{"date":"2026-07-27","title":"What happened in AI — Jul 27, 2026","generated_at":"2026-07-27T21:12:02Z","intro":["Moonshot's 2.8-trillion-parameter Kimi K3 became publicly downloadable today, and vLLM and Modal already have day-0 serving support live — vLLM ships hybrid KDA prefix caching and DSpark speculative decoding, Modal pairs the model with a custom-trained DFlash speculator. A HackerNoon portability audit pushed back on the launch-week hype, arguing K3's headline benchmark scores don't hold up once you account for how it was evaluated.","Two new tools target unsafe agent tool calls directly: Belay adds a local firewall for coding agents, and a static verifier built on the \"Guardians of the Agents\" formal-verification paper checks OpenCode's tool calls before they run. Separately, a widely syndicated report put Chinese models at nearly a third of enterprise LLM tokens at one-tenth the cost of US rivals, even as DeepSeek paused a new funding round."],"highlights":["Moonshot's Kimi K3 (2.8T params) went public with day-0 vLLM and Modal serving support, though a HackerNoon portability audit disputes its launch-week benchmark scores.","Belay and a new OpenCode static verifier both launched to catch unsafe agent tool calls before they execute.","philschmid's EvoCode-Bench tests coding agents across 227 sequential rounds, showing single-turn scores overstate reliability.","A coding agent refactored a 750,000-line codebase over three days with zero human code review, running 31 verification passes to catch 201 errors.","Chinese models now handle nearly a third of enterprise LLM tokens at one-tenth the cost of US rivals, per a widely syndicated report, even as DeepSeek paused a new funding round.","Anthropic named Cognizant a Global Premier Partner after training over 30,000 of its associates on Claude."],"categories":[{"name":"Kimi K3 Goes Public, and the Skepticism Starts Immediately","slug":"kimi-k3-public-launch-skepticism","summary":"Moonshot's 2.8-trillion-parameter Kimi K3 is now publicly downloadable with day-0 vLLM and Modal serving support, but a portability audit is already challenging its launch-week benchmark claims.","articles":[{"title":"Kimi K3 now available for public download","summary":"Moonshot's 2.8-trillion-parameter model is downloadable directly, following its initial launch.","source":"search_cn_open_weight_labs","url":"https://news.google.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?oc=5","published":"Mon, 27 Jul 2026 03:04:11 GMT"},{"title":"A portability audit questions Kimi K3's launch-week scores","summary":"HackerNoon argues K3's headline benchmark numbers don't hold up once you account for how the model was evaluated.","source":"search_cn_open_weight_labs","url":"https://news.google.com/rss/articles/CBMihgFBVV95cUxPUkQ5eko4cWFJZzBiVFdNMmtQRFRQQXBRbHVxTFctMXk4c04yckhWSTRJUk82cGtaVUNJcXVaN3FGZXVRSFF1bWRsdlZIMjhRS3hBRHVvNThMWkFCeFd6NVJPMXJrUlhzb1NJQzM1ZEpSUW5TRk5jeE1yZjdxZ0R1b1BNdVkwQQ?oc=5","published":"Mon, 27 Jul 2026 02:18:56 GMT"},{"title":"vLLM ships day-0 Kimi K3 serving support","summary":"vLLM's Kimi K3 support includes hybrid KDA prefix caching and DSpark speculative decoding, tuned for both NVIDIA and AMD GPUs.","source":"vllm_blog","url":"https://vllm.ai/blog/2026-07-27-k3","published":"Mon, 27 Jul 2026 00:00:00 GMT"},{"title":"Kimi K3 lands on Modal with a custom speculator","summary":"Modal's Kimi K3 deployment pairs the 2.8T-parameter model with a custom-trained DFlash speculative-decoding model.","source":"modal_blog","url":"https://modal.com/blog/kimi-k3-by-moonshot-now-available-on-modal","published":"2026-07-27T00:00:00.000Z"}]},{"name":"Coding Agents Take On Bigger, Less-Supervised Work","slug":"coding-agents-bigger-less-supervised-work","summary":"A case study shows a coding agent refactoring a 750,000-line codebase over three days with no human code review, while GitHub frames its own harness, not any single model, as the unit that makes agentic workflows reliable.","articles":[{"title":"A coding agent refactors a 750k-LOC app with zero human code review","summary":"The agent rebuilt a core system invariant in three days, running 31 verification passes and correcting 201 errors before shipping with no human review.","source":"hackernews_ai","url":"https://news.ycombinator.com/item?id=49068698","published":"Mon, 27 Jul 2026 12:28:35 +0000"},{"title":"GitHub: the harness, not the model, is what makes agentic coding reliable","summary":"GitHub's Copilot workflow guidance argues a consistent prototype-plan-implement-review harness matters more than chasing each new model release.","source":"github_blog_ai_ml","url":"https://github.blog/ai-and-ml/github-copilot/the-harness-is-all-you-need-mostly/","published":"Mon, 27 Jul 2026 18:00:00 +0000"},{"title":"An open-source resume-screening agent pipeline","summary":"Hyre chains an LLM pipeline (PDF-to-markdown, layout analysis, data extraction, scoring) to turn a resume into a numeric evaluation.","source":"hackernews_ai","url":"https://github.com/grandimam/hyre","published":"Mon, 27 Jul 2026 17:32:03 +0000"}]},{"name":"New Tools Aim to Catch Unsafe Agent Actions Before They Run","slug":"catching-unsafe-agent-actions","summary":"Two new builder tools, a local firewall and a static verifier, check agent tool calls before execution, while a new benchmark shows single-turn evals miss the multi-turn regressions that actually break agents.","articles":[{"title":"Belay: a local firewall for AI coding agents","summary":"Belay intercepts and filters coding-agent actions locally, before they reach the file system or shell.","source":"hackernews_ai","url":"https://github.com/SECBLOK/belay","published":"Mon, 27 Jul 2026 14:11:46 +0000"},{"title":"A static verifier blocks unsafe tool calls in OpenCode","summary":"Built on the \"Guardians of the Agents\" formal-verification paper, the plugin statically checks OpenCode's tool calls before execution.","source":"hackernews_ai","url":"https://github.com/albertjoseph0/opencode-plugin-guardians","published":"Mon, 27 Jul 2026 02:45:22 +0000"},{"title":"EvoCode-Bench tests agents across 227 sequential rounds","summary":"Single-turn benchmark scores overstate reliability; the real failure mode is regressions across a persistent workspace, not missing features.","source":"philschmid","url":"https://www.philschmid.de/evocode-bench","published":"Mon, 27 Jul 2026 00:00:00 GMT"},{"title":"NVIDIA and industry partners launch the Open Secure AI Alliance","summary":"The alliance frames open-source security practices as infrastructure for AI safety, extending the model that secured cloud and telecom software.","source":"nvidia_blog","url":"https://blogs.nvidia.com/blog/open-secure-ai-alliance/","published":"Mon, 27 Jul 2026 09:00:07 +0000"}]},{"name":"Platform Teams Build Serving and Data Layers Purpose-Built for Agents","slug":"serving-data-layers-for-agents","summary":"Netflix detailed its in-house Triton/vLLM serving stack, LangChain built a sub-second full-text search index over agent traces, and AWS pitched task-aware compression as a fix for RAG's document-scale ceiling.","articles":[{"title":"Netflix details its in-house LLM serving platform","summary":"Netflix built its own serving layer on Triton and vLLM rather than relying on a third-party inference API.","source":"search_llm_ops_news","url":"https://news.google.com/rss/articles/CBMiZ0FVX3lxTFBpby0wOTVaSTc0LWVRam9WelpQRDdhdW1KMGJYWFRIRWZnQXFGYWdhVjAxUGRhNkRlRGl4VEZnRFVnbEVqeWJtdGVqSFlNa1BIcVhtZzBqdFZHTHlQS3daQXdJRmdrQ0U?oc=5","published":"Mon, 27 Jul 2026 07:33:13 GMT"},{"title":"LangChain's SmithDB adds full-text search over agent traces","summary":"SmithDB indexes large, deeply nested agent-trace JSON in object storage with a median 400ms search latency.","source":"langchain_blog","url":"https://www.langchain.com/blog/full-text-search-in-smithdb-designing-an-inverted-index-for-object-storage","published":"Mon, 27 Jul 2026 04:06:56 GMT"},{"title":"AWS pitches task-aware knowledge compression as RAG's next step","summary":"Traditional RAG breaks down on analytical tasks spanning hundreds of documents; AWS's TAKC pre-compresses knowledge bases into task-specific representations instead.","source":"aws_ml_blog","url":"https://aws.amazon.com/blogs/machine-learning/beyond-rag-task-aware-knowledge-compression-for-enterprise-ai-on-aws/","published":"Mon, 27 Jul 2026 16:11:32 +0000"},{"title":"Why agentic AI needs its own API gateway pattern","summary":"InfoQ argues traditional API gateways assume deterministic services, an assumption agentic AI breaks, pushing engineering leaders toward AI gateways as a dedicated architecture seam.","source":"infoq_ai_ml","url":"https://www.infoq.com/articles/evolutionary-architecture-pattern/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","published":"Mon, 27 Jul 2026 11:00:00 GMT"}]},{"name":"China's Model Cost Advantage Keeps Widening, Even as DeepSeek Wobbles","slug":"china-model-cost-advantage-deepseek","summary":"A widely syndicated report put Chinese models at nearly a third of enterprise LLM tokens at one-tenth the cost of US rivals, even as DeepSeek paused a new funding round and US lawmakers weigh restrictions that could slow the catch-up.","articles":[{"title":"Chinese models now handle nearly a third of enterprise tokens, at one-tenth the cost","summary":"A widely syndicated report puts Chinese open models at roughly 30% of enterprise LLM token volume, priced at about one-tenth of comparable US models.","source":"search_cn_open_weight_labs","url":"https://news.google.com/rss/articles/CBMi9AFBVV95cUxONEQ2WVhlQ2l0ZVVMVlltc3BqNXdtM2tWcDhiRTlNZ0lERFQ4MFdneV8tejVTVnNnTFc2SXJqNHRPN2k3cUU1RUZjUk9PMkNTWWptNDhYY2oyVGVOMC1PUEladElkZG50b3JnRjhtcEZ0S1lnUUZXaFh1VDBzR3EwcEExVDVOVHVCakRZWmNOdTl5QWE0OEk4WGVEQzI5Zm1HUW40bkJwenJMRDZlNTJ6TTdEcXZjZG8yRV9oVlBNRFMxak5PcktnZERKM1dRYUcwM1lKMEozUGRBLVlySkM5c0V3VERjenlKOU1qdjVzVGZyODZ4?oc=5","published":"Mon, 27 Jul 2026 14:45:51 GMT"},{"title":"DeepSeek pauses a new fundraising round","summary":"The pause follows leaked comments attributed to founder Liang Wenfeng that went viral days earlier.","source":"search_cn_open_weight_labs","url":"https://news.google.com/rss/articles/CBMi0gFBVV95cUxNNjllUWhlWEdieUFLX3U3aENPbjk3U0lDcUR0ZERJU05XS2VnWFlsSHpSdjB5UGFXeFh0LTU4WkVLbFhCVDRoX0pSZnBSbTd5Zl9WaUN2QzZabWQzbUVXSTRpdUFoSnVyX0hocDFjN2ZOQmtJUmk2MlNiRG5CbUpyVXE3Q3BoVUdhOFV1ZVFYdERjY2J4dWhibDgwVHRnaGtFR2x3RDhtS0dtZmdsU1hjclkwUDlTSDJrR1Vrd01IOXBjN0RJM0pUQ20xOTZTS0psZUE?oc=5","published":"Mon, 27 Jul 2026 15:19:00 GMT"},{"title":"Congress debates restrictions that could slow the US AI catch-up","summary":"Washington Examiner frames pending congressional measures as a risk to US competitiveness against China's AI pace.","source":"search_cn_open_weight_labs","url":"https://news.google.com/rss/articles/CBMingFBVV95cUxNQXhybzBvZFZfX3VjdGdrd3lDcHItU2pPSVVqLXliN3NPVnhObHM4LXNDY2lHTWFKM18zWTJKUGh6LVVXd1d5VEhEUmxXSFg2cC02aGNtUngtc1JXQllJX0pXWGF6azA0VnU1MGUzeFdNRmpzTE1fYXF3UXZGdlhicWxVdHNYUTV4VlNZdG1yWEhQQmJia1VNa1dqNERJZw?oc=5","published":"Mon, 27 Jul 2026 13:00:00 GMT"}]},{"name":"Enterprise Adoption Signals","slug":"enterprise-adoption-signals","summary":"Anthropic named Cognizant a Global Premier Partner after training over 30,000 of its associates on Claude, and new OpenAI research tracks how ChatGPT is reshaping what workers actually do inside their roles.","articles":[{"title":"Anthropic names Cognizant a Global Premier Partner","summary":"Cognizant has trained 30,000+ associates on Claude and is embedding it across its platforms under the expanded partnership.","source":"anthropic_newsroom","url":"https://www.anthropic.com/news/cognizant-anthropic","published":"2026-07-27T15:32:00+00:00"},{"title":"OpenAI research tracks how ChatGPT is reshaping job tasks","summary":"The study finds ChatGPT users are taking on tasks that cross traditional role boundaries, not just automating existing ones.","source":"openai_blog","url":"https://openai.com/index/how-ai-is-expanding-what-people-do-at-work","published":"Mon, 27 Jul 2026 03:30:00 GMT"}]}],"article_count":20}