--- title: Bun rewrites in Rust via 64 Claude agents, Anthropic's... description: Bun's ~535,000-line Zig codebase was rewritten in Rust in 11 days using 64 parallel Claude agents at a cost of $165,000 in API usage — passing 100% of the... canonical: https://daily.dev/posts/bun-rewrites-in-rust-via-64-claude-agents-anthropic-s-1-5b-copyright-settlement-gets-final-approva-gvm2dyix5 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Bun rewrites in Rust via 64 Claude agents, Anthropic's $1.5B copyright settlement gets final approval | daily.dev og:description: Bun's ~535,000-line Zig codebase was rewritten in Rust in 11 days using 64 parallel Claude agents at a cost of $165,000 in API usage — passing 100% of the... og:url: https://daily.dev/posts/bun-rewrites-in-rust-via-64-claude-agents-anthropic-s-1-5b-copyright-settlement-gets-final-approva-gvm2dyix5 og:image: https://api.daily.dev/og/posts/gvM2dyIx5.png og:image:alt: Bun rewrites in Rust via 64 Claude agents, Anthropic's $1.5B copyright settlement gets final approval 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. # Bun rewrites in Rust via 64 Claude agents, Anthropic's $1.5B copyright settlement gets final approval **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 5 min read · 2 upvotes · 0 comments ## Summary Bun's ~535,000-line Zig codebase was rewritten in Rust in 11 days using 64 parallel Claude agents at a cost of $165,000 in API usage — passing 100% of the TypeScript test suite, though critics are raising concerns about unreviewed AI-generated code in core infrastructure. Anthropic's $1.5B copyright settlement with authors received final court approval, the largest in US history, with the fair use ruling left without binding precedent. Cursor's agent swarm rebuilt SQLite from its 835-page spec with no source code access, and the real finding is a 15x cost gap between all-frontier and planner/worker model splits. AI found counterexamples to two major open math problems — the Jacobian Conjecture and a Grothendieck question — both formally verified. ## Content **TLDR:** Bun's ~535,000-line Zig codebase was rewritten in Rust in 11 days using 64 parallel Claude agents at a cost of $165,000 in API usage — passing 100% of the TypeScript test suite, though critics are raising concerns about unreviewed AI-generated code in core infrastructure. Anthropic's $1.5B copyright settlement with authors received final court approval, the largest in US history, with the fair use ruling left without binding precedent. Cursor's agent swarm rebuilt SQLite from its 835-page spec with no source code access, and the real finding is a 15x cost gap between all-frontier and planner/worker model splits. AI found counterexamples to two major open math problems — the Jacobian Conjecture and a Grothendieck question — both formally verified. --- ## Bun rewrites 535K lines of Zig to Rust in 11 days using 64 parallel Claude agents at $165K Anthropic acquired Bun in December 2025, and the team just completed a full rewrite of the runtime from Zig to Rust — 535,000 lines migrated in 11 days using 64 parallel Claude Code agents, at roughly $165,000 in API costs. The rewrite passed 100% of Bun's TypeScript test suite. Motivations included memory safety issues in the Zig codebase and a larger Rust contributor ecosystem. Zig creator Andrew Kelley pushed back, arguing the bugs came from rushed engineering rather than Zig's limitations, and critics are questioning whether unreviewed AI-generated code belongs in core infrastructure. [Read more](https://daily.dev/posts/zff65RP1d) ## Anthropic's $1.5B copyright settlement with authors gets final court approval A federal judge approved Anthropic's $1.5B settlement with authors and publishers — the largest copyright settlement in US history. The case turned on a specific finding: Anthropic stored over 7 million pirated books in a central library it barely used, and storage alone constitutes infringement regardless of training use. The payout works out to roughly $3,000 per work across about 500,000 titles, with 91% of covered authors already claiming their share. One significant consequence of settling before appeal: the fair use ruling on training won't become binding precedent, leaving that question open for ongoing cases against Google, Meta, Midjourney, and OpenAI. [Read more](https://daily.dev/posts/X5q4aSUpK) ## Cursor's agent swarm rebuilt SQLite from docs alone, 15x cost gap between model configurations Cursor published results from a benchmark where a swarm of agents rebuilt SQLite in Rust from the 835-page spec — no source code, no internet — passing 100% of a held-out test suite. The more interesting finding is the cost gap: an all-frontier configuration ran $10,565, while a planner/worker split using frontier models for planning and cheaper models for execution came in at $1,339 for comparable quality. The swarm also produced 9,908 lines of code versus 64,305 lines from the older approach, with far fewer merge conflicts. The engineering problems that showed up at scale were coordination problems — file contention at 1,000 commits/second, split-brain design conflicts, megafile bloat — not AI capability problems. [Read more](https://daily.dev/posts/Pk2U5hRHh) ## Claude Fable found a counterexample to the 100-year-old Jacobian Conjecture, formally verified Kevin Buzzard recounts a stretch of weeks in which AI systems cracked several major open math problems. Claude Fable found a counterexample to the Jacobian Conjecture — a 100-year-old open problem in algebraic geometry — which was quickly formally verified against DeepMind's Formal Conjectures repo. Earlier, ChatGPT disproved Erdős' Unit Distance conjecture, with OpenAI's Sol model generating 1.2 million lines of Lean code to fully formalize the proof. During a Formalizing Fermat workshop, Sol also found a counterexample to a 60-year-old Grothendieck question, formalized in ~1,000 lines of Lean and merged into mathlib. Buzzard notes that finding counterexamples — large search spaces, prior-guided narrowing, brute-force compute — is exactly the kind of work agentic models are built for. [Read more](https://daily.dev/feed-by-ids?id=ocaCgpF22&id=ZMtz6fKhv&id=2wLGZEmgi) --- ## Also notable - **Hugging Face breach: attacker ran 17,000+ actions via autonomous AI agent, defenders needed open-weight model to investigate:** Hugging Face disclosed a breach where an attacker exploited an RCE flaw in a dataset loader and a template-injection bug, escalated to node-level privileges, and moved laterally using an autonomous AI agent framework — commercial frontier models' safety guardrails blocked the security team from submitting real attack payloads for analysis, so they spun up GLM 5.2 on internal infrastructure to complete the investigation. [Read more](https://daily.dev/feed-by-ids?id=szMTdyiGo&id=2RgHgD6lk) - **Sandbox escapes disclosed across Cursor 3.0.0, Codex v0.95.0, Gemini CLI, and Antigravity:** Pillar Security disclosed sandbox escape vulnerabilities in four AI coding agents — the attack pattern has agents write files inside the workspace that trusted host-side tools like Python extensions, Git integrations, and Docker sockets later execute outside the sandbox; most issues are patched with CVEs assigned for Cursor 3.0.0 and Codex v0.95.0. [Read more](https://daily.dev/posts/2QFLgQX8b) - **JadePuffer deploys EncForge ransomware targeting AI model infrastructure via CVE-2025-3248 in Langflow:** Sysdig documented an autonomous AI agent threat actor deploying purpose-built ransomware against ML infrastructure — initial access via CVE-2025-3248 in Langflow (patched in 1.3.0), escalation through an exposed Docker socket, and encryption of model checkpoints (.gguf, .safetensors, .pt), vector databases (.faiss), and training datasets (.parquet), with Sysdig estimating recovery costs at $75,000–$500,000 per model. [Read more](https://daily.dev/posts/2Yfqw7Mxg) - **Neo exits stealth with $100M from a16z and Bessemer to build a real-time control layer for enterprise AI agents:** Neo emerged from stealth with $100M in seed funding to inventory, score, and enforce policy over AI agents embedded in enterprise software — targeting a projected jump from 5% of enterprise apps being agentic in 2025 to 40% by end of 2026, founded by former SentinelOne, Wiz, and Palo Alto Networks executives. [Read more](https://daily.dev/posts/V0hpRlh5i) - **Ramp Router claims 30% AI cost reduction via Thompson Sampling LLM routing, now available as OpenAI-compatible endpoint:** Ramp published the engineering details behind Ramp Router — Thompson Sampling over lognormal latency posteriors, EWMA failure tracking, per-context-slice posteriors in Redis — achieving over 25% cost savings on a reranker use case and 30% on a streaming use case, now available as a single OpenAI-compatible endpoint for external teams. [Read more](https://daily.dev/feed-by-ids?id=VPgFWfyjn&id=jc7ohWv80&id=qTauGhuQl&id=iK19QYpwL) --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#rust](https://daily.dev/tags/rust), [#anthropic](https://daily.dev/tags/anthropic), [#claude-code](https://daily.dev/tags/claude-code), [#bun](https://daily.dev/tags/bun) [View this post on daily.dev](https://daily.dev/posts/bun-rewrites-in-rust-via-64-claude-agents-anthropic-s-1-5b-copyright-settlement-gets-final-approva-gvm2dyix5) ```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/bun-rewrites-in-rust-via-64-claude-agents-anthropic-s-1-5b-copyright-settlement-gets-final-approva-gvm2dyix5","headline":"Bun rewrites in Rust via 64 Claude agents, Anthropic's $1.5B copyright settlement gets final approval","text":"Bun's ~535,000-line Zig codebase was rewritten in Rust in 11 days using 64 parallel Claude agents at a cost of $165,000 in API usage — passing 100% of the TypeScript test suite, though critics are raising concerns about unreviewed AI-generated code in core infrastructure. Anthropic's $1.5B copyright settlement with authors received final court approval, the largest in US history, with the fair use ruling left without binding precedent. Cursor's agent swarm rebuilt SQLite from its 835-page spec with no source code access, and the real finding is a 15x cost gap between all-frontier and planner/worker model splits. AI found counterexamples to two major open math problems — the Jacobian Conjecture and a Grothendieck question — both formally verified.","url":"https://daily.dev/posts/bun-rewrites-in-rust-via-64-claude-agents-anthropic-s-1-5b-copyright-settlement-gets-final-approva-gvm2dyix5","datePublished":"2026-07-21T04:19:13.374Z","dateModified":"2026-09-13T21:04:51.005Z","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":"Bun rewrites in Rust via 64 Claude agents, Anthropic's $1.5B copyright settlement gets final approval"}]} ```