{"date":"2026-11-10","title":"What happened in AI — Nov 10, 2026","generated_at":"2026-11-10T22:21:57Z","intro":["Agent safety moved to enforcement points: tools that intercept commands, audit access and test database roles shipped alongside NVIDIA donating OpenShell to the CNCF.","Google made Gemini 4 Carbon generally available and DeepMind formally verified Signal's protocol code with agents, while the Fireworks AI breach spread to customer data."],"highlights":["Gemini 4 Carbon is generally available, with a new multimodal agent benchmark (MKA) alongside it.","DeepMind's AlphaProof Code formally verified security properties of Signal's real protocol code.","NVIDIA is donating its OpenShell agent-governance runtime to the CNCF.","A read-only Postgres role does not stop every harmful agent query; a reproducible test shows which ones.","Fireworks AI's breach now includes billing data and some customer secrets; Notion is notifying users.","Epoch: GB300 has 25x H100's agentic inference capacity on only 2.5x the FLOP/s."],"article_count":44,"categories":[{"name":"Guardrails on What Agents Can Touch","slug":"guardrails-agent-access","summary":"Builders are moving agent safety from prompts to enforcement points: database roles, command interceptors, access audits and a CNCF-bound agent runtime.","articles":[{"title":"What a read-only Postgres role does and doesn't stop when an AI agent connects","summary":"A reproducible PostgreSQL 17.11 run shows which harms a read-only role blocks and which session settings and reads still get through.","source":"hackernews_ai","url":"https://vericto.com/blog-post-postgres-read-only-role-ai-agents","published":"Tue, 10 Nov 2026 17:21:23 +0000"},{"title":"Antelog – stops AI coding agents from running dangerous commands","summary":"Local interceptor that checks each shell command a coding agent proposes and pauses destructive ones, like deleting a production volume, for approval.","source":"hackernews_ai","url":"https://antelog.dev/","published":"Tue, 10 Nov 2026 07:57:26 +0000"},{"title":"I Access Auditor – see and revoke what your AI coding agents can access","summary":"Desktop app that inventories what Claude Code, Codex, Gemini CLI, Cursor and OpenCode can access, and lets you revoke it.","source":"hackernews_ai","url":"https://github.com/dirmedmr-alt/ai-access-auditor","published":"Tue, 10 Nov 2026 08:31:13 +0000"},{"title":"NVIDIA Expands Open Source Commitment With OpenShell Donation and New AI Infrastructure","summary":"NVIDIA is donating OpenShell, its runtime for securing and governing autonomous agents, to the CNCF as a Sandbox project.","source":"nvidia_blog","url":"https://blogs.nvidia.com/blog/open-source-openshell-infrastructure/","published":"Tue, 10 Nov 2026 16:00:03 +0000"},{"title":"Show HN: Fix Gate – verifies security-fix PRs, incl. AI-written ones","summary":"Rule-based GitHub Action that checks a security-fix PR: finding gone, nothing new, files parse, tests not newly failed; unrun checks never count as passes.","source":"hackernews_ai","url":"https://github.com/IWEBai/fix-gate-action","published":"Tue, 10 Nov 2026 18:26:35 +0000"}]},{"name":"Agent Experience: Docs and APIs Now Have an Agent Audience","slug":"agent-experience","summary":"Agent traffic to docs is now large enough to measure, and vendors are starting to design SDKs and docs for the agent that picks them.","articles":[{"title":"Measure agent-traffic for any site","summary":"Open-source tool from Mintlify to measure how much of a site's traffic is agents, which ones, and whether a GPTBot request is really OpenAI.","source":"hackernews_ai","url":"https://github.com/mintlify/agent-traffic","published":"Tue, 10 Nov 2026 19:29:46 +0000"},{"title":"What Is Agent Experience (Ax)?","summary":"Microsoft defines AX: an agent picking a rival service or calling a deprecated API makes your product look broken, so measure it.","source":"hackernews_ai","url":"https://developer.microsoft.com/blog/what-is-agent-experience-ax/","published":"Tue, 10 Nov 2026 20:53:00 +0000"},{"title":"Designing for Agents — Zeno Rocha, Resend","summary":"Resend's Zeno Rocha on where answer-engine optimization ends and agent experience begins, from docs Claude Code reads to tools it calls.","source":"latent_space","url":"https://latent.space/p/resend/","published":"Tue, 10 Nov 2026 20:05:17 GMT"},{"title":"Show HN: Get your coding agent off ~2019 JavaScript, onto modern web primitives","summary":"Agent skill that steers coding agents off legacy JavaScript patterns toward native browser primitives like CSS.highlights and dialog.","source":"hackernews_ai","url":"https://github.com/GoogleChrome/modern-web-guidance","published":"Tue, 10 Nov 2026 18:01:15 +0000"}]},{"name":"Verifying and Operating Agent-Built Software","slug":"verifying-operating-agents","summary":"The verification bar for agent output is rising, from formal proofs of real protocol code to evidence that messy codebases derail coding agents.","articles":[{"title":"Introducing AlphaProof Code: a breakthrough in verifying complex software at depth","summary":"DeepMind's agentic framework formally verified key security properties of the actual Signal protocol code.","source":"google_deepmind_blog","url":"https://deepmind.google/blog/introducing-alphaproof-code/","published":"Tue, 10 Nov 2026 12:02:28 +0000"},{"title":"Google Finds that Technical Debt Can Derail Coding Agents Through \"Agent Distraction\"","summary":"Google research finds technical debt causes \"agent distraction\" that throws coding agents off task.","source":"infoq_ai_ml","url":"https://www.infoq.com/news/2026/11/google-agent-distraction/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Tue, 10 Nov 2026 18:00:00 GMT"},{"title":"Article: LLM Reliability Layers Under Test: Accuracy, Yield, and the Limits of Verification","summary":"InfoQ tests LLM reliability layers on accuracy and yield, and maps where verification stops helping.","source":"infoq_ai_ml","url":"https://www.infoq.com/articles/llm-reliability-layers-limits-accuracy/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Tue, 10 Nov 2026 09:00:00 GMT"},{"title":"How Workday Built an Enterprise-Scale AI Agent Platform with LangGraph and LangSmith","summary":"Workday's internal agent platform uses LangGraph for orchestration and LangSmith for tracing and evals, targeting production agents by end of 2026.","source":"langchain_blog","url":"https://www.langchain.com/blog/how-workday-built-an-enterprise-scale-ai-agent-platform-with-langgraph-and-langsmith","published":"2026-11-10T16:00:00+00:00"}]},{"name":"Frontier and Small Model Releases","slug":"model-releases","summary":"Three labs shipped on the same news cycle: Google's top tier, Anthropic's Fable 5.2, and Meta's smallest Muse model.","articles":[{"title":"Gemini 4 Carbon is now available","summary":"Google's strongest model is now generally available, pitched on cybersecurity and long-horizon agentic work, after a two-week defender-only preview.","source":"google_deepmind_blog","url":"https://deepmind.google/blog/gemini-4-carbon-is-now-available/","published":"Tue, 10 Nov 2026 16:03:47 +0000"},{"title":"The Multimodal Knowledge Agent Benchmark","summary":"New Google and Artificial Analysis benchmark for agents that must reason over visually complex files, not just text.","source":"hackernews_ai","url":"https://blog.google/innovation-and-ai/models-and-research/gemini-models/multimodal-knowledge-agent-benchmark/","published":"Tue, 10 Nov 2026 16:12:21 +0000"},{"title":"[AINews] Fable 5.2 retakes World #1 LLM spot as $ANT IPO roadshow begins","summary":"Latent Space covers Fable 5.2 and Mythos 5.2 retaking the top LLM spot as Anthropic's IPO roadshow opened.","source":"latent_space","url":"https://latent.space/p/ainews-fable-5-2-retakes-world-1-llm-spot-as-ant-ipo-roadshow-begins/","published":"Tue, 10 Nov 2026 06:34:44 GMT"},{"title":"Introducing Our Smallest, Most Efficient Model: Muse Ember","summary":"Meta's small, efficient model enters the Haiku/Luna tier, per Simon Willison's notes.","source":"simon_willison","url":"https://simonwillison.net/2026/Nov/10/muse-ember/","published":"2026-11-10T07:28:21+00:00"},{"title":"Kimi K3.1 Pro Coder Arrives at 775 GB, Built to Fit One GPU","summary":"Moonshot's coding model ships as a 775 GB release aimed at fitting on a single GPU.","source":"search_cn_open_weight_labs","publisher_name":"Pasquale Pillitteri","publisher_domain":"pasqualepillitteri.it","url":"https://news.google.com/rss/articles/CBMihAFBVV95cUxPQWVlRVJORHdHUnE4R3pnYm01NzJVNEtrRUdNMUxpdS02azZVS3hzSE5ERjRBVXJzOVZMa1FEajJuRVhhUF9menUybndPdzJ2MXo1UFlZUl9sVTdBLUhWVkgtaHhVMHk2SlNMem1GbVNHaDZEak9ReG02UmY3X0trd1phMWQ?oc=5","published":"Tue, 10 Nov 2026 14:31:32 GMT"}]},{"name":"Inference Economics: Memory, State and Mixed Hardware","slug":"inference-infrastructure","summary":"Inference gains now come from memory, state reuse and plumbing more than raw FLOP/s.","articles":[{"title":"Beyond FLOP/s: comparing AI hardware for inference by memory and communication","summary":"Epoch estimates GB300 delivers about 25x H100's agentic inference capacity per chip with only 2.5x the 8-bit FLOP/s.","source":"hackernews_ai","url":"https://epoch.ai/publications/inference-performance-index","published":"Tue, 10 Nov 2026 12:09:19 +0000"},{"title":"FlashStates Unlocking Flexible Hybrid State Management for Agentic LLM Inference","summary":"Together AI and collaborators decouple recurrent checkpoints from KV-cache blocks so hybrid models can resume agent sessions without dense state storage.","source":"hackernews_ai","url":"https://flashstates.ai/","published":"Tue, 10 Nov 2026 04:29:35 +0000"},{"title":"CUDA+ROCm dual-back ends LLM inference for mixed-vendor GPUs (up to 48% faster)","summary":"Patch splits GLM MoE inference across an RTX 5090 and a Radeon iGPU, CUDA for attention and HIP for experts, reporting up to 48% faster.","source":"hackernews_ai","url":"https://github.com/mw00/project-maya/pull/879","published":"Tue, 10 Nov 2026 02:05:11 +0000"},{"title":"Show HN: QueenMQ – slow LLM calls blocked our Kafka partitions, so we built it","summary":"Apache-2.0 message broker built after slow LLM calls blocked Kafka partitions; it has mostly replaced Kafka at its authors' company.","source":"hackernews_ai","url":"https://github.com/queen-mq/queen","published":"Tue, 10 Nov 2026 14:52:36 +0000"}]},{"name":"Inference Vendor Breach Widens","slug":"inference-vendor-breach","summary":"The Fireworks AI incident now reaches customer data, and downstream apps are notifying their own users.","articles":[{"title":"Fireworks.ai Update on October 2026 security incident","summary":"Fireworks now reports impact to account and billing data and, for some accounts, secrets, datasets and incomplete models.","source":"hackernews_ai","url":"https://fireworks.ai/security/bulletins/october-2026-security-incident-update","published":"Tue, 10 Nov 2026 08:47:00 +0000"},{"title":"Tell HN: Notion AI chat sessions have been exposed via Fireworks AI hack","summary":"Notion is emailing users that AI chat sessions may be exposed through its vendor Fireworks AI.","source":"hackernews_ai","url":"https://news.ycombinator.com/item?id=50436636","published":"Tue, 10 Nov 2026 06:00:48 +0000"}]}],"video":{"youtube_id":"v_q0l1IPNC4","title":"LLM Digest Daily Explainer: Nov 10, 2026","duration_s":116.8}}