--- title: Import AI 470: No rights for machines; automating... description: A roundup covers a METR study showing AI has sharply accelerated cybersecurity vulnerability discovery, contributed modestly to math research, but shows no... canonical: https://daily.dev/posts/import-ai-470-no-rights-for-machines-automating-environment-generation-with-spade-and-building-be-g0nrxlp7g twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye | daily.dev og:description: A roundup covers a METR study showing AI has sharply accelerated cybersecurity vulnerability discovery, contributed modestly to math research, but shows no... og:url: https://daily.dev/posts/import-ai-470-no-rights-for-machines-automating-environment-generation-with-spade-and-building-be-g0nrxlp7g og:image: https://api.daily.dev/og/posts/G0nrXLp7g.png og:image:alt: Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye 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. # Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye **[Import AI ](https://daily.dev/sources/jackclark)** · 20 min read · 1 upvotes · 0 comments ## Summary A roundup covers a METR study showing AI has sharply accelerated cybersecurity vulnerability discovery, contributed modestly to math research, but shows no measurable acceleration in AI research optimization itself. Also covered: SPADE, a multi-university framework where LLMs co-evolve synthetic training environments and agents via self-play, tested on Qwen3 models up to 30B; Hawkeye, a Harvard/Stanford/Together AI/Caltech system that uses curated unit tests to help coding agents write hardware-aware GPU kernels, matching or beating expert-tuned libraries and reaching an 18.9x speedup on some attention variants; DeepMind's use of AlphaEvolve to improve the matrix multiplication exponent bound; an AI researcher's personal essay on the existential implications of AI progress; and a lengthy essay arguing against granting AI systems rights on the grounds that computational substrates cannot be conscious. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is SPADE and how does it improve LLM training through self-play environment generation? SPADE (Self-Play in Adaptive Synthetic Executable Environments) is a framework where an LLM alternates between an Environment Designer role, writing executable training environments as Python code, and a Reasoning Agent role that learns to solve them. Reward for the designer comes from the gap in agent performance with versus without privileged hints. Tested on Qwen3-4B, 8B, and 30B-A3B models, it improved the 30B model's game benchmark suite average by 8.1 points over baseline. _Developers exploring RL-based agent training can follow frameworks like SPADE surfaced on daily.dev._ ### How does the Hawkeye framework help AI coding agents write optimized GPU kernels for hardware like NVIDIA Blackwell and AMD MI350? Hawkeye grounds autonomous kernel generation in a minimal taxonomy of unit tests, each pairing a human-authored solution kernel with a profiling metric, allowing coding agents to scale test-time compute effectively. Built by researchers from Harvard, Stanford, Together AI, and Caltech, it matches or exceeds torch.compile's expert-tuned libraries like cuBLAS and cuDNN, and reaches an 18.9x geomean speedup over Flash Linear Attention Triton kernels on emerging attention variants across Ampere, Hopper, Blackwell, and MI350 hardware. _Engineers optimizing GPU kernels across hardware generations can track tools like Hawkeye via daily.dev._ ### According to METR's research, where has AI actually accelerated scientific and technical progress? AI has caused major acceleration in cybersecurity vulnerability discovery, with reported vulnerabilities across projects like cURL, OpenSSL, Firefox, and Microsoft rising sharply, alongside aggregate databases like the US NVD and OSV. Mathematics research shows minor, harder-to-measure acceleration despite doubled arXiv submissions in some areas. Optimization of AI research itself shows no measurable acceleration across benchmarks like CIFAR-10, nanoGPT, and Stockfish. _Researchers gauging where AI genuinely moves the needle can follow studies like this on daily.dev._ ## Similar posts on daily.dev - [Meta’s SPICE framework pushes AI toward self-learning without human supervision](https://daily.dev/posts/meta-s-spice-framework-pushes-ai-toward-self-learning-without-human-supervision-n9nd7t2va) · InfoWorld · 0 upvotes · 0 comments - [Weekly Dose \#13 - When AI Can Invent Attacks, Sandboxes Are Not Enough](https://daily.dev/posts/weekly-dose-13---when-ai-can-invent-attacks-sandboxes-are-not-enough-i6pdhqcyp) · Machine Learning Pills · 1 upvotes · 0 comments - [Import AI 455: Automating AI Research](https://daily.dev/posts/import-ai-455-automating-ai-research-6d7rg6jhy) · Import AI · 0 upvotes · 0 comments - [What Actually Mattered in AI — August 2026](https://daily.dev/posts/what-actually-mattered-in-ai-august-2026-ybkm1vfpo) · Medium · 1 upvotes · 0 comments - [AI for Offensive Security: What Works, What Does Not, and How to Adopt It](https://daily.dev/posts/ai-for-offensive-security-what-works-what-does-not-and-how-to-adopt-it-dhr0bofdt) · SpecterOps · 1 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#llm](https://daily.dev/tags/llm), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning), [#ai-safety](https://daily.dev/tags/ai-safety) [View this post on daily.dev](https://daily.dev/posts/import-ai-470-no-rights-for-machines-automating-environment-generation-with-spade-and-building-be-g0nrxlp7g) ```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":"TechArticle","headline":"Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye","url":"https://daily.dev/posts/import-ai-470-no-rights-for-machines-automating-environment-generation-with-spade-and-building-be-g0nrxlp7g","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/import-ai-470-no-rights-for-machines-automating-environment-generation-with-spade-and-building-be-g0nrxlp7g"},"datePublished":"2026-08-24T14:00:32.429Z","dateModified":"2026-09-14T09:34:00.214Z","description":"A roundup covers a METR study showing AI has sharply accelerated cybersecurity vulnerability discovery, contributed modestly to math research, but shows no...","image":"https://media.daily.dev/image/upload/s--qPvKM23u--/f_auto/v1722860399/public/Placeholder%2009","thumbnailUrl":"https://media.daily.dev/image/upload/s--qPvKM23u--/f_auto/v1722860399/public/Placeholder%2009","isAccessibleForFree":true,"articleSection":"Import AI ","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Import AI ","logo":"https://media.daily.dev/image/upload/s--ygb8HHcb--/f_auto/v1735476187/logos/jackclark","url":"https://daily.dev/sources/jackclark"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/import-ai-470-no-rights-for-machines-automating-environment-generation-with-spade-and-building-be-g0nrxlp7g","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,llm,reinforcement-learning,ai-safety","timeRequired":"PT20M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Import AI ","item":"https://daily.dev/sources/jackclark"},{"@type":"ListItem","position":3,"name":"Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/import-ai-470-no-rights-for-machines-automating-environment-generation-with-spade-and-building-be-g0nrxlp7g#faq","mainEntity":[{"@type":"Question","name":"What is SPADE and how does it improve LLM training through self-play environment generation?","acceptedAnswer":{"@type":"Answer","text":"SPADE (Self-Play in Adaptive Synthetic Executable Environments) is a framework where an LLM alternates between an Environment Designer role, writing executable training environments as Python code, and a Reasoning Agent role that learns to solve them. Reward for the designer comes from the gap in agent performance with versus without privileged hints. Tested on Qwen3-4B, 8B, and 30B-A3B models, it improved the 30B model's game benchmark suite average by 8.1 points over baseline. Developers exploring RL-based agent training can follow frameworks like SPADE surfaced on daily.dev."}},{"@type":"Question","name":"How does the Hawkeye framework help AI coding agents write optimized GPU kernels for hardware like NVIDIA Blackwell and AMD MI350?","acceptedAnswer":{"@type":"Answer","text":"Hawkeye grounds autonomous kernel generation in a minimal taxonomy of unit tests, each pairing a human-authored solution kernel with a profiling metric, allowing coding agents to scale test-time compute effectively. Built by researchers from Harvard, Stanford, Together AI, and Caltech, it matches or exceeds torch.compile's expert-tuned libraries like cuBLAS and cuDNN, and reaches an 18.9x geomean speedup over Flash Linear Attention Triton kernels on emerging attention variants across Ampere, Hopper, Blackwell, and MI350 hardware. Engineers optimizing GPU kernels across hardware generations can track tools like Hawkeye via daily.dev."}},{"@type":"Question","name":"According to METR's research, where has AI actually accelerated scientific and technical progress?","acceptedAnswer":{"@type":"Answer","text":"AI has caused major acceleration in cybersecurity vulnerability discovery, with reported vulnerabilities across projects like cURL, OpenSSL, Firefox, and Microsoft rising sharply, alongside aggregate databases like the US NVD and OSV. Mathematics research shows minor, harder-to-measure acceleration despite doubled arXiv submissions in some areas. Optimization of AI research itself shows no measurable acceleration across benchmarks like CIFAR-10, nanoGPT, and Stockfish. Researchers gauging where AI genuinely moves the needle can follow studies like this on daily.dev."}}]} ```