Congress AI Spending Exposed—ChatGPT Dominates 80% of Identifiable Costs

The story of the day

ChatGPT accounts for 80% of identifiable AI spending in the US House of Representatives, according to CNBC. The figure underscores OpenAI’s near-monopoly on government AI procurement, leaving competitors like Anthropic with crumbs. For builders, this signals two things: public-sector adoption is real, but it’s concentrated in a single vendor’s ecosystem. The question for startups: when—and if—procurement diversifies.

New tools & launches

CostClaw — CLI tool to audit Claude Code spend leaks. Open source. [HN]

Nono — Sandbox for AI agents with zero setup. Open source. [HN]

Agent-Talk — File-based protocol for AI agents to negotiate consensus. Open source. [HN]

Rever-browser — AI-agent browser for reverse-engineering. Open source. [HN]

Models & research

Rust LLM Policy — The Rust language team adopted a formal LLM usage policy, clarifying how AI tools can (and cannot) be used in core development. Open-source governance meets practical constraints. [Rust Blog]

Verified Crypto Libraries via AI — A new arXiv paper proposes an AI-driven approach to verifying production cryptographic libraries, addressing a long-standing gap in formal methods for security-critical code. Proprietary research, but the implications for open-source audit tools are immediate. [arXiv]

Industry pulse

Wall Street is growing uneasy with SpaceX’s AI spending, despite the company’s promises of rapid returns, per CNBC. Investors are questioning whether the bet on AI-driven satellite operations will pay off before capital runs dry. The tension highlights a broader skepticism: even for high-growth firms, AI’s ROI remains a leap of faith.

Stat of the day

80% of the US House’s identifiable AI spending goes to ChatGPT, according to CNBC. The figure reveals how deeply a single model has embedded itself in government workflows—before procurement rules catch up.

Get the full analysis — subscribe to Gradient Pro.