An Anthropic Anthology

The Claude Era A Guide to Enlightened Intelligence

"In my younger and more vulnerable years, my father gave me some advice that I've been turning over in my mind ever since: the right tools change everything."

Volume I · MMXXVI
Chapter I

The Foundation

◆
⬥ I ⬥
Claude.ai
The Grand Salon
  • The Web Application
  • Desktop & Mobile Apps
  • Projects & Artifacts
  • The Support Center
Enter the Salon →
⬥ II ⬥
Claude Code
The Architect's Workshop
  • Official Documentation
  • The Agent SDK
  • MCP Configuration
  • Community Handbook
Begin the Craft →

So we beat on, boats against the current, borne back ceaselessly into the future of intelligent machines.

— A Variation on Fitzgerald
Chapter II

The Architecture

◆
⬥ III ⬥
Claude API
The Conduit of Power
  • Introduction & Quickstart
  • Models Overview
  • API Reference
  • The Anthropic Console
Wield the Power →
⬥ IV ⬥
Certifications
The Order of Excellence
  • CCA Foundations — Live Now
  • 5 Domains · 6 Scenarios
  • Passing Score: 720 / 1000
  • Free for Partner Employees
View Exam Guide →
The Trinity of Intelligence
Current Models · Anno Domini MMXXVI
✥ New · May 2026
The Sovereign
Opus
version 4.8
For matters of grand consequence: deep reasoning, complex creation, and intellectual excellence. Now with 1M-token native context.
1,000,000 tokens
The Diplomat
Sonnet
version 4.6
The harmonious balance of speed and intelligence, perfectly suited for most endeavors.
200,000 tokens
The Messenger
Haiku
version 4.5
Swift as a champagne toast, perfect for instantaneous responses and simple tasks.
200,000 tokens
The Selector’s Field Guide

The Model Matrix

◆
◆ The Art of Model Selection · A Consulting Field Guide
Which Model for Which Task?

The wise consultant selects the finest instrument for the occasion — speed for the drawing room, depth for the study, balance for the salon. This matrix is your field guide to the art of selection. Match the task to the model; let the conversation flourish.

Legend:
★  Ideal choice
◇  Capable
—  Not recommended
Use Case Haiku
Swift · Light
 Sonnet
Balanced · Reliable
 Opus
Deep · Supreme
◆ Speed & Efficiency
Real-time / interactive chat ★◇—
High-volume classification & tagging ★◇—
Content moderation at scale ★◇—
Cost-sensitive production pipelines ★◇—
◆ Balanced Workloads
Standard content & copywriting ◇★◇
Code generation & review —★◇
Data analysis & summarization ◇★◇
Agentic task orchestration —★◇
Customer support & Q&A ◇★—
RAG document retrieval & synthesis —★◇
◆ Deep Intelligence
Complex multi-step reasoning chains —◇★
Strategic consulting & analysis reports —◇★
Long-form creative & professional writing —◇★
Enterprise architecture design ——★
Million-token document analysis ——★
Critical decision support & scenario planning —◇★

◆ As a rule: default to Sonnet. Reach for Haiku when cost & latency govern; reach for Opus when the question keeps you up at night.

✥ Breaking Dispatch · May 2026 ✥
Anthropic Enterprise Certification · Second Cohort · May 20, 2026
Claude Enterprise Certification Program Opens Second Cohort

Anthropic’s Claude Enterprise Certification program opened its second cohort on May 20, 2026, with curriculum focused on Claude 4 system prompting, tool use, and enterprise deployment patterns. The program builds on the first cohort’s foundation and runs alongside the broader Claude Partner Network, which commits $100M in 2026 to training, sales enablement, and Applied AI engineering support for firms deploying Claude at scale. Partners completing the program gain priority access to the Claude Certified Architect, Foundations exam and upcoming seller and developer certifications.

Cohort 2 · Opened May 20 Claude 4 System Prompting Tool Use & Deployment Patterns CCA Foundations Cert Live Partner Network · $100M 2026

Source: Anthropic · Claude Partner Network · March 12, 2026 · Verify second-cohort expansion at claude.com/partners

✥ Exam Chronicle · Available Now ✥
Claude Certified Architect · Foundations · Issued by Anthropic
The Architect’s Examination: Claude Certified Architect — Foundations

Anthropic’s first official technical certification tests your mastery of production-grade Claude architecture across 5 domains and 6 real-world scenarios. The exam is scenario-based, multiple-choice, with a passing score of 720 out of 1,000. Four of six scenarios are randomly selected per sitting. Currently free for the first 5,000 partner company employees.

◆ The Five Domains
Domain I · ~25%
Agentic Architecture & Orchestration
Agentic loops, multi-agent orchestration, hooks, session management & task decomposition.
Domain II · ~20%
Tool Design & MCP Integration
Tool description best practices, structured error responses, MCP configuration & built-in tools.
Domain III · ~20%
Claude Code Config & Workflows
CLAUDE.md hierarchy, custom commands, plan mode, CI/CD integration & batch processing.
Domain IV · ~20%
Prompt Engineering & Structured Output
Few-shot prompting, tool_use for structured output, JSON schema design & validation-retry loops.
Domain V · ~15%
Context Management & Reliability
Progressive summarization, context positioning, escalation patterns & error propagation.
◆ The Six Scenarios (4 Randomly Selected Per Sitting)
ICustomer Support Resolution Agent
IICode Generation with Claude Code
IIIMulti-Agent Research System
IVDeveloper Productivity with Claude
VClaude Code for CI/CD
VIStructured Data Extraction
Multiple Choice · Scenario-Based Passing Score: 720 / 1000 5 Domains · 6 Scenarios Free · First 5,000 Partners Issued by Anthropic
◆   Study the Exam Guide Register for the Exam →

Source: claudecertifications.com · Exam Guide · Registration via Anthropic Skilljar Portal

✥ New · March 2026 ✥
The Anthropic Institute · Launched March 11, 2026 · Led by Jack Clark
Introducing The Anthropic Institute

Anthropic has established a new division dedicated to confronting the societal challenges posed by powerful AI. Led by co-founder Jack Clark in the new role of Head of Public Benefit, the Institute operates as an independent research arm — committed to reporting candidly on what frontier AI builders observe in the wild, even when those observations are uncomfortable. The Institute’s founding thesis: the organizations building the most powerful AI are also best positioned to sound the alarm about what it does to society.

◆ Three Research Teams
Team I
Frontier Red Team
Stress-testing frontier models before and after deployment. Finding failure modes society will encounter before society does.
Team II
Societal Impacts
Studying how AI reshapes democratic institutions, rule of law, and social trust — led by founding hire Matt Botvinick (Yale Law).
Team III
Economic Research
Measuring AI’s real effects on labor markets and wages — led by Anton Korinek (UVA) and Zoë Hitzig (formerly OpenAI).
Frontier Red Team Societal Impacts Economic Research Head: Jack Clark Est. March 11, 2026
◆   Visit the Institute

Source: anthropic.com · The Anthropic Institute Announcement · March 11, 2026

✥ Breaking · May 29, 2026 ✥
Model Development · Claude Opus 4.8 · Released May 29, 2026
Claude Opus 4.8 Arrives — 1M Native Context, 12% RAG Uplift

Anthropic has released Claude Opus 4.8, the most capable model in the Claude 4 family, featuring a 1 000 000-token native production context window — the first in the industry to offer this at general availability. Internal benchmarks show a 12% improvement over Claude 3.5 Sonnet on enterprise RAG tasks as measured by HELMET, and significant gains in structured-output reliability across financial, legal, and scientific domains. The model ships with enhanced constitutional AI guardrails and expanded tool-use capabilities optimized for multi-agent orchestration workflows. Pricing is tiered: standard context (up to 200K) at the existing Opus rate; the extended million-token window carries a premium multiplier for tokens beyond 200K.

Context Window
1,000,000 Tokens
Native GA production context — full codebases, legal corpus, research archives in a single pass.
RAG Performance
+12% on HELMET
Statistically significant improvement over Claude 3.5 Sonnet on enterprise retrieval benchmarks.
Model String
claude-opus-4-8
Available now in Anthropic API, Claude.ai, and all enterprise deployments via the Console.
PM / AI Angle
Scope Immediately
Clients with large-document workflows — legal, compliance, codebase review — should pilot Opus 4.8 this sprint.
claude-opus-4-8 1M Token Context +12% RAG Benchmark GA · May 29, 2026 Constitutional AI v3
◆   Read the Announcement Access in Console →

Source: Anthropic News · May 29, 2026 · Model string confirmed via docs.claude.com

Chapter III

The Mastery

◆
⬥ V ⬥
Prompt Engineering
The Art of Conversation
  • The Official Guide
  • Best Practices Manual
  • Research & Insights
  • Interactive Tutorial
Refine Your Craft →
⬥ VI ⬥
The Console
The Inner Sanctum
  • Test Your Prompts
  • Manage API Keys
  • Monitor Usage
  • Track Performance
Enter the Sanctum →
Chapter IV

The Academy

◆
✥ VII ✥
Anthropic Academy
The Learning Salon
  • Build with Claude
  • Claude for Work
  • Claude for Personal
  • AI Fluency Newsletter
Enter the Academy →
✥ VIII ✥
Featured Courses
The Curriculum
  • Claude 101
  • Claude Code in Action
  • MCP & Agentic Workflows
  • Earn Certificates
Begin the Course →
The Moving Picture Salon

Anthropic on Film

▸
▶ The Silver Screen of Intelligence
Anthropic on YouTube

From interpretability breakthroughs to live product demonstrations, Anthropic's YouTube channel is the definitive moving-picture record of the Claude era — research talks, safety deep-dives, enterprise showcases, and the occasional glimpse behind the curtain of the most consequential laboratory of our age.

▶   Enter the Salon
youtube.com/@anthropic-ai  ·  Subscribe for weekly transmissions
📊
Research Talks
The Symposium Floor
  • Interpretability Findings
  • Alignment Theory
  • Constitutional AI
  • Scaling & Safety
⚙
Claude in Action
The Demonstration Hall
  • Product Launches
  • Feature Walkthroughs
  • Agentic Workflows
  • Tool Use Demos
🏛
Enterprise
The Board Room
  • Customer Stories
  • Partner Showcases
  • Deployment Patterns
  • ROI Case Studies
📚
Policy & Ethics
The Great Debate
  • AI Governance
  • Responsible Scaling
  • Congressional Testimony
  • Global AI Policy
The Wireless Dispatches

The Social Gazette

◆
◆ Official Voices · Follow the Dispatches in Real Time
Anthropic & Claude on X

The salon extends to the telegraph wire. Follow these official handles for model releases, research dispatches, product announcements, and the occasional glimpse into the minds shaping the intelligence of our age.

@AnthropicAI
Anthropic
The official company voice — model releases, safety research, policy announcements & company news.
@ClaudeAI
Claude
The Claude product account — feature launches, tips, and dispatches from the model itself.
@claudecode
Claude Code
The agentic coding CLI — new releases, SDK updates, MCP server announcements & developer tips.
@DarioAmodei
Dario Amodei
CEO of Anthropic — long-form thinking on AI safety, policy, and the future of beneficial intelligence.
@AmandaAskell
Amanda Askell
Alignment & character research lead — the mind behind Claude’s values, tone, and personality.
@jackclarkSF
Jack Clark
Head of Public Benefit & Anthropic Institute — AI policy, societal impact & the frontier red team.
◆ @AnthropicAI — Official Dispatches
◆ @claudecode — Developer Channel
All handles are official Anthropic & team accounts · Follow for real-time intelligence · X / Twitter

And so with the sunshine and the great bursts of leaves growing on the trees, the future of artificial intelligence has that familiar way of beginning again with each conversation.

— The Eternal Promise of Claude