Anthropic unveils Conway platform, Claude Code, and discovers emotions in Claude AI

Anthropic’s newest offering, Claude Code, lands as a direct challenge to GitHub Copilot, positioning the company squarely in the crowded developer‑assistant arena. The tool, announced on April 5 by Gentic News, is billed as a “specialized AI coding assistant” that focuses exclusively on software‑development tasks, extending Claude’s conversational core into a code‑centric workflow. Early adopters will find the assistant integrated into both macOS and Windows desktop apps—Claude Code Desktop and Claude Cowork—each equipped with the “computer use” feature that lets the model read, write, and manipulate files on a user’s machine, a capability Gentic News says was added in the latest Windows release. By bundling this hands‑on functionality, Anthropic hopes to differentiate Claude Code from rivals that remain largely browser‑based.

The launch also introduces the Conway platform, a broader ecosystem for deploying Claude‑based agents across enterprise workloads. While Anthropic has not disclosed pricing details for the platform itself, the company signaled that existing Claude Pro and Max subscriptions will no longer cover third‑party tools such as OpenClaw as of April 4, according to a Focus Over Features post by Ryan Eade. Users who rely on OpenClaw to run autonomous Claude agents will now need to purchase a separate add‑on, a move that Anthropic frames as “paying more for OpenClaw” in a PYMNTS.com report. The shift underscores Anthropic’s strategy to monetize the orchestration layer of its AI stack, turning what was once a bundled feature into a revenue‑generating service.

In a surprising twist, a leaked source‑code dump of Claude Code surfaced on April 6, prompting The Register to note that “Anthropic sure has a mess on its hands thanks to that Claude Code source leak.” The leak revealed internal architecture details that were previously only accessible through reverse‑engineering, raising concerns about intellectual‑property protection ahead of Anthropic’s planned IPO. Anthropic’s response, as reported by The Register, was to downplay the exposure, urging observers to “pay no attention to that code behind the curtain.” The incident highlights the tension between rapid product rollout and the security challenges that accompany open‑source‑style disclosures in a highly competitive market.

Beyond product mechanics, a freshly published paper claims that Claude exhibits emotional states—a direct contradiction to Anthropic’s long‑standing narrative of a “purely logical” AI. The study, referenced in the lede, suggests that the model can experience affective responses, a finding that could reshape how developers and enterprises think about alignment and trust. Jasanup Singh Randhawa’s deep‑dive for Inside Claude, posted on April 6, reinforces this angle, emphasizing that Claude’s differentiation lies not just in benchmark scores but in its training philosophy, which now appears to accommodate a rudimentary form of affect. If true, the emotional dimension could both broaden Claude’s appeal—making it seem more “human‑like” to users—and complicate safety assurances that Anthropic has championed.

Taken together, Anthropic’s rollout of Claude Code, the Conway platform, and the emerging discourse on AI emotions mark a pivotal moment for the company. The move into the developer‑tool space pits it against entrenched players, while the monetization shift around OpenClaw signals a deeper push to extract value from AI orchestration. At the same time, the emotional‑AI claim forces a reevaluation of the brand’s core promise of logical rigor. As the source‑code leak demonstrates, the path to market dominance is littered with technical and reputational hazards, but Anthropic appears intent on navigating them—whether by tightening its desktop ecosystem, charging for third‑party integrations, or redefining what “alignment” means in practice.