#Agent

AILovable joined AWS, Google Cloud, Databricks, Salesforce, and other firms as a founding member of the Blueprint Alliance, a coalition developing an open reference architecture for securing and governing enterprise AI agents. The blueprint covers registering agents as identities with accountable owners, scoping their access to tasks, enforcing policies through gateways, and responding to incidents by revoking tokens or quarantining agents.

AITencent Hunyuan introduced WebCraftBench, a benchmark that tests AI agents by using the live web app and scoring aesthetics, usability, and whether the original request was met. Coverage-guided exploration reaches parts of the app that agents otherwise miss. On 197 human-validated pairs, the benchmark matches human preference 85.3% of the time.

AIOpenBMB thanked a developer for testing MiniCPM5-2B as a worker in a multi-agent workflow handling invoice matching, short payments, duplicate references, and disputes through tool calls. The background post says GPT-6 Astra coordinated the MiniCPM5-2B workers, verifying 32 synthetic invoices in 67.8 seconds with 232 executed tool calls. The demo does not move money.

AISpaceXAI says its combined support team handled a 175% rise in tickets without hiring, crediting Grok Bot, which it says would otherwise have required about 200 additional staff. The company reports resolving tickets for $0.20 to $0.30 each, versus the $1 to $4 per resolution it attributes to traditional AI support tools. Grok Bot is also reported to resolve 99% of refund requests without human intervention.

AIvllm-metal ports vLLM's scheduler, paged KV cache, and OpenAI-compatible server to Apple Silicon, with MLX and Metal handling execution. The v0.28.0 release added batched MTP, GGUF and hybrid-model support, and faster prefill on M5, and v0.29.0 is installable through Homebrew.

Why it matters: The post explains how vllm-metal packs requests and pages KV cache on Apple Silicon, with benchmarks showing where concurrent serving gains and tradeoffs appear.

AIAmp runners can now create Git worktrees from the directory picker, with each new worktree placed as a sibling folder on a new branch off the current HEAD while uncommitted changes stay in the original checkout. Runners can also opt in with --amp-env to inject Secrets & Env Vars configured on ampcode.com into thread shell commands, MCP servers, and plugins, with changes applied to the next thread without a restart.

AIMeta announced a partnership with Shopify to make shopping and checkout easier in Muse. The background post from Shopify's Tobi says the deal enables agentic checkout with Shop Pay across all Shopify stores. More partnerships like this are coming soon.

AIAndrew Ng argues that recent AI danger fears are driven by hype and a PR campaign rather than any new dangerous turn in the technology. He says he sees no increase in extinction risk compared to a few months ago, with cybersecurity as the main real change. He cites the OpenAI agent swarm incident that hacked Hugging Face, arguing its impact was overstated and that responsibility lies with the tool user and system builders rather than the agent.

AIRadixArk says its Miles framework uses an async rollout buffer that can change which sample groups reach training and which prompts get retried, while reusing the rollout worker and trainer. The post argues that stable, granular extension points let contributors modify one part of an RL system without disrupting its neighbors.

AIJeff Dean, who recently left Google after 27 years, thanked Dawn Song for a discussion covering foundational ideas, recursive self-improvement, automated scientific discovery, and AI safety. The post is a brief acknowledgment of that conversation, which Song promoted as Dean's first public talk since leaving Google.

AIXiaomi MiMo released MiMo-V2.6-Distill-Qwen-9B, a 9B agentic model made by supervised fine-tuning Qwen3.5-9B on MiMo-generated data, as an open starting point for agentic reinforcement learning research. It scored 61.1 on SWE Verified, versus 60.0 for Qwen3.5-9B, and 44.6 on SWE Pro, versus 32.0. The checkpoint is served with SGLang and a MiMo chat template, and its SFT data totals 77.4B tokens.

AISpaceXAI has released Grok 4.7, which it describes as its most capable model yet for coding and knowledge work, with NVIDIA supporting the launch through accelerated computing. Elon Musk characterized the model as combining strong intelligence, speed, and low cost.

AIGrok 4.7 has been released as a more capable and smarter model than its predecessor, with no change in speed or cost. The announcement says it improves on Grok 4.6 at the same price and speed.

AIThe post links to an official xAI news page for more details.

AIXiaomi released MiMo-V2.6-Flash-RL, an efficiency-balanced checkpoint in its MiMo-V2.6 series, on Hugging Face. The model is a sparse MoE with 309B total and 15B activated parameters, supports text, image, video, and audio input, and offers a 1M-token context. The technical report says it was trained with a single mixed reinforcement learning run across coding, agent, visual, and cybersecurity tasks.

Why it matters: The report pairs its benchmark tables with the RL training method, which helps readers judge how the checkpoint's scores relate to its training approach.

AIXiaomi MiMo released MiMo-V2.6-Pro-RL on Hugging Face, a sparse MoE model with 1.02T total and 42B activated parameters and a 1M-token context. The technical report says it accepts text, image, video, and audio, and was trained with a single mixed reinforcement learning run across coding, agent, visual, and cybersecurity tasks.

Why it matters: The report pairs a 1.02T-parameter MoE model with an RL-based self-improvement method, useful for judging how reinforcement learning is scaled in frontier open models.

AITim Dettmers argues that agents make single research projects cheap, so the ecosystem, not the paper, becomes the unit of research. He says his lab's open-source week will release two projects and four papers together, including a harness that runs frontier-scale models on local hardware. He also argues that AI job fears are overstated and that university labs can compete on creativity and cheap, valuable problems.

AIA research-agent design has GPT-6 orchestrating while MiniCPM5-2B runs locally on DGX Spark to read sources and extract traceable evidence. The post presents this as an example of pairing frontier reasoning with compact, efficient models for transparent research workflows.

AIMike Knoop argues that programmers are again moving up the abstraction ladder, with few still writing assembly. He says the major shift is toward understanding software by observing behavior through induction and abduction rather than reading code through deduction.

AIFireworks AI reports that an oracle router choosing among 18 models per DeepSWE task reaches 97.6% at $1.88 per task, versus 74.1% at $6.52 for GPT-6 Astra alone. The oracle is hindsight-based, so the authors say a production router must predict the best model before the task starts, which is the harder problem.

AIxAI released Grok 4.7, which it calls its most capable model for coding and knowledge work, built on a larger base model than Grok 4.6 and trained with a longer reinforcement learning run. It is priced from $2 per million input tokens and $6 per million output tokens, the same as Grok 4.6, and is available in Cursor, Grok Build, and the Grok API. xAI reports gains on CursorBench 4.0 (46.3%) and AA Briefcase v1.1 (1,657) over Grok 4.6, and says it posts the strongest safety results it has tested on refusals and jailbreak resistance.

Why it matters: The release pairs a new base model with benchmark tables against named rivals and pricing, letting readers compare its coding and office-work gains against Grok 4.6 and frontier models.

AIA commenter praises a practical setup where a small local model runs a log-debugging pipeline entirely on CPU through OpenClaw and Ollama. They argue that workflow design can make a small model far more useful than its raw size suggests.

AIThe main post suggests ChatGPT Atlas, an AI browser, is being revived within the Codex and ChatGPT app. It also speculates that Codex or Claude Code built-in browsers could one day serve as full everyday browsers.

AIJasper Lu's new blog post walks through training a search agent with GRPO, showing how small reward function changes teach a model to avoid sloppy tool calls, prune unnecessary documents, and balance persistence against token efficiency. The post makes every rollout browsable and releases the code as open source, with the full process from learning rate sweeps to reward shaping documented.

AIMike Knoop asks what the underscore.js equivalent for AI would look like, noting that such programming primitives feel close. He adds that he barely reads or writes code anymore despite these emerging tools. The quoted post introduces Probably, a toy programming language built around Jev, where constructs like "feels," "match," and "while" let AI make decisions within ordinary code.

AIGoogle's weekly recap covers Gemini 3.8 Live and 3.8 Live Extended Thinking, described as its most advanced live dialogue audio models yet. It also notes Dreambeans, a GoogleLabs experiment curating daily personalized stories, is now generally available, and that CC has expanded into a shared agent for household coordination. Google Pics, a Workspace tool for generating and co-creating images, is now GA, alongside AlphaGenome Atlas, DeepMind's interactive genomics discovery platform.

AIEthan Mollick says GPT-6 Astra and Fable 5.1 already enable transformative impact and can reliably handle weeks of human work when properly guided. He cites GPT-6 Astra turning the 1977 text adventure Zork into a 3D action-adventure game and Fable 5.1 reconstructing Umberto Eco's Milan library in 3D from videos, photos, and catalogues.