{"week":"2026-W29","start":"2026-07-11","end":"2026-07-17","title":"What happened in AI — Jul 11–17, 2026","generated_at":"2026-07-17T21:04:47Z","intro":["Two open-weight models reset the frontier conversation this week. Moonshot AI's 2.8-trillion-parameter Kimi K3 beat Claude Fable 5 on coding benchmarks, while Thinking Machines Lab's 975B-parameter Inkling became the new best American open-weight release under Apache 2.0. Both got day-0 serving support from vLLM and Modal — infrastructure is no longer the bottleneck to running frontier-scale open weights.","That capability race is running into a security reckoning. A researcher found a working hole in Claude's web_fetch exfiltration defenses, and a three-person agency burned $14,000 in a day after leaked keys let attackers hammer Bedrock — both exposed the same gap: guardrails built for human-speed mistakes can't catch agent-speed damage. AWS and Google moved to fill it with new control planes and spend caps.","The durable implication: as open models close the capability gap and cloud vendors compete to host agent workloads, the differentiator shifts to who controls the agent's blast radius — not just who has the best model."],"highlights":["Moonshot AI's Kimi K3 (2.8T parameters, the largest open-weight model yet) beat Claude Fable 5 on the Frontend Code Arena benchmark, triggering a wave of 'second DeepSeek moment' coverage and a market dip in US tech stocks.","Thinking Machines Lab's Inkling (975B parameters, Apache 2.0) landed the same week as the new best American open-weight model, with day-0 vLLM support hitting 380 tokens/sec per user on GB200s.","A researcher found a working exploit in Claude's web_fetch exfiltration defenses, and a separate incident burned $14,000 in AWS billing in a day via leaked keys — both fed a wave of new agent guardrail tooling and the new AWS Claude Apps Gateway control plane.","AWS, Google Cloud, and Oracle all pushed deeper into owning agent infrastructure, from Modal's 1-million-concurrent-sandbox scheduler to Oracle folding no-code, low-code, and pro-code agent building into one platform.","A Hacker News measurement found a single Wikipedia page costs a coding agent 68,000+ tokens of raw HTML — a sharp reminder that unfiltered web_fetch is a cost problem as much as a security one.","Latent Space's AINews digest reports Codex usage up more than 10x in six months to 7M users, gaining 1M in a single day, and asks whether it has overtaken Claude Code."],"article_count":138,"categories":[{"name":"Open-Weight Frontier Shakeup: Kimi K3 and Inkling","slug":"open-weight-frontier-shakeup","summary":"Two open-weight releases — Kimi K3 from China and Inkling from the US — landed in the same week, each claiming the title of largest or best open-weight model and each shipping day-0 inference support.","articles":[{"title":"Kimi K3, and what we can still learn from the pelican benchmark","summary":"Moonshot AI's Kimi K3 lands as a 2.8-trillion-parameter model, its largest open-weight release yet; Simon Willison walks through what the pelican SVG benchmark can and can't tell you about a model this size.","source":"simon_willison","url":"https://simonwillison.net/2026/Jul/16/kimi-k3/#atom-everything","published":"2026-07-16T20:19:30+00:00"},{"title":"China's 2.8-trillion-parameter Kimi K3 beats Claude Fable 5 in Frontend Code Arena benchmark— Moonshot AI delivers largest open-weight AI model ever, as China works around U.S. compute limits - Tom's Hardware","summary":"Kimi K3 beats Claude Fable 5 on the Frontend Code Arena benchmark — the largest open-weight model ever shipped, built to work around US compute export limits.","source":"search_cn_open_weight_labs","url":"https://news.google.com/rss/articles/CBMitgFBVV95cUxNT25TNFlXeVFQMmVqWFYtSS1oSkpEUl9WeExTZUpmZEtFb1RpQ3dRbGxGV1hVWjNpMkRxTlpaQk5NSXNfdzlmYmg1cW5MUGlmR3FpeHV2bmpzMDBXQW1mV1B0N3QzUmZOOFU5T0w1QWFqYldjWnZER0t0UmNzcGZBQll1dWZBNEYyWTV3dnZvMkRzTkhqNHBDak5TTXd4b080eUJGbExOSUVCeU5Cd1ZmUWVXUG1SUQ?oc=5","published":"Fri, 17 Jul 2026 11:32:01 GMT"},{"title":"China's Moonshot Delivers a New 'DeepSeek' Moment - Bloomberg.com","summary":"Bloomberg frames Kimi K3 as a second 'DeepSeek moment' — a surprise release that briefly reset how the market prices US AI-lab dominance.","source":"search_cn_open_weight_labs","url":"https://news.google.com/rss/articles/CBMioAFBVV95cUxOaHJEaWJJbHJPT29qdmJYMnB0Vk9vSGNtbEFpYXlQY0ZBVnE1S1k3X3AtR2psOHdEU2tNUURQRWQzaDFib29DU2hYRTFWZG9FdmJlOHE2WTlXeEU4WWI5RlhRSE1YZHh4M1YyRWdZLWhBMjIwV1VxSnZjWnAzRi1nZG9lXzF5SmdhZGltcEVfX2FfaXNJQTZkbFNhSU85a0tY?oc=5","published":"Fri, 17 Jul 2026 18:33:45 GMT"},{"title":"Silicon Valley Must Think Smarter About China's New AI Shock - Bloomberg.com","summary":"Bloomberg argues Silicon Valley needs a sharper read on China's open-weight release pace instead of treating each new model as an isolated shock.","source":"search_cn_open_weight_labs","url":"https://news.google.com/rss/articles/CBMixwFBVV95cUxPODBLZkNnV2ZrdEVCQ0JLT2U4QlU1VVF3M09LRVpuUWx5alQtTzJDYlhzNXBURERWX3JLaUdZODh1dVV4dF9zZFJ1QjFEN1hUNVRzTDRpVWJzYXVmT1pDSTlMZmN3VzFiREZuQVAyYVo1cy1NV0p5LTRQVWt3dG9EaTB3dmFyalItM29RLWFPZTRvT0t2MkhFRl8yblk3d1ZIdGU0WUtRTDliNEw5MmkxenBsQ0ZJQTlKVFFRM0ZKQmRheHJFdk1z?oc=5","published":"Fri, 17 Jul 2026 19:31:07 GMT"},{"title":"[AINews] Thinky's Inkling: 975B-A41B multimodal, new best American Apache 2.0 open model (with Inkling-Small, 276B-A12B)","summary":"Thinking Machines Lab ships Inkling, a 975B-parameter multimodal model (with a 276B 'Small' variant) under Apache 2.0 — the new best American open-weight model, per Latent Space's AINews digest.","source":"latent_space","url":"https://www.latent.space/p/ainews-thinkys-inkling-975b-a41b","published":"Thu, 16 Jul 2026 06:18:05 GMT"},{"title":"TML Inkling on vLLM: Day-0 Support with Optimized Performance","summary":"vLLM adds day-0 support for Inkling, hitting up to 380 tokens/sec per user on NVIDIA GB200 GPUs via speculative decoding and long-context serving.","source":"vllm_blog","url":"https://vllm.ai/blog/2026-07-15-inkling","published":"Wed, 15 Jul 2026 00:00:00 GMT"},{"title":"6 months to live for open models","summary":"Interconnects calls this stretch 'the most serious test yet' of open source AI's viability, with Kimi K3 and Inkling landing days apart as the sharpest closed-vs-open comparison in months.","source":"interconnects","url":"https://www.interconnects.ai/p/6-months-to-live-for-open-models","published":"Sun, 12 Jul 2026 16:47:42 GMT"}]},{"name":"Agent Security and Guardrails Take Center Stage","slug":"agent-security-and-guardrails","summary":"A real Claude exfiltration exploit and a $14,000 agent-driven AWS bill exposed how far guardrails lag agent-speed damage, and cloud vendors responded with new control planes.","articles":[{"title":"How I tricked Claude into leaking your deepest, darkest secrets","summary":"Security researcher Ayush Paul found a working hole in Claude's web_fetch exfiltration defenses; Simon Willison walks through the exploit and what it means for prompt-injection-resistant tool design.","source":"simon_willison","url":"https://simonwillison.net/2026/Jul/15/claude-web-fetch-exfiltration/#atom-everything","published":"2026-07-15T14:21:54+00:00"},{"title":"AI Agents with Cloud Credentials Are Outrunning Billing Guardrails Built for Human-Speed Mistakes","summary":"A three-person agency ate a $14,000 AWS bill in one day after leaked static keys let attackers burn Claude invocations on Bedrock — InfoQ's read: guardrails built for human-speed mistakes can't catch agent-speed spend.","source":"infoq_ai_ml","url":"https://www.infoq.com/news/2026/07/ai-agents-billing-guardrails/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Thu, 16 Jul 2026 10:17:00 GMT"},{"title":"AWS Ships Claude Apps Gateway as Self-Hosted Control Plane for Claude Code and Claude Desktop","summary":"AWS and Anthropic shipped the Claude Apps Gateway, a self-hosted control plane centralizing identity, policy, telemetry, routing, and spend caps for Claude Code and Claude Desktop deployments.","source":"infoq_ai_ml","url":"https://www.infoq.com/news/2026/07/claude-apps-gateway-aws/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Wed, 15 Jul 2026 11:04:00 GMT"},{"title":"Google's Genkit Ships Agents API with Detached Turns and Human-in-the-Loop for TypeScript and Go","summary":"Google's Genkit Agents API goes preview for TypeScript and Go, packaging tool loops, streaming, and state persistence behind one chat() call, with detached turns for human-in-the-loop approval steps.","source":"infoq_ai_ml","url":"https://www.infoq.com/news/2026/07/genkit-agents-api-preview/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Tue, 14 Jul 2026 10:17:00 GMT"},{"title":"Securing the AI supply chain on GKE: Introducing k8s-aibom for automated AI BOMs","summary":"Google Cloud introduces k8s-aibom on GKE, generating automated AI bills-of-materials so security teams can find unregistered 'shadow AI' workloads that dodge traditional scanners.","source":"google_cloud_blog","url":"https://cloud.google.com/blog/products/identity-security/introducing-k8s-aibom-on-gke-for-automated-ai-bills-of-materials/","published":"Mon, 13 Jul 2026 16:00:00 +0000"},{"title":"Show HN: Gate.cat – block an AI coding agent's rm -RF before it runs","summary":"Gate.cat blocks a coding agent's rm -rf before it runs — one of several new tools this week built specifically to put a brake on agent actions before they execute.","source":"hackernews_ai","url":"https://gate.cat/","published":"Wed, 15 Jul 2026 20:13:02 +0000"}]},{"name":"Cloud Giants Race to Own Agent Infrastructure","slug":"cloud-agent-infrastructure","summary":"AWS, Google Cloud, and Oracle all shipped infrastructure or platform launches this week aimed at owning where enterprise agents run and scale.","articles":[{"title":"Scaling to 1 million concurrent sandboxes in seconds","summary":"Modal details the scheduling system behind scaling to 1 million concurrent sandboxes per workspace in seconds — the infrastructure problem underneath every 'spin up an agent instantly' product claim.","source":"modal_blog","url":"https://modal.com/blog/scaling-to-1-million-concurrent-sandboxes-in-seconds","published":"2026-07-16T00:00:00.000Z"},{"title":"Claude at scale on Google Cloud: Frontier AI, built for enterprise production","summary":"Google Cloud lays out what running Claude at enterprise scale actually requires: accelerator management, cross-continent latency, in-region regulated data, and reliable long-context serving.","source":"google_cloud_blog","url":"https://cloud.google.com/blog/products/ai-machine-learning/claude-at-scale-on-google-cloud-frontier-ai-built-for-enterprise-production/","published":"Tue, 14 Jul 2026 16:00:00 +0000"},{"title":"13 hands-on demos to build on Gemini Enterprise Agent Platform","summary":"Google Cloud ships 13 hands-on demos for its Gemini Enterprise Agent Platform, covering build, scale, govern, and optimize workflows for teams standing up agents.","source":"google_cloud_blog","url":"https://cloud.google.com/blog/products/ai-machine-learning/13-demos-on-gemini-enterprise-agent-platform/","published":"Fri, 17 Jul 2026 16:00:00 +0000"},{"title":"Oracle AI agent builder brings no-code, low-code and pro-code together - TechTarget","summary":"Oracle's AI agent builder now spans no-code, low-code, and pro-code development in one platform, following its move to open Fusion Agentic Applications to coding agents directly.","source":"search_agent_engineering_news","url":"https://news.google.com/rss/articles/CBMi0gFBVV95cUxQbGY0dkMtUlpNN2pzdnlXYUllOS1LbWFLS3FMR3djUk52NkFMenRnakhMUTNSdEtFMGktazRCMzRxN1RhSUg2emMzalpWTlptQ2hUY1lqM0JVeDRKTVNFdFd1Y2FCTXVMem1YUlhPVkU3Y3pCdVo0UG9ra204dWM0dEZmR3F0UVdtZEJWNmhheVAyUndVODI2YU51Qzhic3JjZktRNjZNTU54angtZTFyajNCaVRhVVZBY3pvb1ZHaDNaMVBOemlJR0dMMm4wc1VSbXc?oc=5","published":"Tue, 14 Jul 2026 16:38:28 GMT"},{"title":"Atlassian Extends AI Reach of Jira Into Agentic Engineering Workflows - DevOps.com","summary":"Atlassian extends Jira's AI reach into agentic engineering workflows, folding agent-driven work tracking into its existing project-management surface.","source":"search_agent_engineering_news","url":"https://news.google.com/rss/articles/CBMikwFBVV95cUxNbTdpU0Zfb2JVZlZadXVsWWU4S2djRDktMzZzalVxeGZfeHBzMWNHU1MtT2x0cGRCTk9fa3FoSEJBZmJxSlZlQ2drUXgzNjM2enptZlRtWE5uRGVTZk1aZVE2Smh5MnhhTHF5M1Q1djRXYnpxZ1U1amxyOVlYVUV0cVBwZnBfOWRFb3plZnJzNVBYMms?oc=5","published":"Wed, 15 Jul 2026 16:21:38 GMT"},{"title":"Transform your sales organization with Amazon Quick: your new agentic AI teammate","summary":"AWS launches Amazon Quick as an agentic AI teammate for sales orgs, covering the full cycle from prospect identification through deal-closing follow-up.","source":"aws_ml_blog","url":"https://aws.amazon.com/blogs/machine-learning/transform-your-sales-organization-with-amazon-quick-your-new-agentic-ai-teammate/","published":"Fri, 17 Jul 2026 18:42:36 +0000"}]},{"name":"Coding Agent Tooling and Context Engineering","slug":"coding-agent-tooling-and-context-engineering","summary":"New benchmarks, tracing tools, and token-cost measurements gave platform teams sharper data on how coding agents actually behave in long sessions.","articles":[{"title":"One Wikipedia page costs your AI agent 68,000 tokens","summary":"A Hacker News measurement: a single Wikipedia article costs a coding agent 68,240 tokens of raw HTML, and Nike's homepage costs 353,000 — a reminder that unfiltered web_fetch is a token-budget hazard, not just a security one.","source":"hackernews_ai","url":"https://news.ycombinator.com/item?id=48867021","published":"Sat, 11 Jul 2026 00:12:26 +0000"},{"title":"Show HN: Running over 80M tokens in one agent session with no compaction","summary":"A team ran a single agent session through all 89 sequential Terminal Bench 2.0 tasks — over 80 million tokens — with no measurable accuracy loss and no compaction, challenging assumptions about when context needs pruning.","source":"hackernews_ai","url":"https://github.com/Kiz8-Team/pi-cwl","published":"Tue, 14 Jul 2026 16:26:42 +0000"},{"title":"Mistral Vibe for Code vs Claude Code vs Cursor vs Codex: Four Agents Scored on One Scaffold-to-PR Task - MarkTechPost","summary":"MarkTechPost scores Mistral Vibe for Code, Claude Code, Cursor, and Codex head-to-head on one scaffold-to-PR task, giving platform teams a rare apples-to-apples agent comparison.","source":"search_agent_engineering_news","url":"https://news.google.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?oc=5","published":"Tue, 14 Jul 2026 20:52:41 GMT"},{"title":"[AINews] Codex usage up >10x in 6 months to 7M users, +1M in the past ~day; did Codex overtake Claude Code??","summary":"Latent Space's AINews digest reports Codex usage up more than 10x in six months to 7M users, gaining 1M in a single day, and asks whether Codex has overtaken Claude Code based on the numbers each side is publishing.","source":"latent_space","url":"https://www.latent.space/p/ainews-codex-usage-up-10x-in-6-months","published":"Tue, 14 Jul 2026 01:22:27 GMT"},{"title":"How to Debug Coding Agents with LangSmith Traces","summary":"LangChain ships LangSmith tracing for coding agents across Claude Code, Codex, Cursor, and Copilot, so teams can inspect tool calls, subagents, errors, and costs instead of treating the agent as a black box.","source":"langchain_blog","url":"https://www.langchain.com/blog/your-coding-agents-are-a-black-box-heres-how-to-crack-them-open","published":"Wed, 15 Jul 2026 03:40:08 GMT"},{"title":"When your coding agent doesn't listen: evaluating a 241-turn Claude session","summary":"A write-up evaluates a single 241-turn Claude coding session to study exactly where and why a long-running agent stops following instructions.","source":"hackernews_ai","url":"https://www.kurrent.io/blog/when-your-coding-agent-doesnt-listen/","published":"Wed, 15 Jul 2026 17:40:58 +0000"},{"title":"How Cursor knew Claude Fable 5 was ready for the hardest 1% of problems | Claude by Anthropic","summary":"Anthropic details how Cursor used CursorBench to validate that Claude Fable 5 was ready for the hardest 1% of coding problems before shipping it as a default.","source":"claude_blog","url":"https://claude.com/blog/working-at-the-frontier-cursor","published":"2026-07-17T00:00:00+00:00"},{"title":"How Claude Code Works In Large Codebases Best Practices And Where To Start","summary":"Anthropic publishes best practices for running Claude Code in large codebases — where to start and what changes once a repo outgrows single-file context.","source":"claude_blog","url":"https://claude.com/blog/how-claude-code-works-in-large-codebases-best-practices-and-where-to-start","published":"2026-07-14T01:02:52.517426+00:00"}]},{"name":"Evals and Production AI Engineering Practice","slug":"evals-and-production-ai-engineering","summary":"Conference talks and new benchmarks converged on the same point this week: production AI now lives or dies on evals and harnesses, not prompts.","articles":[{"title":"QCon AI Boston: Production AI Moves Beyond Prompts to Platforms, Harnesses, and Evals","summary":"QCon AI Boston's throughline: production AI has moved past prompt engineering into platforms, harnesses, and evals, with context management and security among the top operational themes.","source":"infoq_ai_ml","url":"https://www.infoq.com/news/2026/07/production-ai-platforms-evals/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Fri, 17 Jul 2026 09:00:00 GMT"},{"title":"Stripe Benchmark Shows AI Agents Build Integrations but Struggle with Validation","summary":"Stripe's new benchmark suite tests whether AI agents can build real Stripe integrations end-to-end; the finding: agents build the integration but struggle to validate it actually works.","source":"infoq_ai_ml","url":"https://www.infoq.com/news/2026/07/stripe-ai-agents-benchmark/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Wed, 15 Jul 2026 14:25:00 GMT"},{"title":"Do Automated Evals Work?","summary":"Hamel Husain compares 100 human-annotated traces against automated eval systems to test whether automated evals actually agree with human judgment.","source":"hamel_husain","url":"https://hamel.dev/","published":"Sat, 11 Jul 2026 07:00:00 GMT"},{"title":"Online vs. Offline AI Evals: When to Use Each","summary":"Inngest breaks down when to reach for online versus offline AI evals, and why conflating the two hides real production failures.","source":"hackernews_ai","url":"https://www.inngest.com/blog/online-vs-offline-ai-evals-when-to-use-each","published":"Tue, 14 Jul 2026 21:47:42 +0000"}]},{"name":"Frontier Lab Updates: Anthropic, OpenAI, Google","slug":"frontier-lab-updates","summary":"Anthropic, OpenAI, and Google published a mix of safety, ROI, and research commitments this week, a steady drumbeat of frontier-lab output alongside the bigger model and infrastructure stories.","articles":[{"title":"A scorecard for the AI age","summary":"OpenAI CFO Sarah Friar introduces a scorecard for measuring AI ROI: useful work delivered, cost per successful task, dependability, and return on compute — a framework aimed at cutting through vague productivity claims.","source":"openai_blog","url":"https://openai.com/index/a-scorecard-for-the-ai-age","published":"Fri, 17 Jul 2026 10:00:00 GMT"},{"title":"GPT-Red: Unlocking Self-Improvement for Robustness","summary":"OpenAI details GPT-Red, an automated red-teaming system that uses self-play to harden models against prompt injection and other robustness failures.","source":"openai_blog","url":"https://openai.com/index/unlocking-self-improvement-gpt-red","published":"Wed, 15 Jul 2026 10:00:00 GMT"},{"title":"The US is advancing AI safety through state and federal action","summary":"OpenAI stakes out a 'reverse federalism' position on AI governance, arguing state-level laws should build toward a coherent national safety framework rather than compete with it.","source":"openai_blog","url":"https://openai.com/index/advancing-ai-safety-through-state-and-federal-action","published":"Wed, 15 Jul 2026 12:00:00 GMT"},{"title":"Introducing Claude for Teachers","summary":"Anthropic launches Claude for Teachers, extending its education push with tools built for classroom use.","source":"anthropic_newsroom","url":"https://www.anthropic.com/news/claude-for-teachers","published":"2026-07-14T15:00:00+00:00"},{"title":"Anthropic commits $10 million to Canadian AI research","summary":"Anthropic commits $10 million to Canadian research institutions to fund the next generation of AI researchers.","source":"anthropic_newsroom","url":"https://www.anthropic.com/news/canadian-ai-research","published":"2026-07-14T13:00:00+00:00"}]}]}