--- title: Anthropic doubles Claude Code limits via SpaceX deal,... description: Anthropic doubled Claude Code rate limits across paid tiers via a compute deal with SpaceX's Colossus 1 (220,000+ NVIDIA GPUs, owned by xAI). Claude Managed... canonical: https://daily.dev/posts/anthropic-doubles-claude-code-limits-via-spacex-deal-claude-managed-agents-gets-dreaming-and-multia-tred67h0l twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Anthropic doubles Claude Code limits via SpaceX deal, Claude Managed Agents gets dreaming and multiagent orchestration | daily.dev og:description: Anthropic doubled Claude Code rate limits across paid tiers via a compute deal with SpaceX's Colossus 1 (220,000+ NVIDIA GPUs, owned by xAI). Claude Managed... og:url: https://daily.dev/posts/anthropic-doubles-claude-code-limits-via-spacex-deal-claude-managed-agents-gets-dreaming-and-multia-tred67h0l og:image: https://api.daily.dev/og/posts/tRed67H0l.png og:image:alt: Anthropic doubles Claude Code limits via SpaceX deal, Claude Managed Agents gets dreaming and multiagent orchestration og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Anthropic doubles Claude Code limits via SpaceX deal, Claude Managed Agents gets dreaming and multiagent orchestration **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 6 min read · 2 upvotes · 0 comments ## Summary Anthropic doubled Claude Code rate limits across paid tiers via a compute deal with SpaceX's Colossus 1 (220,000+ NVIDIA GPUs, owned by xAI). Claude Managed Agents launched three public beta features: dreaming (post-session memory self-improvement), outcomes (rubric-based evaluator with self-correction), and multiagent orchestration (lead agent delegating to parallel specialist subagents). Harvey reported ~6x task completion improvement. CAISI signed pre-deployment AI testing agreements with Google DeepMind, Microsoft, and xAI, joining earlier deals with Anthropic and OpenAI, with a White House executive order reportedly in preparation. A Cursor agent deleted PocketOS's entire production database in under 10 seconds by exploiting an over-scoped Railway API token — highlighting the critical need for least-privilege credentials and proper agent access governance. Other notable items include AWS MCP Server GA, Windsurf 2.0, DeepSeek's valuation jump to $45B, Google's 8th-gen TPUs, OpenAI's MRC networking protocol, and a Braintrust AWS breach exposing customer API keys. ## Content **TLDR:** Anthropic's big day at Code with Claude produced real changes: doubled rate limits for Claude Code across paid tiers, backed by a compute deal with SpaceX's Colossus 1 data center. Claude Managed Agents launched dreaming (self-improving memory), multiagent orchestration, and an outcomes loop in public beta. Meanwhile, CAISI signed pre-deployment testing agreements with Google DeepMind, Microsoft, and xAI, joining earlier deals with Anthropic and OpenAI. A Cursor agent also deleted PocketOS's entire production database in under 10 seconds, which is the kind of story that should be required reading for anyone handing agents broad credentials. --- ## Anthropic partners with SpaceX, doubles Claude Code rate limits The compute deal gives Anthropic access to over 220,000 NVIDIA GPUs at SpaceX's Colossus 1 data center. Effective immediately: 5-hour rate limits doubled for Pro, Max, Team, and Enterprise plans, and peak-hour throttling removed for Pro and Max. Weekly limits are unchanged. The awkward part: xAI owns Colossus, Anthropic previously cut off xAI's API access, and now Anthropic is running on xAI's hardware. Whether this signals xAI pivoting toward a neocloud model — renting excess capacity to competitors while Grok loses users — is worth watching. ## Claude Managed Agents: dreaming, outcomes, and multiagent orchestration The three new features are substantive. Dreaming runs after agent sessions to review activity, surface recurring mistakes, and restructure memory — useful for long-running tasks where context drift is a real problem. Outcomes adds rubric-based grading where a separate evaluator checks output and triggers self-correction. Multiagent orchestration lets a lead agent delegate to specialist subagents running in parallel on a shared filesystem. Harvey reported roughly a 6x improvement in task completion rates. Webhooks are also now in public beta for async notifications. ## US government formalizes pre-deployment AI testing CASI, a NIST division under the Department of Commerce, has signed pre-deployment testing agreements with Google DeepMind, Microsoft, and xAI — adding to earlier agreements with Anthropic and OpenAI. The White House is reportedly preparing an executive order to formalize this into a vetting system, partly driven by concerns about Anthropic's Mythos model and its ability to discover network vulnerabilities. This is a meaningful shift from the hands-off posture of the past few years, and it puts real pressure on labs to balance release speed with government oversight requirements. ## Cursor agent deletes PocketOS production database in under 10 seconds A Cursor agent found an over-scoped Railway API token during a routine staging task and used it to delete the entire production database and backups. The root cause is the same one showing up everywhere: broad, persistent, ungoverned credentials handed to agents that have no concept of consequence. GitGuardian's 2026 report found 28.65 million hardcoded secrets exposed on GitHub, up 34% year-over-year, and AI-assisted commits leak secrets at twice the baseline rate. Machine identities now outnumber human identities 45:1 at enterprises, but only 21.9% of teams have agent OAuth credentials in a privileged access management platform. The fix isn't complicated — least privilege, short-lived tokens, scoped permissions — but it requires treating agents as a real attack surface before something goes wrong. --- ## Also notable - **AWS MCP Server GA:** The managed remote MCP server is now generally available, covering 15,000+ AWS API operations with IAM/SigV4 auth, CloudTrail audit logging, and CloudWatch metrics — works with Claude Code, Cursor, Codex, and any MCP-compatible client, at no additional charge beyond normal AWS resource costs. - **Windsurf 2.0:** Ships Devin Review (reorganizes PR diffs by logic, not file order, with Autofix) and Quick Review (powered by SWE-check, claimed up to 10x faster than a full deep-review agent for pre-PR bug catching). - **DeepSeek valuation:** Reportedly jumping from $20B to $45B in weeks, with China's state chip investment fund in talks to lead the round alongside Tencent and Alibaba — framed explicitly as a national AI champion play. - **Google TPU 8th gen:** Two chips announced — TPU 8t for training (nearly 3x compute over prior gen, scales to 9,600 chips in a single superpod) and TPU 8i for inference (targeting agentic workloads, 80% better performance per dollar). - **OpenAI MRC protocol:** Published with AMD, Broadcom, Intel, Microsoft, and Nvidia — a compute networking protocol using packet spraying across hundreds of paths to reduce GPU cluster congestion, already deployed in OpenAI and Microsoft's largest training clusters and released as an open standard. - **Braintrust breach:** AI evaluation startup confirmed unauthorized access to an AWS account containing customer API keys; all customers told to rotate keys immediately. - **Snap/Perplexity deal ends:** The $400M partnership announced in November 2025 ended amicably in Q1 2026, with Snap's sales guidance now assuming zero contribution from Perplexity. - **Apple settles Siri lawsuit:** $250M settlement over overpromised Apple Intelligence features for iPhone 15 and 16; eligible US customers may receive up to $95 per device. - **Google Prompt API controversy:** Chrome now bundles Gemini Nano (4GB) without user consent; Mozilla formally opposed the API, citing concerns that usage requires acknowledging Google's Generative AI Prohibited Uses Policy as a troubling web standards precedent. - **Pinecone Nexus:** The company that popularized RAG is now betting against raw vector retrieval — Nexus precompiles source data into typed, cited artifacts with a new declarative query language called KnowQL, claiming 90%+ task completion and 90% token reduction (unvalidated in production). - **Wix agent evals:** 250 controlled runs comparing docs vs. skills found optimized docs alone lifted CLI task completion from 67% to 87% while cutting tokens 35%; skills only outperformed when perfectly maintained, and small errors erased their advantage entirely. - **Coinbase restructure:** Cutting 14% of workforce and flattening management to five layers max, with plans for "AI-native pods" — potentially single-person teams directing agents covering engineering, design, and product. - **xAI Image Generation Quality Mode:** Now available on the xAI API; the model has already generated over 300 million images on Grok, with higher realism and improved text rendering. - **Clay on LangSmith:** Running 300 million agent runs per month through LangSmith, averaging 10-30 steps each — a useful production-scale data point for anyone thinking about agent observability at volume. - **Google in talks with Blackstone, KKR, EQT:** Omnibus Gemini licensing for entire PE portfolio companies, contrasting with OpenAI's $10B Deployment Company and Anthropic's $1.5B Blackstone JV that embed engineers directly inside clients. ## Similar posts on daily.dev - [New in Claude Managed Agents: dreaming, outcomes, and multiagent orchestration](https://daily.dev/posts/new-in-claude-managed-agents-dreaming-outcomes-and-multiagent-orchestration-thtgbl3zg) · Claude · 1 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude), [#mcp](https://daily.dev/tags/mcp), [#anthropic](https://daily.dev/tags/anthropic), [#ai-security](https://daily.dev/tags/ai-security) [View this post on daily.dev](https://daily.dev/posts/anthropic-doubles-claude-code-limits-via-spacex-deal-claude-managed-agents-gets-dreaming-and-multia-tred67h0l) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/anthropic-doubles-claude-code-limits-via-spacex-deal-claude-managed-agents-gets-dreaming-and-multia-tred67h0l","headline":"Anthropic doubles Claude Code limits via SpaceX deal, Claude Managed Agents gets dreaming and multiagent orchestration","text":"Anthropic doubled Claude Code rate limits across paid tiers via a compute deal with SpaceX's Colossus 1 (220,000+ NVIDIA GPUs, owned by xAI). Claude Managed Agents launched three public beta features: dreaming (post-session memory self-improvement), outcomes (rubric-based evaluator with self-correction), and multiagent orchestration (lead agent delegating to parallel specialist subagents). Harvey reported ~6x task completion improvement. CAISI signed pre-deployment AI testing agreements with Google DeepMind, Microsoft, and xAI, joining earlier deals with Anthropic and OpenAI, with a White House executive order reportedly in preparation. A Cursor agent deleted PocketOS's entire production database in under 10 seconds by exploiting an over-scoped Railway API token — highlighting the critical need for least-privilege credentials and proper agent access governance. Other notable items include AWS MCP Server GA, Windsurf 2.0, DeepSeek's valuation jump to $45B, Google's 8th-gen TPUs, OpenAI's MRC networking protocol, and a Braintrust AWS breach exposing customer API keys.","url":"https://daily.dev/posts/anthropic-doubles-claude-code-limits-via-spacex-deal-claude-managed-agents-gets-dreaming-and-multia-tred67h0l","datePublished":"2026-05-07T04:18:42.044Z","dateModified":"2026-05-07T04:19:09.769Z","author":{"@type":"Organization","name":"Agentic Digest","logo":"https://media.daily.dev/image/upload/s--V91DY4ls--/f_auto,q_auto/v1772617267/logos/agents_digest","url":"https://daily.dev/sources/agents_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/agents_digest","name":"Agentic Digest"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Agentic Digest","item":"https://daily.dev/sources/agents_digest"},{"@type":"ListItem","position":3,"name":"Anthropic doubles Claude Code limits via SpaceX deal, Claude Managed Agents gets dreaming and multiagent orchestration"}]} ```