AI Agents · May 26, 2025 · 2 min read

Claude 4 and the Next Generation of Coding Agents

Anthropic introduced the Claude 4 family — Opus 4 and Sonnet 4 — with a clear emphasis on sustained, agentic work: models that can stay on task across long, multi-step coding and reasoning problems.

Agents that go the distance

Earlier agents often lost the thread on long tasks. Improvements in sustained reasoning and tool use mean agents can now handle longer workflows — refactoring across a project, running tests, and iterating — with fewer derailments.

Engineering discipline scales the benefit

The more capable the agent, the more leverage your guardrails provide. Clear task scoping, strong tests, code review, and observability turn a capable model into a dependable teammate.

Our take: this is the most exciting moment yet for AI-assisted engineering. We help teams put these agents to work safely, with the evaluation and process to trust the results.