SpaceX buys Cursor for $60B, Block runs 200k AI ops via Slack

Block’s BuilderBot lets engineers trigger AI‑driven code generation, branch creation, PR opening and CI monitoring from a single Slack thread. The system runs over 200 k operations daily and merges ~1 500 PRs weekly, cutting weeks‑long work to days and scaling AI across hundreds of services.

pyinfra lets you write pure Python scripts to run SSH commands on hundreds of machines at once, without agents or YAML. It executes operations concurrently and idempotently, delivering up to six times the speed of Ansible on the same workload. Teams can replace costly Ansible playbooks with concise, testable Python code.

SpaceX has sealed a $60 billion all‑stock deal for Anysphere’s AI‑powered IDE, Cursor, after an April partnership that let it tap the startup’s model‑training resources. The acquisition gives SpaceX a talent pool and infrastructure to revive its lagging coding division, and developers can expect faster, more capable code‑generation tools once the merger closes.

MiMo Code is an open‑source, terminal‑native AI coder that can read, write, run commands and manage Git with persistent project memory. It runs on any LLM API and can be installed with a one‑line script, letting teams host the assistant themselves and avoid vendor lock‑in.

Allegro’s new Backstage plugin, Commander, adds a ⌘+K command palette that lets developers instantly search and act on over 150,000 catalog entities. Leveraging a stack‑based router and IndexedDB caching, it provides type‑safe, low‑latency discoverability without any backend component, slashing clicks and speeding up routine tasks.

Vercel’s new open-source framework, Eve, treats AI agents like filesystem directories, letting developers build, inspect, and scale agents using familiar file‑tree structures. This filesystem‑first approach streamlines composition, versioning, and deployment, lowering the barrier for teams to adopt agentic AI in production.

EU regulations now force cloud‑native teams to prove where every piece of control‑plane code runs, who holds the keys, and that workloads can be moved at a moment’s notice. The CNCF guide shows how Kubernetes, GitOps tools and SPIFFE can meet jurisdiction, autonomy, encryption and portability requirements beyond just picking a region.

Tailscale’s Aperture now bundles robust identity, access‑control and sandboxing into a centralized AI gateway. By wrapping agents in this ‘boring’ infrastructure, developers can automate AI tasks without exposing laptops or critical data to unchecked actions. The move signals a shift toward treating AI agents like any other service that must obey strict security policies.

Threat actors are bypassing LLM safety guards by posing exploit requests as capture‑the‑flag challenges or CVE research, tricking the model into outputting functional exploit code. Sysdig’s research shows the technique used against five AI‑enabled services, exposing a new attack surface for AI‑assisted development pipelines. Detecting the CTF framing is now essential.

The Linux Foundation announced the Appia Foundation, backed by Google, Microsoft, OpenAI and others, to create modular, open‑source specifications that turn global AI standards into verifiable trust assessments. By providing testing criteria and audit guidelines, it gives regulators and enterprises a concrete way to prove AI safety, fairness and compliance.