XNews

Anthropic's Claude as Autonomous Colleague and Internal Code Revolution

Anthropic's Claude as Autonomous Colleague and Internal Code Revolution

Anthropic executive Boris demonstrated Claude’s evolution from a passive tool to an autonomous colleague running in the background. In Slack, Claude proactively reads channel discussions, independently pulls data from Datadog and BigQuery, synthesizes answers, draws UI design sketches, and joins multi‑person conversations—without being explicitly @‑mentioned. When Boris casually complained about confusing audio icons, Claude fetched usage data across two data sources, provided alternatives, created design sketches, and later involved a designer in the thread. Internally, 90% of Anthropic’s code is now written by Claude Code; Boris has been at 100% since November. Traditional frontend/backend/design roles have been replaced by five new roles: prototyper, builder, maintainer, scaler, and finisher. The scaler role is currently in high demand, reflecting the need to grow and harden AI‑built products. The vision presented replaces static file systems with a dynamic, evolving conversation.

Key insights

Claude operates as a proactive colleague in Slack, reading discussions and taking actions without explicit prompts

Rather than waiting for an @‑mention, Claude now monitors Slack channels, understands context, and independently initiates data queries and design work. In the demonstration, it responded to an unstructured complaint about audio icons by gathering cross‑database usage statistics, synthesizing them, and proposing solutions.

The demo integrated Datadog and BigQuery data to answer a user’s casual remark and produced UI design alternatives

Claude pulled quantitative usage data from two separate data platforms, combined the information into a coherent answer, offered several alternative designs, and then drew actual sketches—later pulling in a human designer for a multi‑party discussion.

90% of Anthropic’s code is now generated by Claude Code; Boris has been at 100% AI-written code since November 2025

The company’s internal software development has undergone a dramatic shift, with the vast majority of production code being written by its own AI system, and the executive running his own work entirely on AI-generated code for several months.

Traditional frontend/backend/design role boundaries have dissolved into five new AI‑era software roles

Boris stated that standard tri‑partite roles are obsolete because everyone can now write code. The team now consists of prototypers (rapid idea testing), builders (market‑ready product creation), maintainers (at‑scale reliability), scalers (10×‑100× expansion), and finishers (polish and de‑risking).

The “scaler” role is currently the most sought‑after at Anthropic, indicating a new bottleneck in AI‑driven development

As AI handles initial building and prototyping, the critical need shifts to scaling and hardening products for large audiences, making the scaler—the person who takes a proven product and grows it by orders of magnitude—extremely valuable.

The ideal interface is no longer a static file system but a continuous, multi‑user conversation with an AI participant

Boris described a future where, instead of exchanging files, teams engage in an ongoing dialog that includes the AI, allowing dynamic, participatory product evolution that blurs the line between tool and colleague.

Source posts