Meta released Muse Code in beta on Wednesday, a terminal-based agentic coding tool built on the Muse Spark 1.2 model and priced below Claude Code and OpenAI Codex.

• Muse Code handles planning, writing, and validating code against large codebases autonomously, positioning it as a full software engineering agent rather than an autocomplete assistant.
• Pricing undercuts rivals as Meta's primary competitive lever, following the same playbook used with Llama to pressure closed-model incumbents on cost.

Tenable today expanded its One Exposure Management Platform to cover Google Gemini, MCP servers, and major AI developer tools, completing coverage across all three frontier model providers.

• Tenable One AI Exposure now spans OpenAI, Anthropic Claude, and Google Gemini surfaces, meaning security teams can assess risk across every major enterprise LLM deployment from a single platform.
• MCP server coverage is the notable addition: as enterprises adopt model context protocol integrations, Tenable is surfacing them as a discrete attack surface category.

Cloudflare launched an Identity-Aware AI Gateway that ties every AI model request to a named employee or agent, enabling per-user spend limits and sensitive data flagging.

• The product addresses two problems simultaneously: runaway AI API costs from unattributed requests and lack of visibility into what data workers are sending to external models.
• Agent identity is a first-class concept in the product, meaning automated workflows and human users are tracked through the same policy engine.

Anthropic disclosed it is actively recruiting engineers with silicon shipping experience to build a custom AI chip co-designed with future Claude models.

• Co-design is the stated goal, meaning chip specs will be tailored to Claude's specific memory and compute profile rather than running on general-purpose accelerators with mismatched specs.
• The move follows Google TPUs and Amazon Trainium as the third major frontier lab to commit to proprietary inference and training silicon, tightening the vertical integration race.

HubSpot's Q2 2026 earnings call revealed that its AI agents are seeing rapid enterprise uptake, with revenue of $911.7M beating estimates as AI-driven products become a measurable growth driver.

• Agent activation numbers: Data Agent surpassed 16,000 customers (up 80% QoQ), Prospecting Agent reached nearly 17,000 (up 28%), and Customer Agent crossed 10,000, showing broad-based adoption rather than a single product spike.
• Upmarket momentum tied to AI: Deals over $120K ARR grew 38% year-over-year, and 64% of new Pro Plus customers activated AI features, suggesting AI is the primary driver pulling buyers into higher tiers.