📡 AI Digest — 2026-05-09#
🔥 トレンド（Hacker News）#
- A New Engineering Metric for the AI-Transformation Age [Hacker News] https://www.webwire.com/ViewPressRel.asp?aId=354394
- A New Era of Security: Frontier AI Defense [Hacker News] https://www.paloaltonetworks.com/blog/2026/05/frontier-ai-defense/
- AI Is Distorting Practically Everything About the Economy [Hacker News] https://www.wsj.com/tech/ai/ai-is-distorting-practically-everything-about-the-economy-4ca6fcff
- AI’s Big Messaging Pivot [Hacker News] https://www.noahpinion.blog/p/ais-big-messaging-pivot
- AI, the Poor, and the Ignorant [Hacker News] https://user8.bearblog.dev/ai-the-poor-and-the-ignorant/
- Aesthetic Layout in LLM-Based Slide Generation via Verifiable Rewards [Hacker News] https://arxiv.org/abs/2604.22840
- AniTroves – An anime database with a custom LLM-based discovery hub [Hacker News] https://news.ycombinator.com/item?id=48057592
- Ask HN: How do we handle the rise of low quality “This is LLM” comments? [Hacker News] https://news.ycombinator.com/item?id=48063759
- Cloudflare’s slowing growth disappoints investors betting on AI boost [Hacker News] https://www.reuters.com/business/cloudflares-slowing-growth-disappoints-investors-betting-ai-boost-2026-05-08/
- Debt Behind the AI Boom: A Large-Scale Study of AI-Generated Code in the Wild [Hacker News] https://arxiv.org/abs/2603.28592
- Frontier models refuse to help organizers, so we built our own activist AI [Hacker News] https://www.outcryai.com/research/how-to-create-activist-ai
- How LLM Inference Works [Hacker News] https://twitter.com/akshay_pachaar/status/2050941458614751327
- I’ve been using Empirical as my memory layer across AI tools [Hacker News] https://empirical.gauzza.com/blog/codex-session-tone-voice-how-i-used-codex-empirical-to-lock-in-my-writing-voice/
- IatroBench: Pre-Registered Evidence of Iatrogenic Harm from AI Safety Measures [Hacker News] https://arxiv.org/abs/2604.07709
- LLM Inference Series: 4. KV caching, a deeper look [Hacker News] https://medium.com/@plienhar/llm-inference-series-4-kv-caching-a-deeper-look-4ba9a77746c8
- LLM-driven security reports disrupt coordinated disclosure [Hacker News] https://lwn.net/SubscriberLink/1070698/708a56108d2a9e2e/
- Linux 7.1: Kicinski Called It ‘LLM-Pocalypse.’ Then Deleted 138,000 Lines [Hacker News] https://canartuc.medium.com/linux-7-1-kicinski-called-it-llm-pocalypse-then-deleted-138-000-lines-afa3cb6136dc
- Mesh LLM to build private personal AI, using open models [Hacker News] https://www.anarchai.org
- Meta plans advanced ‘agentic’ AI assistant for users [Hacker News] https://www.reuters.com/business/meta-plans-advanced-agentic-ai-assistant-users-ft-reports-2026-05-05/
- NL Autoencoders Produce Unsupervised Explanations of LLM Activations [Hacker News] https://transformer-circuits.pub/2026/nla/
- Novel macro signals for AI-related job loss? [Hacker News] https://news.ycombinator.com/item?id=48066573
- Pagaya investor seeks refund, claiming AI real estate fund wiped 80% of capital [Hacker News] https://www.calcalistech.com/ctechnews/article/3a0p63mof
- Pentagon will ’never again’ rely on a single AI provider, official says [Hacker News] https://www.nextgov.com/artificial-intelligence/2026/05/pentagon-will-never-again-rely-single-ai-provider-official-says/413399/
- Phishing Arena – multi-agent LLM tournament to study adversarial email security [Hacker News] https://github.com/Krabby24/phishing-arena
- SafeSandbox – infinite undo for AI coding agents (Cursor, Claude Code, Codex) [Hacker News] https://github.com/Baukaalm/safesandbox
- Show HN: Describe what makes a photo “bad” and let a local LLM flag them [Hacker News] https://github.com/iamnotagentleman/bad-photos-out
- Show HN: Openloom – Turn Loom links into transcripts and frames an LLM can watch [Hacker News] https://www.useopenloom.com/
- Show HN: Tokid – Token-native IDs for LLM-facing systems [Hacker News] https://github.com/Tetra-Research/tokid
- Show HN: UltraCompress – first mathematically lossless 5-bit LLM compression [Hacker News] https://github.com/sipsalabs/ultracompress
- Show HN: When the LLM Accidentally [Hacker News] https://news.ycombinator.com/item?id=48059025
- SubQ: Sub-quadratic LLM built for 12M-token reasoning [Hacker News] https://subq.ai/
- Subquadratic LLM 12 million tokens ctx [video] [Hacker News] https://www.youtube.com/watch?v=kEiuMnosuxU
- Tell HN: Prayer with AI models is suboptimal [Hacker News] https://news.ycombinator.com/item?id=48068661
- The Abstraction Fallacy: Why AI Can Simulate but Not Instantiate Consciousness [Hacker News] https://philarchive.org/rec/LERTAF
- Trump jumps from ‘anything goes’ to ‘strict regulation’ AI policy [Hacker News] https://www.theregister.com/ai-and-ml/2026/05/08/trump-jumps-from-anything-goes-to-strict-regulation-ai-policy/5234687
- Trusted Remote Execution: Policy-Enforced Scripts for AI Agents and Humans [Hacker News] https://aws.amazon.com/blogs/opensource/introducing-trusted-remote-execution-policy-enforced-scripts-for-ai-agents-and-humans/
- Why LLM Outputs Need a Deterministic Evaluation Layer [Hacker News] https://avectic.com/engineering/deterministic-evaluation-layer
- Why do LLM outputs get worse even when metrics stay stable? [pdf] [Hacker News] https://huggingface.co/datasets/realitydriftproject/ai-drift-detection-frameworks/blob/main/llm-drift-detection-why-ai-outputs-degrade-without-errors.pdf
- Windows 11 Registry mod blocks automatic download of 4GB AI model on Chrome [Hacker News] https://www.neowin.net/news/official-windows-11-registry-mod-blocks-automatic-download-of-4gb-ai-model-on-google-chrome/
- Yes, I set up Karpathy’s LLM wiki. Now what? [Hacker News] https://twitter.com/keane42443/status/2052426761477255448
💬 Reddit (MachineLearning / Artificial)#
- [D] Self-Promotion Thread [Reddit (↑9)] https://www.reddit.com/r/MachineLearning/comments/1t1d2m0/d_selfpromotion_thread/
- [D] Monthly Who’s Hiring and Who wants to be Hired? [Reddit (↑8)] https://www.reddit.com/r/MachineLearning/comments/1t0gle1/d_monthly_whos_hiring_and_who_wants_to_be_hired/
- People Interested in Continual Learning Research[R] [Reddit (↑65)] https://www.reddit.com/r/MachineLearning/comments/1t72u1r/people_interested_in_continual_learning_researchr/
- Getting harassed by an aggressive “independent researcher” demanding very specific citations and phrasing in my paper [D] [Reddit (↑107)] https://www.reddit.com/r/MachineLearning/comments/1t6vvjc/getting_harassed_by_an_aggressive_independent/
- Disillusionment with mechanistic interpretability research [D] [Reddit (↑46)] https://www.reddit.com/r/MachineLearning/comments/1t6zdj6/disillusionment_with_mechanistic_interpretability/
- Formalizing statistical learning theory in Lean 4 [R] [Reddit (↑10)] https://github.com/Robby955/FormalSLT
- Interactive KL Divergence Visualisation [P] [Reddit (↑2)] https://www.reddit.com/r/MachineLearning/comments/1t7lmu4/interactive_kl_divergence_visualisation_p/
- Embedding models for time series data [D] [Reddit (↑4)] https://www.reddit.com/r/MachineLearning/comments/1t7avgp/embedding_models_for_time_series_data_d/
- Backcasting forecast errors: model collapsing to mean [P] [Reddit (↑0)] https://www.reddit.com/r/MachineLearning/comments/1t7knyd/backcasting_forecast_errors_model_collapsing_to/
- Steam Similarity Recommender [P] [Reddit (↑20)] https://www.reddit.com/gallery/1t6x2zw
- Marc Andreessen Mocked for Accidentally Revealing That He Seems to Have a Deep Misunderstanding of How AI Actually Works [Reddit (↑637)] https://futurism.com/artificial-intelligence/marc-andreessen-mocked-ai-works
- I like ChatGPT, I like AI [Reddit (↑10)] https://i.redd.it/bd1pzq8mczzg1.jpeg
- New AI model spots pancreatic cancer up to 3 years earlier than human doctors in test [Reddit (↑18)] https://www.livescience.com/health/cancer/new-ai-model-spots-pancreatic-cancer-up-to-3-years-earlier-than-human-doctors-in-test
- CFS - Conditional Field Subtraction [Reddit (↑6)] https://medium.com/@mauro.dev/cfs-conditional-field-subtraction-43a3c4eb80f4
- AI tooling is starting to feel like PC modding culture [Reddit (↑6)] https://www.reddit.com/r/artificial/comments/1t7gfud/ai_tooling_is_starting_to_feel_like_pc_modding/
- AMD’s local, open-source AI can now easily interact with your Gmail [Reddit (↑11)] https://www.phoronix.com/news/AMD-GAIA-Gmail-Integration
- Compiled every national AI strategy in Asia — Vietnam has the most comprehensive standalone law, Japan has no penalties, Korea just eliminated Naver from sovereign LLM competition for using Qwen weights [Reddit (↑2)] https://www.reddit.com/r/artificial/comments/1t7h9gt/compiled_every_national_ai_strategy_in_asia/
- Inside the AI Sweatshops Powering ChatGPT [Reddit (↑4)] https://www.reddit.com/r/artificial/comments/1t7aejp/inside_the_ai_sweatshops_powering_chatgpt/
- TRANSMISSION LOG — UNVERIFIED SOURCE [Reddit (↑0)] https://www.reddit.com/r/artificial/comments/1t7ll5t/transmission_log_unverified_source/
📰 ニュース#
- Cloudflare says AI made 1,100 jobs obsolete, even as revenue hit a record high [TechCrunch AI] https://techcrunch.com/2026/05/08/cloudflare-says-ai-made-1100-jobs-obsolete-even-as-revenue-hit-a-record-high/
- Live only at TechCrunch Disrupt 2026: Why most founders are already behind on raising a Series A in 2027 [TechCrunch AI] https://techcrunch.com/2026/05/08/live-only-at-techcrunch-disrupt-2026-why-most-founders-are-already-behind-on-raising-a-series-a-in-2027/
- The biggest US power grid is under strain from AI — and no one is happy [TechCrunch AI] https://techcrunch.com/2026/05/08/the-biggest-u-s-power-grid-is-under-strain-from-ai-and-no-one-is-happy/
- Airbnb says AI now writes 60% of its new code [TechCrunch AI] https://techcrunch.com/2026/05/08/airbnb-says-ai-now-writes-60-of-its-new-code/
Generated at 2026-05-09 07:00 JST by Emma 🍻