{"date":"2026-11-19","title":"What happened in AI — Nov 19, 2026","generated_at":"2026-11-19T22:16:57Z","intro":["The tool layer moved out of the prompt today. Gemini now loads tool schemas on demand, LangSmith put credentials and permissions behind a gateway, and Vercel shipped eve 1.0.","Security evidence cut both ways. An audit found AgentDojo scored 433 real attacks as misses, while an autonomous Codex run reported 507 Webmin bugs for about $4,578."],"highlights":["AgentDojo scored 433 runs as no attack even when the agent paid the attacker; re-check security claims built on it.","Gemini API Tool Retrieval (preview): set defer_loading and add tool_search to load tool schemas only when needed.","An autonomous Codex campaign reported 507 reproduced Webmin vulnerabilities for about $4,578 in tokens.","Vercel's eve agent framework reached 1.0; LangSmith Tool Gateway entered public beta for governed tool calls.","agent-framework follow-up: PyPI quarantined Microsoft's publisher; the attacker may still have publish access.","DeepSeek dropped separate fees for model-triggered search and web fetch."],"article_count":23,"categories":[{"name":"Tool layers: load less, govern more","slug":"tool-layers-load-less-govern-more","summary":"Vendors moved tool handling out of the prompt: schemas load on demand, credentials sit behind a gateway, and output JSON gets compressed in a proxy.","articles":[{"title":"Introducing Tool Retrieval in the Gemini API","summary":"Preview in the Interactions API: add a tool_search tool with defer_loading true and Gemini fetches parameter schemas only when needed, cutting input tokens for large tool libraries.","source":"google_ai_blog","url":"https://blog.google/innovation-and-ai/technology/developers-tools/gemini-api-tool-retrieval/","published":"Thu, 19 Nov 2026 18:00:00 +0000"},{"title":"LangSmith Tool Gateway: Secure, governed tool use for agents","summary":"Public beta managed tool layer that holds credentials, decides which tools each agent may call and under whose identity, and logs every action.","source":"langchain_blog","url":"https://www.langchain.com/blog/langsmith-tool-gateway","published":"2026-11-19T18:00:00+00:00"},{"title":"Show HN: Wiretuple – Cut LLM structured-output tokens with positional tuples","summary":"Zero-dependency Go reverse proxy that rewrites repeated JSON keys into positional tuples, claiming 35–60% fewer output tokens for structured outputs and function calls.","source":"hackernews_ai","url":"https://github.com/Dimoncelli/wiretuple","published":"Thu, 19 Nov 2026 14:33:17 +0000"},{"title":"Discovery: Fail-closed authority checks for AI agent tool calls (v1.0-v1.1.8)","summary":"Solo project that refuses tool calls when untrusted prompt text tries to mint authority or replay receipts; publishes 3,315/3,315 verified receipt hashes.","source":"hackernews_ai","url":"https://github.com/2fastthompson/the-discovery","published":"Thu, 19 Nov 2026 16:57:48 +0000"}]},{"name":"Agent harnesses grow up: frameworks, state, sandboxes","slug":"agent-harnesses-frameworks-state-sandboxes","summary":"Agent runtimes are becoming infrastructure with GA labels, in-house standards, and durable state, not per-team experiments.","articles":[{"title":"Eve 1.0 – Agent framework from Vercel","summary":"Vercel's open-source agent framework hits GA after 230+ preview releases, adding approval controls for actions and an eval suite for how well coding agents build with it.","source":"hackernews_ai","url":"https://vercel.com/blog/eve-is-now-generally-available","published":"Thu, 19 Nov 2026 11:39:25 +0000"},{"title":"Aviator: HubSpot's Agent Harness","summary":"HubSpot explains why it built one internal agent harness in 2025 after teams kept making divergent model and architecture choices.","source":"hackernews_ai","url":"https://hubspot.engineering/blog/aviator-hubspots-agent-harness/","published":"Thu, 19 Nov 2026 16:42:34 +0000"},{"title":"Coordinated Snapshots of Agent State","summary":"Hatchet argues agent sandboxes need coordinated snapshots of state across processes, not just isolated compute.","source":"hackernews_ai","url":"https://hatchet.run/blog/agent-statefulness","published":"Thu, 19 Nov 2026 13:21:33 +0000"},{"title":"Docker AI Rebranded","summary":"Docker docs now lead with agent sandboxes: scheduled tasks run in a fresh cloud sandbox that is deleted when the task ends or times out.","source":"hackernews_ai","url":"https://docs.docker.com","published":"Thu, 19 Nov 2026 14:28:57 +0000"},{"title":"OSS Minymaton (event-driven MCP message bus for AI agents, bundled SQLite)","summary":"Durable, event-driven MCP message bus with bundled SQLite so several coding agents can hand off API changes and test failures without a human relaying them.","source":"hackernews_ai","url":"https://github.com/immutlex-dev/minymaton","published":"Thu, 19 Nov 2026 17:59:20 +0000"}]},{"name":"Evals that check what the agent did","slug":"evals-that-check-what-the-agent-did","summary":"Today's eval work audits actions and traces instead of trusting a model's reply text or one automated verdict.","articles":[{"title":"Did the attack work? Auditing AI-agent security benchmark verdicts","summary":"Audit of AgentDojo: the benchmark scored 433 runs as no attack even though the agent paid the attacker or took another decisive action, 11.9% of real attacks. Preprint, not peer reviewed.","source":"hackernews_ai","url":"https://zenodo.org/records/24321025","published":"Thu, 19 Nov 2026 16:47:14 +0000"},{"title":"Show HN: Agentest – offline tests for AI agents from OpenTelemetry traces","summary":"Open-source, offline suite of 165 deterministic tests over OpenTelemetry GenAI traces that flags unearned success claims, hidden-instruction follows, and forbidden tool calls with no LLM judge.","source":"hackernews_ai","url":"https://github.com/srpadmanabhuni/agentest","published":"Thu, 19 Nov 2026 14:22:18 +0000"},{"title":"LLM call is doing more than you think","summary":"Audiala swapped an LLM router for a decision model and rolled back in 23 minutes: the old call also resolved cities, references, and tool flags nobody evaluated.","source":"hackernews_ai","url":"https://seldon-ai.com/blog/your-llm-call-is-doing-more-than-you-think","published":"Thu, 19 Nov 2026 12:59:52 +0000"},{"title":"Genie One Accuracy: Why Curated Agents Need Governed Context","summary":"Databricks used 125 insurance questions to show agents need governed business definitions; Finance and Underwriting gave opposite, correct loss-ratio answers.","source":"databricks_blog","url":"https://www.databricks.com/blog/genie-one-accuracy-why-curated-agents-need-governed-context","published":"Thu, 19 Nov 2026 20:15:41 GMT"}]},{"name":"Security: AI finds bugs, supply chain still leaks","slug":"security-ai-finds-bugs-supply-chain-leaks","summary":"Autonomous auditing got cheap enough to run on a whole codebase while package and model supply chains stayed the soft spot.","articles":[{"title":"Autonomous AI Security Auditing of Webmin: 507 Newly Discovered Vulnerabilities","summary":"A fully autonomous Codex campaign reported 507 reproduced vulnerabilities in Webmin 2.670 using 8.35B tokens, about US$4,578 at standard API rates.","source":"hackernews_ai","url":"https://supuk.ch/papers/webmin-ai-security-audit-507-vulnerabilities","published":"Thu, 19 Nov 2026 12:27:53 +0000"},{"title":"Official Microsoft agent-framework PyPI package compromised","summary":"Follow-up: PyPI quarantined the Microsoft publisher because the attacker is believed to still have publishing access; treat any 1.26.1 or 1.26.2 install as a credential compromise.","source":"hackernews_ai","url":"https://opensourcemalware.com/blog/agent-framework-pypi-compromised","published":"Thu, 19 Nov 2026 04:29:50 +0000"},{"title":"Three breaches, endless leaks: Your AI supply chain needs canaries","summary":"Dragos argues for canary tokens across the AI supply chain so a leaked key or poisoned dependency trips an alarm when used.","source":"hackernews_ai","url":"https://www.dragos.com/blog/ai-supply-chain-security-canary-tokens","published":"Thu, 19 Nov 2026 16:51:00 +0000"}]},{"name":"Agents doing production engineering work","slug":"agents-doing-production-engineering-work","summary":"Large operators described agents that run inside their own engineering loops, from performance fixes to RL rollouts.","articles":[{"title":"How OpenAI’s Performance Factory keeps ChatGPT fast","summary":"OpenAI runs Codex-powered workflows that find and fix performance regressions to keep ChatGPT fast.","source":"openai_blog","url":"https://openai.com/index/how-openai-performance-factory-keeps-chatgpt-fast","published":"Thu, 19 Nov 2026 08:00:00 GMT"},{"title":"How Meta Uses AI Agents to Scale ML Innovations Across Ads Ranking Models","summary":"Meta uses agents to port proven ranking techniques across its many ads models, each with different architecture, data, and serving limits.","source":"hackernews_ai","url":"https://engineering.fb.com/2026/11/18/ml-applications/ai-agents-to-scale-ml-innovations-across-ads-ranking-models/","published":"Thu, 19 Nov 2026 16:06:28 +0000"},{"title":"How Abundant designed efficient rollout architecture with Modal Sandboxes","summary":"Abundant runs RL rollouts for frontier labs on Modal Sandboxes, giving each task its own isolated environment, Dockerfile, and verifier.","source":"modal_blog","url":"https://modal.com/blog/abundant-efficient-rollout-architecture-with-modal-sandboxes","published":"2026-11-19T00:00:00.000Z"}]},{"name":"Token prices and compute deals","slug":"token-prices-and-compute-deals","summary":"Chinese and Indian labs competed on price while compute commitments shifted in both directions.","articles":[{"title":"DeepSeek Removes Separate Fees for Model-Triggered Search and Web Fetch","summary":"DeepSeek no longer charges separately when the model triggers search or web fetch, which simplifies cost math for tool-using agents.","source":"search_cn_open_weight_labs","publisher_name":"Pandaily","publisher_domain":"pandaily.com","url":"https://news.google.com/rss/articles/CBMikgFBVV95cUxQakxYOFZROC1Xblc5S2hTUWdSb1ZOMVBNZmJ2YWt0TmRfY2tFc3hXUkIwZXAxNk5HUUZZNVRrMTlKR1FTMnBZd1F3X0ZRMTNnY3B2NmJSVW13QXJTMm5YeHlpNnY2a2JELS01WXFtU2szeXo4ZHVQcEh1YlJFUTlIZFY1b3FaSFNYekFZeFFpWkFpUQ?oc=5","published":"Thu, 19 Nov 2026 01:53:49 GMT"},{"title":"Sarvam Prices Its Own Model Below DeepSeek In New Token Factory","summary":"Sarvam's new token factory lists its own model below DeepSeek's price.","source":"search_cn_open_weight_labs","publisher_name":"Forbes","publisher_domain":"forbes.com","url":"https://news.google.com/rss/articles/CBMiuAFBVV95cUxQekxSVzdzd1ZVSF9pT3MwSTV1TkJnSnVEU19WVTlkblNCbXRZN3hvREVkdHdFeEtpODVfTW9oeFNsbTFBSE9xZ3lRLTlhbWJGUHpuMi1acHRKZ0xLdXg0bGVoa2NfSTlLU2djVmMxNDhzSG56UEt5anpxclUyS0RXZzFSR3AyVU8yM1N0LUE1UnhWUlV1LVUtb2lMejllaDFDWEQtcmplcFVyMlVoN3UtVjhaUVZDdl8z?oc=5","published":"Thu, 19 Nov 2026 12:31:02 GMT"},{"title":"Moonshot AI Cancels $150 Million Nvidia B300 GPU Rental Deal in Indonesia","summary":"Moonshot AI cancelled a $150 million Nvidia B300 GPU rental deal in Indonesia.","source":"search_cn_open_weight_labs","publisher_name":"Gadget Review","publisher_domain":"gadgetreview.com","url":"https://news.google.com/rss/articles/CBMiowFBVV95cUxNdVgtNlR6STVqOVNwSEZHZVVJZzVxb1MzbTZfVXdIcTI5Z1A4LUFlNjRKTmRZWS14eUFFOEhROGFnZkF2alAxamJiRDhJNnFXLS1jOFNBQVJmRGtDdEVoc3U2cktuYk9SQ3pfOURHT0V6QURDY09IQkJhMlhaeVRWMGJtVXdXaXVpd1R3TnVGLTNFZEp5R0hVdzcwQ016ajlHNVBV?oc=5","published":"Thu, 19 Nov 2026 18:14:24 GMT"},{"title":"NVIDIA and Microsoft Expand Accelerated Computing From Azure to Windows PCs","summary":"At Ignite, Microsoft widens access to Azure VMs on NVIDIA Vera Rubin NVL72 systems and extends the stack to Windows PCs.","source":"nvidia_blog","url":"https://blogs.nvidia.com/blog/microsoft-ignite-accelerated-computing-azure-windows-pcs/","published":"Thu, 19 Nov 2026 20:00:15 +0000"}]}],"video":{"youtube_id":"X0N_-ugxZBA","title":"LLM Digest Daily Explainer: Nov 19, 2026","duration_s":119.5}}