PILLAR_L01 UPDATED QUARTERLY
The agentic landscape, mapped.
Where we came from. Who's leading. What's next. A plain-English orientation map for the agentic era.
TIMELINE // 2017→PRESENT
Where we came from
2017
Transformer arch (Vaswani et al.)
foundation
2020
GPT-3, large-context language models
foundation
2022
ChatGPT — chatbot era kicks off
consumer
2023
Tool calling, function calling, ReAct loops
agentic
2023
GPTs, Custom GPTs, Anthropic Claude
platform
2024
MCP introduced (Anthropic, Nov)
protocol
2024
Computer Use API (Anthropic)
agentic
2025
A2A protocol (Google), Apps SDK (OpenAI)
protocol
2025
Claude Skills, Operator, Agent 365
platform
2026
Multi-agent systems mainstream
current
▸ VENDOR MAP // WHO'S BUILDING WHAT
The big four agentic platforms.
VENDOR PROFILE
Microsoft
FOCUS Copilot Studio · Foundry · Connectors · Agent 365
STRATEGY M365 / Azure-native enterprise agents
VENDOR PROFILE
Anthropic
FOCUS Claude · MCP · Skills · Computer Use
STRATEGY Protocol-led, capability-led, safety-led
VENDOR PROFILE
OpenAI
VENDOR PROFILE Google
FOCUS Gemini · Vertex Agents · A2A · Notebook LM
STRATEGY Long-context, image/audio/video integration, deep workspace ties
STATE OF · 2026 Q2
What's actually shipping right now
▲ Trending up
- MCP adoption — 348 servers and growing
- Computer Use APIs (Anthropic, OpenAI Operator)
- Agent SDKs from every major vendor
- Self-hosted agents (Ollama + LM Studio + Open WebUI)
- Multi-agent orchestration (LangGraph, CrewAI)
▼ Cooling off
- "Single agent does everything" pitch
- AutoGPT-style infinite loop demos
- Prompt-engineering-as-a-job
- Older agent frameworks consolidating — AutoGen in maintenance mode, Microsoft Agent Framework as the successor path
◎ To watch
- Protocol war: MCP vs A2A vs Apps SDK
- Agent-to-agent payments + commerce
- On-device agents (Apple Intelligence, Phi)
- Standards consolidation
- Specialised agents replacing horizontal SaaS