ToolsHow to evaluate AI‑generated frontend code beyond surface polishAI can rapidly produce seemingly complete frontend UIs, but rendered screens and passing builds are only the starting point.8 min read
ToolsUsing NeMo Relay to Trace and Evaluate Hermes Agent BehaviorNVIDIA’s NeMo Relay integration with Hermes Agent provides structured traces (ATOF, ATIF, and OpenTelemetry) to inspect model and tool calls, errors, retries, timing, and token usage.8 min read
ToolsOpenAI's new Decisions API echoes Jev-style decision models for cheaper, faster agent supervisionAt OpenAI's Dev Day, CEO Sam Altman introduced a limited-preview Decisions API that offers Jev-like functionality by letting the Luna model choose from predefined options quickly and cheaply.3 min read
ToolsDoorDash launches text-based AI agent for ordering via Apple MessagesDoorDash announced a text-to-order AI agent that lets U.S.2 min read
ToolsInstinct's new 'Selections' pushes product recommendations, some users call them unsolicitedAI startup Instinct launched "Instinct Selections," a human-curated product-recommendation feature announced by founder Noah Shinn.3 min read
ToolsRestate raises $20M to scale durable workflow infrastructure for AI agentsBerlin startup Restate raised $20 million in a Series A led by Singular to expand its durable execution platform that supports long-running, failure-resilient AI agent workflows.3 min read
ToolsGoogle transitions Gems to 'Skills' — user-created AI assistants will be preservedGoogle will retire the Gems feature of its Gemini app and relaunch it under the name Skills after November 17, 2026.2 min read
ToolsAI SDK reaches 30 million downloads per weekAccording to AI SDK, the software has now reached 30 million downloads per week; the announcement highlighted the 5M/10M/20M/30M milestones.1 min read
ToolsJesse Vincent’s Superpowers: building disciplined agent workflows and preserving human expertiseJesse Vincent’s Superpowers framework teaches Claude Code to behave like a disciplined senior engineer by codifying processes, rationales and checks into reusable skills and agent colleagues.5 min read
ToolsDesigning an AI-native Open Source: Lessons from NVIDIA TensorRT Model ConnectNVIDIA’s open source TensorRT Model Connect is a C++ collection of reference implementations built on NVIDIA TensorRT that aims to make inference performance accessible to model developers who are not TensorRT experts.7 min read
ToolsNVIDIA VSS Blueprint 3.3 lowers development and runtime costs for visual AI agentsNVIDIA released version 3.3 of the Metropolis Blueprint for Video Search and Summarization (VSS), adding a Build Vision Agent skill and Adaptive Efficient Video Sampling (EVS) to cut both development time and VLM runtime compute.6 min read
ToolsHow to Make Knowledge Your Own in the Age of AIMike Loukides reflects on what it means to truly 'possess' knowledge when artificial intelligence can discover, summarize, and write about findings.3 min read
