{"date":"2026-11-20","title":"What happened in AI — Nov 20, 2026","generated_at":"2026-11-20T22:16:32Z","intro":["AI providers leak what you send them. Dragos seeded about 2 million canary prompts and caught three provider compromises, resellers reading prompts and ignored Zero Data Retention settings.","Builders answered at the edge. New tools put coding agents on isolated workstations and microVMs, and a local scanner checks agent configs for live keys before a run."],"highlights":["Dragos canary test: three AI provider compromises and ignored Zero Data Retention settings across about 2M prompts.","CVE-2026-28400: a default-privilege container could corrupt Docker Desktop's VM disk through Docker Model Runner.","incus-spawn and Sandscale give coding agents their own Linux machine or Firecracker microVM, not your credentials.","Enola ties agent memory to code symbols so a later session re-checks notes against changed code.","Inkling 2 Mini loops 12 layers three times per token; vLLM serves it on day 0."],"article_count":19,"categories":[{"name":"Agent machines: isolate the box, not the user","slug":"agent-machines-isolate-the-box-not-the-user","summary":"Four launches move coding and computer-use agents off the developer's own desktop and credentials, onto separate VMs, headless sessions or simulated hardware.","articles":[{"title":"Isolated Linux workstations for autonomous AI agents","summary":"incus-spawn gives each coding agent its own isolated Linux workstation, so the agent never sees your API keys, GitHub token, ~/.ssh or other checked-out repos.","source":"hackernews_ai","url":"https://github.com/Sanne/incus-spawn","published":"Fri, 20 Nov 2026 19:04:23 +0000"},{"title":"Show HN: Sandscale – Firecracker microVMs for AI coding agents","summary":"Firecracker microVMs for agent-written code: start one in a single API call, pause and resume it; US East today, Frankfurt coming.","source":"hackernews_ai","url":"https://sandscale.dev","published":"Fri, 20 Nov 2026 15:29:29 +0000"},{"title":"Agent should not take the cursor","summary":"Field report on why computer-use agents that drive your real mouse and focus block half-hour tasks, and what it took to run them headless instead.","source":"hackernews_ai","url":"https://minea.minea.workers.dev/cursor","published":"Fri, 20 Nov 2026 09:46:45 +0000"},{"title":"Show HN: BSWorld – AI agents can now test Bluetooth code without hardware","summary":"Xiaomi's open-source simulator runs Zephyr, NuttX, NimBLE, BlueZ and other Bluetooth hosts deterministically in pytest, so agents and CI can test BLE code without radios.","source":"hackernews_ai","url":"https://github.com/XiaoMi/bsworld","published":"Fri, 20 Nov 2026 06:06:42 +0000"}]},{"name":"Coding-agent context: remember, then re-verify","slug":"coding-agent-context-remember-then-re-verify","summary":"Builders attacked stale and bloated context: memory tied to code symbols, org standards fed into generation, and page summaries instead of full DOM dumps.","articles":[{"title":"Keeping AI Coding Agent Memory Relevant as Code Changes","summary":"Enola links saved agent investigations to source symbols, so a later session can check whether the function it remembers has since moved or changed.","source":"hackernews_ai","url":"https://enola.tech/blog/agent-memory-needs-a-code-reference","published":"Fri, 20 Nov 2026 18:18:13 +0000"},{"title":"OutcomeOps: Context Engineering for AI-Assisted Development OSS GitHub Repo","summary":"Apache-2.0 platform that deploys into your own AWS account and grounds code generation and PR review in your ADRs, code-maps and architecture decisions.","source":"hackernews_ai","url":"https://github.com/bcarpio/outcome-ops-ai-assist","published":"Fri, 20 Nov 2026 12:59:28 +0000"},{"title":"Show HN: Frowse – Drop-in agent-browser CLI that cuts token spend by up to 90%+","summary":"agent-browser replacement that sends a compact page summary and resolves elements with a local CPU model (10–25 ms, under 250 MB RAM); claims up to 90% fewer tokens per step.","source":"hackernews_ai","url":"https://github.com/figranium/frowse","published":"Fri, 20 Nov 2026 14:50:44 +0000"},{"title":"Building the Agent Host Protocol for sessions that continue across devices","summary":"GitHub engineering post on the Agent Host Protocol, built so an agent session can continue when you switch devices.","source":"hackernews_ai","url":"https://github.blog/engineering/engineering-principles/building-the-agent-host-protocol-for-sessions-that-continue-across-devices/","published":"Fri, 20 Nov 2026 13:52:36 +0000"},{"title":"Humans.md: tell your AI agents how to work with you","summary":"Draft 0.3 spec for one ~/.humans/HUMANS.md file that coding agents read at session start: how to write to you, when to ask, when to interrupt.","source":"hackernews_ai","url":"https://humansmd.dev/","published":"Fri, 20 Nov 2026 05:42:30 +0000"}]},{"name":"Inference supply chain: what you send does not stay put","slug":"inference-supply-chain-what-you-send-does-not-stay-put","summary":"Evidence piled up that prompts, tokens and local model runners are attack surface, not neutral pipes.","articles":[{"title":"Canary Token Test Exposes AI Provider Breaches","summary":"Dragos planted about 2 million prompts with canary URLs, DNS names and AWS keys from July to November; it found three provider compromises, resellers reading prompts and ignored Zero Data Retention settings.","source":"hackernews_ai","url":"https://epyhia.com/dragos-canary-token-test-exposes-three-ai-provider-breaches-prompt","published":"Fri, 20 Nov 2026 13:20:06 +0000"},{"title":"Hostile Inference: Exploiting Docker Model Runner for Container Escapes","summary":"ZDI details CVE-2026-28400 in Docker Model Runner: an unauthenticated argument injection let a default-privilege container corrupt the Docker Desktop VM disk, and further work reached container escapes.","source":"hackernews_ai","url":"https://www.thezdi.com/blog/2026/11/17/hostile-inference-exploiting-docker-model-runner-for-container-escapes","published":"Fri, 20 Nov 2026 07:42:18 +0000"},{"title":"Token fraud is here and it’s BAD (YouTube)","summary":"Simon Willison links Claire Vo's video on token fraud from running ChatPRD, a live AI product since the GPT-4 era; treat paid LLM usage as an abuse target.","source":"simon_willison","url":"https://simonwillison.net/2026/Nov/20/token-fraud-is-here-and-its-bad/","published":"2026-11-20T02:49:57+00:00"}]},{"name":"Agent-setup hygiene: keys, hooks and root causes","slug":"agent-setup-hygiene-keys-hooks-and-root-causes","summary":"The fixes on offer are boring and local: scan the agent's config, clean up pasted keys, and read past incidents for the repeat causes.","articles":[{"title":"Show HN: Preflight-agent – five checks your AI-agent setup needs before it runs","summary":"Deterministic local scanner with five pre-run checks: auto-executing cloned folders, live keys in config, hooks that never fire, ambiguous agents and skill sprawl.","source":"hackernews_ai","url":"https://github.com/jasveerai88/preflight-agent","published":"Fri, 20 Nov 2026 15:14:03 +0000"},{"title":"Ultimate guide for fixing your team’s API keys","summary":"Google Cloud guide for teams whose Gemini prototypes started with an AI Studio key pasted into a coding agent, and how to clean that up without slowing prototyping.","source":"google_cloud_blog","url":"https://cloud.google.com/blog/topics/developers-practitioners/how-to-fix-your-teams-api-keys/","published":"Fri, 20 Nov 2026 17:00:00 +0000"},{"title":"AI incidents of 2026: Rogue agents, stolen keys, and other familiar stories","summary":"Aikido coded the stated root causes of 50 AI-related incidents disclosed in 2026 through Nov 19; many trace back to familiar problems such as stolen keys.","source":"hackernews_ai","url":"https://www.aikido.dev/blog/ai-incidents-root-causes-patterns","published":"Fri, 20 Nov 2026 09:42:13 +0000"}]},{"name":"Models: looped weights and provenance fights","slug":"models-looped-weights-and-provenance-fights","summary":"One new architecture shipped with day-0 serving, while labs and buyers argued over who copied, swapped or should avoid whose weights.","articles":[{"title":"Inkling 2 Mini on vLLM: Day-0 Support for a Looped Model","summary":"Thinking Machines' Inkling 2 Mini runs its middle 12 layers three times per token; 26B total / 3.5B active multimodal MoE, 1M context, served in vLLM on day 0 via vllm/vllm-openai:inkling-2-mini.","source":"vllm_blog","url":"https://vllm.ai/blog/2026-11-20-inkling2-mini","published":"Fri, 20 Nov 2026 00:00:00 GMT"},{"title":"Combatting adversarial distillation","summary":"OpenAI describes how it fights unauthorized distillation of its models' outputs into copies that ship without safeguards, citing cyber capabilities as one risk.","source":"openai_blog","url":"https://openai.com/index/combatting-adversarial-distillation","published":"Fri, 20 Nov 2026 10:00:00 GMT"},{"title":"Prime Intellect discloses unauthorized DeepSeek model swap that sharply reduced refusals","summary":"Report: Prime Intellect disclosed that a DeepSeek model was swapped without authorization, and the swap sharply reduced refusals.","source":"search_cn_open_weight_labs","publisher_name":"Superpower Daily","publisher_domain":"superpowerdaily.com","url":"https://news.google.com/rss/articles/CBMivwFBVV95cUxOYzRuZUFqYnV4WTU2dkZPSnFJX3dXRldmdG14aEJvQkg4UXlHTHVQQks5X0Y0MXh0U1phVGtyMW0yOTZfS2plemFiZ2FDRVhSRXdRU0h6aVlqZV9XcnN6cGRtd2pkYVRvUktaSmR5MDhBMWV2dzJfbVVDb2FscDc0TkNNQTNJeEVhcHRjT2VYNnlBXzc0Q0hGcENleTMwMXV6QmZSMDNCbFFGOEExVjM1NWZEYV9iVlQ0TWpCMU9Ncw?oc=5","published":"Fri, 20 Nov 2026 00:11:00 GMT"},{"title":"Cursor reportedly eyes the exit for Chinese open-weight AI models","summary":"Report: Cursor is weighing a move away from Chinese open-weight models.","source":"search_cn_open_weight_labs","publisher_name":"Digg","publisher_domain":"digg.com","url":"https://news.google.com/rss/articles/CBMiQkFVX3lxTE1yeUlIbmJod0ZnM3pkYzI4MG1GOGVBdmVVbUtLdUNoaFVJcE9WbDZ6STJTaWJ0YWtqNkdVaWhwUE8xQQ?oc=5","published":"Fri, 20 Nov 2026 20:34:56 GMT"}]}],"video":{"youtube_id":"2Lq1WoDQFJw","title":"LLM Digest Daily Explainer: Nov 20, 2026","duration_s":119.8}}