AI Impact Research

Comprehensive research on AI platform usage from frontier labs (2025)

View the Project on GitHub vishalsachdev/ai-impact

← Back to Overview Data & Analysis →

Agentic Capabilities: Sources

Computer Use Benchmarks

OSWorld

  1. OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
    • Type: Academic benchmark
    • Key Data: Real VM task completion methodology
  2. Claude Opus 4.5 Announcement
    • URL: https://www.anthropic.com/news/claude-opus-4-5
    • Published: November 2025
    • Key Data: 61.4% OSWorld score
  3. Anthropic Transparency Hub
    • URL: https://www.anthropic.com/transparency/model-report
    • Key Data: Computer use capabilities
  4. VentureBeat: Claude Computer Use Analysis
    • URL: https://venturebeat.com/ai/anthropics-claude-opus-4-5-is-here-cheaper-ai-infinite-chats-and-coding
    • Published: November 2025

Agent Capability Research

Self-Refinement

  1. Anthropic: How AI Is Transforming Work at Anthropic
    • URL: https://www.anthropic.com/research/how-ai-is-transforming-work-at-anthropic
    • Key Data: Autonomous refinement in 4 iterations
  2. Anthropic Research Blog
    • Type: Technical documentation
    • Key Data: Agent architecture details

Multi-Step Tasks

  1. Claude 4 Performance Analysis (Data Studios)
    • URL: https://www.datastudios.org/post/claude-4-in-2025-performance-safety-benchmarks-ecosystem-news-and-real-world-impact-for-enterpr
    • Key Data: Agentic capabilities breakdown

Agentic AI Industry Coverage

Platform Announcements

  1. Microsoft Build 2025: Age of AI Agents
    • URL: https://blogs.microsoft.com/blog/2025/05/19/microsoft-build-2025-the-age-of-ai-agents-and-building-the-open-agentic-web/
    • Published: May 19, 2025
    • Key Data: Agentic AI vision
  2. OpenAI: Agentic AI Foundation
    • URL: https://openai.com/index/agentic-ai-foundation/
    • Key Data: Agent architecture approach
  3. Google Gemini 3 Agent Capabilities
    • URL: https://deepmind.google/models/gemini/
    • Key Data: Multi-modal agent features

Industry Analysis

  1. The Superblogs: Agentic AI 2025
    • URL: https://thesuperblogs.com/agentic-ai-2025-gemini-3-gpt-5-1-qwen-enterprise-workflows/
    • Key Data: Enterprise workflow automation
  2. Keepler: Gemini 3 Agent Analysis
    • URL: https://keepler.io/2025/11/27/google-gemini-3-a-new-paradigm-in-frontier-ai/
    • Published: November 27, 2025

Productivity with Agents

Enterprise Adoption

  1. Microsoft Copilot Usage Report
    • URL: https://microsoft.ai/news/its-about-time-the-copilot-usage-report-2025/
    • Published: December 10, 2025
    • Key Data: Agentic workflow adoption
  2. Menlo Ventures: State of GenAI in Enterprise
    • URL: https://menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise/
    • Key Data: Agent adoption patterns

Research Applications

  1. Computing at School: AI Agents in Education
    • URL: https://www.computingatschool.org.uk/forum-news-blogs/2025/april/empowering-educators-insights-from-anthropic-s-report-on-claude-s-role-in-higher-education/
    • Published: April 2025

Safety and Reliability

Agent Safety Research

  1. Anthropic: Model Safety Evaluations
    • URL: https://www.anthropic.com/transparency/model-report
    • Key Data: Agent safety testing
  2. AI Magazine: Microsoft AI Agent Analysis
    • URL: https://aimagazine.com/news/microsofts-report-inside-the-global-ai-adoption-divide
    • Key Data: Enterprise safety considerations

Academic Research

  1. Multi-Agent Systems for AI
    • Type: Academic literature
    • Key Topics: Agent architecture, coordination
  2. Human-AI Collaboration Research
    • Type: Academic literature
    • Key Topics: Oversight, intervention patterns
  3. Agent Reliability Studies
    • Type: Academic literature
    • Key Topics: Failure modes, recovery mechanisms

Source Categories

Category Count
Computer use benchmarks 4
Agent capability research 3
Industry coverage 5
Productivity studies 3
Safety research 2
Academic research 3
Total 20