AUGUST 9, 2026 — NVIDIA NOOA OPEN-SOURCE · ANTHROPIC $1.5B ENTERPRISE JV
- NVIDIA NOOA open-sourced: Apache 2.0 Python agent framework — agent = one Python class (methods are actions, fields are state, docstrings are prompts). SWE-bench Verified 82.2%, CyberGym L1 86.8%, ARC-AGI-3 85.1%. Model-agnostic via LiteLLM. v0.0.8 alpha, Python 3.12-3.13. Security warning: AST checks are NOT containment — always run in container or VM. Full review + install guide →
- Anthropic launches Ode With Anthropic: $1.5B JV with Blackstone and Hellman & Friedman now live with 100 engineers. Backed by Goldman, General Atlantic, Apollo, GIC, Sequoia. Targets mid-sized banks, health systems, manufacturers. Deploys Claude in regulated sovereign configurations. Launched same week as CGAO hire — coordinated enterprise + regulatory push. Full analysis →
Story 1 — NVIDIA NOOA: The Cleanest Agent Framework So Far
Per NVIDIA's technical blog, NOOA's key innovation is structural: an agent is a Python class, not a collection of prompt templates and callback graphs. Methods whose body is ... are completed by an LLM at runtime. Methods with normal bodies are deterministic Python. Everything — state, prompts, actions, contracts — lives in one class that can be tested, traced, and version-controlled like ordinary software. SWE-bench Verified 82.2% is a strong result for an alpha framework. The containment warning in the README is important: AST checks do not stop code that is actively trying to escape a sandbox. Run NOOA agents in a container, VM, or NVIDIA OpenShell. Install with pip install nooa. Full review →
Story 2 — Anthropic's Enterprise Push: Ode + CGAO + FLI C+
Ode With Anthropic is the implementation vehicle that turns Anthropic's model API into an enterprise product a bank compliance team can sign off on. The $1.5B JV with Blackstone and H&F provides the private equity distribution relationships that Anthropic's direct sales motion doesn't reach. 100 engineers are live deploying Claude at mid-sized banks, health systems, and manufacturers in data-sovereign, compliance-ready configurations. Combined with Tino Cuéllar's CGAO appointment (former CA Supreme Court Justice) and Anthropic's Politico-confirmed state-by-state AI safety law push, this week's Anthropic news tells one consistent story: the company is making a coordinated move into regulated enterprise as its primary growth market. The FLI C+ safety rating — the highest of any frontier lab — is the commercial differentiator in those conversations. Full analysis →