of-concept abandoned; BCG median ROI: 10%; Goldman: macro impact “basically zero”
- SR 11-7 vs nondeterministic LLMs, CFPB explainability enforcement, BCBS 239 still unresolved at most banks
05 Consolidation & Economics
Consolidation & Enterprise
The thin wrapper death spiral, five defensible moats, and the economic dynamics of enterprise AI adoption.
- Five moats: compliance, data flywheel, workflow integration, vertical specificity, system-of-record
- 966 startup shutdowns in 2024, $37B enterprise GenAI spend
- 95% of pilots fail P&L, shadow AI in 98% of orgs
Sci-Fi & Power Users
Cultural predictions versus reality, developer adoption patterns, and the behavior loops driving platform stickiness.
- Sci-fi predictions: HAL, JARVIS, Her, Star Trek, Cortana vs reality
- Developer adoption: 84% using AI, Cursor $2B ARR
- Adoption arc: explore → primary → consolidate → infrastructure
06 OpenClaw Deep Dive
Five sources on a single subject — justified by its role as architectural counter-thesis to the platform-centric model.
OpenClaw Architecture
Analysis of the agentic harness architecture that reached 250K GitHub stars in 60 days.
- Agentic harness (not a model), 250K GitHub stars in 60 days
- Steinberger → OpenAI transition, Tencent/WeChat integration
- Cisco security findings, naming drama, China ban implications
OpenClaw Practical Setup
Hands-on guide covering installation, cost models, and real-world use cases for the agentic harness.
- 10-step setup guide from beginner to production
- OAuth cost model: $20/mo vs API direct pricing
- Content pipeline (7-step), conversational CRM use cases
Jensen Huang (All-In)
Jensen Huang's vision for $50T Physical AI, 10,000x compute scaling, and OpenClaw as the new OS.
- $50T Physical AI thesis, Groq acquisition rationale
- OpenClaw as “new OS” (memory + skills + resources + scheduling)
- $250K token spend per engineer, autoresearch (30min = 7yr PhD)
Steinberger (OpenAI)
Interview with the creator covering build workflow, model trust hierarchy, and agent resourcefulness.
- 90K contributions/yr across 120+ projects
- Model trust hierarchy: Claude → Codex preference
- “Vibe coding is a skill” — agent resourcefulness patterns
OpenClaw (Udacity)
Balanced explainer positioning the project as infrastructure rather than AI, with skill ecosystem analysis.
- “Plumbing not AI” — architectural positioning
- 15K Clawhub skills in the marketplace
- Real bottleneck: delegation skill, not technology
07 Enterprise Deep Dives
Oracle Cloud ERP AI
Full-stack deep dive covering Oracle's AI strategy from OCI through Fusion to Agent Studio, with pricing and competitive analysis.
- 50+ prebuilt agents, Agent Studio (Supervisory → Functional → Utility)
- Pricing: $50/user or $2.50/employee, Q3 FY26 $8.9B cloud (84% OCI growth)
- Competitive vs SAP/Microsoft/Workday/ServiceNow, $553B RPO ($300B+ OpenAI)
Source Credibility Framework
| Tier | Definition | Examples | Usage Rule |
|---|---|---|---|
| Gold | Independent, rigorous methodology, large sample | Gallup (n=20K+), Pew, Stanford HAI, Stack Overflow (n=65K), NIST, OWASP, MITRE, SEC, FINRA, CISA | Cite without caveats |
| Silver | Credible but potential bias | McKinsey (executive bias), Gartner, Forrester, IBM, CrowdStrike, Palo Alto Networks, EY, Menlo Ventures | Cite with stated caveats |
| Bronze | Useful signal, interpret with caution | Slack/Salesforce surveys, security vendors, market research firms | Order-of-magnitude only |
| Contextual | Primary source material | Platform announcements, CEO interviews, Lex Fridman transcripts | Qualitative claims + direct quotes |
Research File Index
Direct links to all research source files. Each file is plain-text markdown accessible at the URL path shown.
- landscape.md — research/sources/landscape.md
- usage-patterns.md — research/sources/usage-patterns.md
- platforms-consumer-developer.md — research/sources/platforms-consumer-developer.md
- platforms-enterprise.md — research/sources/platforms-enterprise.md
- global-non-us-landscape.md — research/sources/global-non-us-landscape.md
- architecture-trends.md — research/sources/architecture-trends.md
- technical-foundations.md — research/sources/technical-foundations.md
- cybersecurity-ai-threats.md — research/sources/cybersecurity-ai-threats.md
- cybersecurity-enterprise-ai.md — research/sources/cybersecurity-enterprise-ai.md
- cybersecurity-regulatory-compliance.md — research/sources/cybersecurity-regulatory-compliance.md
- ai-financial-services.md — research/sources/ai-financial-services.md
- consolidation-enterprise.md — research/sources/consolidation-enterprise.md
- scifi-powerusers.md — research/sources/scifi-powerusers.md
- openclaw.md — research/sources/openclaw.md
- openclaw-practical-setup.md — research/sources/openclaw-practical-setup.md
- jensen-huang-allin-mar2026.md — research/sources/jensen-huang-allin-mar2026.md
- steinberger-openai-interview.md — research/sources/steinberger-openai-interview.md
- openclaw-udacity-explainer.md — research/sources/openclaw-udacity-explainer.md
- oracle-cloud-erp-ai.md — research/sources/oracle-cloud-erp-ai.md
- cybersecurity-implications.md — research/sources/cybersecurity-implications.md