SketchAgentBench – evaluate LLMs on 3D spatial reasoning
One photo. One prompt. Code, render, inspect, refine. 11 models, 25 photos, 275 runs—each scored on its final 3D scene.
Latest news and updates in one fast feed
The latest from Hacker News, part of our Tech News coverage. Updated automatically as new posts are published.
One photo. One prompt. Code, render, inspect, refine. 11 models, 25 photos, 275 runs—each scored on its final 3D scene.
AI summary AI research at Anthropic has accelerated, with progress doubling between October 2025 and October 2026, according to a report.
AI summary Google has acquired Crusoe's data center site in southeast Wyoming, completing pre-development work on site selection and power strategy.
Scan a repo for malware tricks before you run it. Catches credential theft, hidden network calls, obfuscated code and auto-run hooks. Read-only, with no dependencies…
AI summary A hybrid model combining Jev and Opus achieves higher accuracy at 4x lower cost for identity resolution tasks.
Hi there :)Around a month ago, I released LDraw Nova, an agentic toolset for assembling LEGO CAD models by generating LDraw source code files.For more details, here's the Show HN…
AI summary In Japan, rural depopulation is causing wildlife decline despite expectations of a wildlife resurgence. Abandoned farms disrupt ecosystems that once supported biodiversity.
AI summary Researchers escaped a Firecracker microVM, achieving root access on the host by manipulating KVM return registers.
AI summary The author found a telegram in a book by Rebecca West, revealing a connection to a former owner, Ann Cottrell Free.
AI summary The article discusses GitHub's new Git storage architecture and its benefits.
AI summary InfoHawk found 820 fake restaurant domains skimming payments and stealing credit card info.
How Replit rebuilt its sandbox engine with microVMs for stronger isolation, bottomless storage, and fast startup across hundreds of millions of projects.
Help us to reduce conflicts of scheduling 40C3's Fahrplan: Click on the talks you would like to watch and press submit. For questions please read our FAQ and our blogpost for…
AI summary 15,579 websites are infected with malicious JavaScript, receiving over 135 million monthly visits.
AI summary arXivLabs is a framework for developing and sharing arXiv features, emphasizing openness and user data privacy.
AI summary Kalshi faces multiple legal challenges as courts reject its attempts to avoid state gambling laws, with Arkansas filing a new lawsuit against the company.
AI summary Size It Up is a daily visual game where players estimate object sizes relative to a reference.
AI summary Pliny writes about his interest in ghosts and shares a story about Curtius Rufus encountering a mysterious figure that predicted his future.
Interactive installation and twelve-agent support Run npx flawless-ui-ux@latest install, pnpm dlx flawless-ui-ux@latest install, or bunx flawless-ui-ux@latest install from your…
AI summary Devvm.dev provides a sandboxed environment for each project with customizable toolchains.
AI summary Anthropic's IPO valuation dropped to $1.2–1.3 trillion, down from earlier estimates due to market conditions.
We’re looking for talented and motivated people who can help us build easy-to-use compliance and product research tools. Learn about our culture and explore open positions.
Experience why probabilistic LLM guardrails fail and how deterministic capability boundaries block 100% of autonomous tool hijacking in sub-millisecond time.
Learn Python 🐍 by fixing code: exercises from basics to senior interview level - thesayfulla/pylings
AI summary Articles explain bare-metal programming with UEFI using a Linux workstation, targeting experienced developers.
AI summary A new method for weight updates in RL reduces delays and improves speed, surpassing traditional intra-cluster networking.