Tech • IA • Crypto
Anthropic Cloud Agents, Claude Memory, Legora $5B Surge
Briefing
Anthropic unveils cloud agent infrastructure
Anthropic introduced cloud-managed agent infrastructure designed to support autonomous, long-running AI workflows at scale. The system shifts focus from model capability to execution layers like orchestration, memory, and security. It enables agents to complete multi-step objectives independently, returning results only when tasks are finished. This marks a transition from prompt-response interactions to outcome-driven automation.
API-first agents redefine development stack
The platform exposes agent capabilities through API endpoints built around primitives like agents, sessions, environments, and events. Developers can assemble production systems without building orchestration, retries, or state handling from scratch. Fine-grained tool permissions allow controlled access to resources such as bash, web search, or private integrations. This modular architecture positions agents as composable infrastructure rather than monolithic applications.
Claude memory enables persistent learning
Anthropic introduced a file-based memory system that allows agents to retain knowledge across tasks. Agents can reuse prior strategies, learn from failures, and improve performance over time instead of restarting from scratch. The design leverages familiar file operations, aligning with model strengths in reading and editing structured data. Shared memory also enables collaboration between multiple agents within the same environment.
Agents demand enterprise-grade security controls
Autonomous agents require secure access to systems like GitHub repositories, internal tools, and communication platforms. This introduces the need for identity frameworks, permission layers, and auditability. Developers can configure which tools run automatically and which require human approval. The approach balances autonomy with safeguards, especially in sensitive enterprise environments.
Legora hits $5B using agent playbook
Legora, an AI legal workspace, surpassed 1,000 customers and reached a valuation above $5 billion with $100 million ARR. The company applies patterns from coding agents—such as planning, sub-agents, and human-in-the-loop review—to legal workflows. Legal and software domains share structured, text-heavy processes, enabling transfer of these designs. This demonstrates how vertical AI products can scale rapidly using agent frameworks.
Omni scales analytics with Claude models
Omni integrated Claude models to power natural-language data querying and accelerate engineering output. A team of about 25 engineers significantly increased development velocity after adoption. Its semantic layer standardizes business definitions and enforces permissions across datasets. The result is more reliable query generation and improved accessibility for non-technical users.
Man Group deploys AI trading signals
Man Group, managing over $200 billion, is using AI to generate end-to-end systematic trading signals. Models scan global markets to rank securities, with strategies validated عبر 15+ years of backtesting. Metrics like Sharpe ratio and drawdowns remain central to evaluation. The firm emphasizes governance and reliability, as errors directly impact real-world financial outcomes.
Coding shifts bottleneck to verification
AI-assisted coding has moved the constraint from writing code to verifying correctness and reliability. Teams now prioritize testing, validation, and automated checks as output scales rapidly. Legacy processes like heavy upfront planning are becoming less effective in this new environment. Human judgment remains critical, especially in code review and high-stakes decision points.
Videos covered
How to get to production faster with Claude Managed Agents
- •Shift from model limits to infrastructure bottlenecks
- •Rise of outcome-oriented agents
- •Need for secure access and identity
Build a production-ready agent with Claude Managed Agents
- •API-first agent infrastructure
- •Core building blocks: agents, environments, sessions, events
- •Fine-grained tool access and permissions
What legal agents inherit from coding agents: Lessons from Legora
- •Rapid growth of Legora
- •Coding agents as a blueprint
- •Shared structure of knowledge work
Building the best agentic analytics harness: Powered by Claude, built with Claude Code
- •Rapid engineering acceleration
- •Natural language to data queries
- •Semantic layer as core infrastructure
Trading signals that trade themselves
- •AI enters high-stakes asset management
- •Systematic trading as a prime AI use case
- •Backtesting remains central to validation
Picking the right model
- •Model selection is more complex than it թվում
- •Three core decision pillars: quality, latency, cost
- •Public benchmarks offer limited real-world value
Memory and dreaming for self learning agents
- •Agents Are Expanding in Capability and Duration
- •Memory Enables Continuous Learning
- •File-Based Design Aligns With Model Strengths
The prompting playbook
- •Prompting remains a core engineering skill
- •Evaluation frameworks are essential
- •Common failure in legacy prompts
The Problem Solvers: Kay Zhu at Genspark
- •AI reshaping career paths
- •Expansion of AI-powered workplaces
- •Startups relying on partnerships
How AirOps chases friction to build AI products with Claude
- •Shift from workflow builders to agents
- •Targeting marketers in AI search
- •Launch of AirOps Next and Quill agent
Running an AI-native engineering org
- •Bottlenecks move from coding to verification
- •Legacy processes “quietly stop working”
- •AI transforms development practices like testing and refactoring
Making agentic workflows trustworthy and verifiable with a custom DSL
- •Mechanism over output
- •Trade-off between speed and rigor
- •Three requirements for trustworthy agents