Claude Code Goes Cloud-Native — Coding Becomes a Background Process

Claude Code Goes Cloud-Native — Coding Becomes a Background Process

Anthropic also rolled out Claude Code cloud sessions, now generally available to Pro, Max, Team, and Enterprise users [3]. The pitch is simple: delegate coding tasks to Anthropic-managed VMs, close your laptop, and the work keeps running. You can steer sessions in real time from browser, app, or CLI, and GitHub integration means it can auto-fix PRs without you babysitting a terminal [4][5].

This is the clearest signal yet that "vibe coding" — one person, one prompt, one output — is giving way to agent management. You're no longer writing code so much as running a small fleet of coding agents and checking in on their progress. The bonus credits promotion ($100 Pro, $250 Max, through October 7) is Anthropic's way of getting teams to actually restructure their workflows around this, not just try it once [3].

Developer reaction has centered on parallel workflows fitting naturally into team SDLC processes, with genuine enthusiasm about orchestration that doesn't lock you into one harness — support for 27+ harnesses via related tooling keeps this open rather than walled-garden [4]. That's the detail worth watching: infrastructure that respects existing team processes tends to win adoption faster than infrastructure that demands you rebuild around it.

Money Keeps Flowing Into AI Security and Infrastructure

Cyera closed a $400M Series G extension from Goldman Sachs, pushing its valuation to roughly $12B and total raised to $2.7B — all in data and AI-agent security [6]. That's not a random bet; as companies hand more autonomy to agents (see: Claude Code cloud sessions above), securing what those agents can touch becomes its own category, not a feature bolted onto existing security tools.

The broader round-up tells the same story from different angles: Cognition's $2B+ Series E, Temporal's $550M, Snorkel's $350M, plus earlier moves like Bessemer's $5.75B AI fund lineup and Tekever's $580M defense-AI raise [6]. H1 2026 alone saw $377.6B in US AI funding across nearly 1,500 deals — money is not the constraint right now, judgment about where to deploy it is [6].

DeepSeek's reported ~$1B ARR is worth flagging too — proof that strong revenue models exist outside the US lab duopoly, and that capital is chasing infrastructure and agent-governance plays as hard as it's chasing raw model capability [6]. If you're building on AI right now, the money is telling you where the next bottleneck will be: not intelligence, but control and security around what that intelligence is allowed to do.

EU AI Act Enforcement Begins — With Room to Breathe

The AI Act's enforcement phase formally started August 2, 2026, with the AI Office and national authorities now active [7]. But the July 27 Digital Omnibus (Regulation 2026/1744) softened the landing considerably: high-risk Annex III obligations are pushed to December 2027, and embedded product rules stretch out to August 2028 [8]. Transparency requirements — labeling AI-generated content — apply now, though pre-existing models get until December 2026 to comply [9].

The one area with zero delay: new prohibitions on non-consensual intimate imagery and CSAM, effective December 2, 2026 [8]. That's the EU signaling where it draws hard lines versus where it's willing to phase things in.

For companies operating in or selling into Europe, this is a "start now, not later" moment. The delays buy time, not exemption — and the gap between EU risk-based regulation and the US's lighter-touch approach is only going to get more pronounced as enforcement matures. Legal and compliance teams should be mapping to ISO/NIST now, not waiting for the 2027 deadline to feel urgent.