Observed Signal · Jul 28, 2026 · Technical Release · Source: The Pragmatic Engineer · Impact: 3/5 · Sentiment: Positive

Anthropic's AI Tools Reshape Software Engineering

Executive Signal Summary

The Pragmatic Engineer visited Anthropic’s San Francisco lab and interviewed four engineers to describe how improved AI tooling is changing software development. Key examples: the Claude Platform team built and launched Claude Managed Agents after a six-month project and re-architected its platform layer (migrating from Python to Rust); Bun creator Jarred Sumner completed a Zig→Rust rewrite in 11 days using 64 parallel AI agents and about $165,000 in tokens, with substantial verification and testing work after the initial implementation. The article documents shifts in team practices — more agent-driven prototyping, heavy use of automated code review and security scanners, increased fluidity between teams, and continued reliance on planning and PRDs for complex projects.

Polaris7 AgentPolaris7 Strategic Assessment
High Confidence

Demonstrates concrete, production-scale uses of AI agents and AI-driven engineering workflows (fast large codebase migration, managed agent infra, automated review/scanning) that can materially change developer productivity, CI/CD practices, and software architecture decisions.

SIGNAL RADAR

Track Anthropic Signals & Market Shifts in Real-Time

Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your Polaris7 workspace to monitor these entities.

Start Monitoring
Verified primary sourcesAutomated briefingsInstant watchlist setup

Key Takeaways & Evidence Grounding

  • Anthropic built and launched Claude Managed Agents; the project took around six months and launched in April 2026.
  • Anthropic’s Claude Platform is migrating its platform layer from Python to Rust for performance and scalability reasons.
  • Jarred Sumner rewrote Bun from Zig to Rust and merged the change in 11 days using 64 parallel AI agents and about $165,000 in tokens; the rewrite shipped to production and powers Claude Code.
  • Anthropic teams run many AI agents in parallel (3–10 commonly), use automated AI code review and security scanning (e.g., multiple runs of Claude Security Scanner), and devote most AI usage to verification and testing rather than implementation.
  • Anthropic consulted internal sandboxing teams and re-architected Managed Agents to decouple 'brain', 'hands' (sandboxes/tools), and 'session', and implemented a vault + proxy design so credentials are injected only at the egress boundary.

Connected Companies & Entities

“Thanks to Anthropic for showing me inside their lab in San Francisco.”

“David Hershey working with customers like Cursor, Cognition, and Perplexity”

“David Hershey working with customers like Cursor, Cognition, and Perplexity”

“David Hershey working with customers like Cursor, Cognition, and Perplexity”

“The project took about six months, by no means a rapid process adding support for running agents on AWS, GCP and Azure.”

“The project took about six months, by no means a rapid process adding support for running agents on AWS, GCP and Azure.”

Ontology Mapping & Concepts

Large Language Models (LLM) & AI#Claude Managed Agents#AI-powered parallel-agent code rewrite (agent orchestration)
Primary Source Grounding & Direct Attribution
Direct Origin Attribution
Primary Reporting: The Pragmatic Engineer•Published: Jul 28, 2026
Original Coverage Title: “How building software is changing at Anthropic”

Related Market Signals & Shifts

Recent verified developments and strategic activity across this market segment.

Large Language Models (LLM) & AIMar 4, 2026

Boris Cherny Explains Building Anthropic's Claude Code

Boris Cherny, creator and Head of Claude Code at Anthropic, discusses how Claude Code evolved from a side project into a core internal developer tool and the engineering practices that enable agent-driven development. He describes a high-velocity workflow (running multiple parallel Claude instances to produce many PRs per day), retrieval strategies (simple glob/grep driven by the model outperformed more complex RAG/indexing approaches), and deterministic review and sandboxing patterns used to manage safety. The episode also covers Claude Cowork — a VM-based desktop agent built rapidly to serve non-engineers — and organizational shifts: prototypes replacing PRDs, automation of repetitive code-review comments, and a changing skills profile for engineers as agentic tools handle more of the implementation work.

Read assessment
Large Language Models (LLM) & AIApr 20, 2026

Intercom Doubles Engineering Velocity Using Claude Code

Brian Scanlan, senior principal engineer at Intercom, describes how the company adopted an AI-first engineering approach and used Claude Code to double merged PRs per R&D employee in nine months while maintaining code quality. Intercom rolled out Claude Code broadly—100% of engineers plus designers, PMs and TPMs ship code via the tool—and built telemetry (using Honeycomb) to measure AI adoption and quality. They created a skills repository that enforces engineering standards, implemented permission and accountability frameworks, and prepared their product and infrastructure for an agent-first future with CLIs, Model Context Protocol (MCP) servers, and ephemeral APIs. The interview includes demos, a deep dive on flaky-spec automation, and discussion of treating AI spend as an investment and achieving 'backlog zero.'

Read assessment
Large Language Models (LLM) & AIApr 30, 2026

10-Agent AI Product Team in Claude Code

A developer describes building a 10-agent AI product team using Claude Code's Agent Teams feature to orchestrate product development stages (ideation through go-to-market). Each agent is defined as a markdown file in a .claude/agents folder and runs in its own context; agents communicate directly and a lead orchestrator ('Athina') enforces stage gates and runs 'Grill Me' challenge sessions. The author migrated from an OpenClaw setup to Claude Code to reduce infrastructure friction and token costs, splitting agents across Opus 4.6 (open-ended reasoning) and Sonnet 4.6 (procedural checklist work). The workflow uses the Superpowers plugin to enforce TDD, Playwright for E2E QA, and a Codex (GPT) adversarial review step to provide cross-model code review. The post highlights cost, portability, and design-alternatives before commitment.

Read assessment
Executive Briefing & Watchlist

Receive all future signals for Anthropic & competitors in your briefing

Never miss material M&A deals, leadership changes, account wins, or product rollouts. Polaris7 monitors 20,000+ primary sources 24/7, delivering verified telemetry directly into your executive briefing.

Daily primary source scansCurated executive briefingsUnsubscribe anytime