Anthropic 'Code w/ Claude 2026' Developer Event and Same-Day Announcements

On May 6, 2026, Anthropic held its 'Code w/ Claude 2026' developer conference in San Francisco, paired with announcements of a compute deal with SpaceX's Colossus 1 data center (300+ megawatts, 220,000+ NVIDIA GPUs) and doubled Claude Code rate limits for Pro, Max, Team, and Enterprise plans.[2][1] The SpaceX partnership — part of a stated 10+ GW multi-provider compute portfolio — is drawing both straightforward tech press coverage and pointed skepticism about Anthropic's safety-first brand pairing with a company controlled by Elon Musk.[4][3] By mid-May, Claude Code itself had reached version 2.1.139, introducing Agent View and Background Sessions features that signal continued rapid product development beyond the conference.[12]

Anthropic is making a coordinated bet that massive compute investment translates directly into developer loyalty — and the SpaceX partnership tests whether its safety-and-values brand can coexist with infrastructure deals that invite political scrutiny. The rate-limit doubling is a concrete near-term benefit for practitioners, but the gap between orbital-compute ambition and day-to-day reliability complaints reveals the distance between Anthropic's aspirational framing and ground-level developer experience.

On May 6, 2026, Anthropic held its 'Code w/ Claude 2026' developer conference in San Francisco. Simon Willison attended and live-blogged the morning keynote sessions, serving as a primary public relay for announcements.[1] The same day, Anthropic published a formal announcement disclosing that it had signed an agreement with SpaceX to access the full Colossus 1 data center — over 300 megawatts and more than 220,000 NVIDIA GPUs — with access expected within a month.[2] Alongside the compute deal, Anthropic doubled Claude Code's five-hour rate limits for Pro, Max, Team, and seat-based Enterprise plans and removed the peak-hours rate-limit reduction for Pro and Max subscribers, framing both changes as a direct consequence of the new capacity.[2]

The SpaceX deal sits within a broader infrastructure buildout: Anthropic disclosed a multi-provider compute portfolio exceeding 10 GW of future capacity spanning Amazon, Google/Broadcom, Microsoft/NVIDIA, and Fluidstack, and stated interest in working with SpaceX to develop 'multiple gigawatts of orbital AI compute capacity.'[2] Anthropic tied partner selection explicitly to geopolitical criteria — 'democratic countries whose legal and regulatory frameworks support investments of this scale' with secure hardware supply chains — a framing that immediately invited scrutiny given SpaceX's control by Elon Musk.[2] Wired's headline 'Anthropic Gets in Bed with SpaceX' captured the pointed editorial reaction, while Ars Technica covered the story factually and without editorial coloring.[3][4]

Developer community response to the announcement was shaped by real-world context that had been building for weeks. Fiberplane published a detailed case study on its internal Claude Code and agent usage.[5] Individual practitioners shared first-person accounts: one developer migrated 16 websites between Linode servers in a single day using Claude Code,[6] while a technical post explored pro-grade Java development in Claude Code via LSP integration.[7] Against this backdrop of enthusiasm, a credibility note landed on May 1 when users reported that Claude Opus 4.7 quotas on AWS Bedrock had abruptly dropped to 0 tokens per minute without warning — raising questions about the consistency of access across Anthropic's distribution channels even as massive new compute was being announced.[8] Separately, a Hacker News thread asked whether any open-source model and harness had reached Claude Sonnet or Opus performance levels, reflecting a segment of developers actively watching for alternatives.[9]

Post-conference momentum continued on multiple fronts. Attendee write-ups appeared in the days following the event,[10] and by May 9, a full YouTube playlist of the Code with Claude 2026 sessions made conference content available beyond live-blog summaries.[11] By May 13, Claude Code had reached version 2.1.139, with a Hacker News discussion focused on new Agent View and Background Sessions capabilities — indicating that Anthropic's product release cadence through the conference period was sustained well into the weeks that followed.[12] OpenAI remained a competitive backdrop throughout: it was reported in early April to be targeting Anthropic directly with $100/month ChatGPT Pro subscriptions.[13]