Databricks built an internal benchmark from real pull requests on their large codebase to evaluate coding agents. Key findings: frontier models form a Pareto frontier; open models like GLM 5.2 reach top tier; per-task cost is more revealing than per-token cost; the harness (framework) significantly impacts cost and quality. They sealed git history to avoid cheating. The approach enables data-driven model selection.

How to build governance into autonomous security agents from the architecture up.

Webdog.ai is an open-source web monitoring tool that tracks changes on any website, providing AI-generated summaries, visual diffs, and instant alerts. It notifies users within seconds via various channels, includes screenshots, and can be self-hosted. Start watching for free.

An open-source project using Telnyx API to enable foreign language practice with AI over a phone call. The system leverages speech recognition, AI inference, and text-to-speech for real-time interaction.

Constellation Network launches Gate AI, a proxy gateway that protects AI agents from prompt injection attacks, automatically saves tokens, and creates a verifiable audit trail. It ranks #1 across 16 public benchmarks with a 97.4% F1 score. Compression saves 20%+ on tokens automatically with zero code changes. Free tier available.

Learn how to go from an empty repository to a live custom domain with HTTPS in about 14 minutes, without manually editing a single DNS record, using GitHub Copilot CLI and a Namecheap skill.

Flint is an open-source visualization intermediate language from Microsoft Research, designed to help AI agents create expressive, polished charts from compact, human-editable specifications. It handles low-level design details automatically via semantic types, supports multiple rendering backends, and powers the Data Formulator project.

This post shows you two architecture patterns that address the health check challenge when using an internet-facing ALB with AWS WAF to secure Amazon Bedrock AgentCore Runtime. Pattern 1 uses a Lambda proxy for request transformation and custom logging, while Pattern 2 targets VPC Endpoint ENI IPs directly for lower latency. Both patterns include guidance on closing the direct-access backdoor with a resource policy and have been tested with SigV4 and OAuth authentication.

Startup MorphMind launches Academic Humanizer, a tool that uses AI to remove AI-writing tells from academic papers and grant proposals, raising concerns about integrity and the proliferation of low-quality AI-generated research.

Learn how to use Jamf's AI Governance with Amazon Bedrock to configure, deploy, and validate managed settings for AI applications across a Mac fleet, including Claude Code, Claude Desktop, and OpenAI Codex.

Analysis of trends in AI agents based on data from 1,080 Y Combinator startups.

An AI agent simulated the 2026 FIFA World Cup knockout stages, predicting outcomes from the Round of 32 through the Quarter-finals. Notable results include France edging Morocco 2-1 in the quarters, Argentina surviving Egypt on penalties, and a major upset as Norway defeated Brazil in the Round of 16.

A guide for iOS developers to ship private, local AI on Apple devices. Covers decision frameworks, production use of foundation models, model ownership with MLX Swift and Ollama, and shipping concerns like memory, privacy, and evaluations. All code is compiler-verified against shipping SDKs.

A series of violent protests and attacks against AI companies and data centres are on the rise, highlighting growing opposition to AI development.

Cruxible is an open-source state layer that provides a typed, verifiable hard state for AI agents and human teams. It models domains with Terraform-like configuration and enforces write rules through deterministic workflows and proposal/review governance. Unlike RAG or vector memory, it enables reproducible queries, multi-hop traversals, and staleness checks, allowing agents to base decisions on accountable truth. It is designed to work with any agent framework and includes a demo for supply-chain scenarios.

Cowork's move to mobile and the web shows how AI is extending into how people work.

Gmail's AI failed at a nuanced research task, but Claude Cowork found the right pitches, quotes, and permissions, proving connected AI assistants may finally help tackle some aspects of email overload.

LangChain and NVIDIA launch the NemoClaw Deep Agents blueprint, combining Deep Agents Code, Nemotron 3 Ultra, and OpenShell for open, governed enterprise agents.

By tuning only the harness (scaffolding) around the Nemotron 3 Ultra, we achieved a best run of 0.86 on the Deep Agents suite, nearly matching Opus 4.8's best of 0.87, at roughly 10x lower cost. This article details the eval-driven approach, prompt engineering, middleware optimizations, and what didn't work.

They suck up energy and water, and blast out heat. Just who is better off from all this investment – aside from tech bros?

NVIDIA Nemotron 3 Ultra is offering leading performance at lower cost than top closed models with the largest and most widely adopted AI agent orchestration platform. LangChain tuned its Deep Agents harness for NVIDIA Nemotron 3 Ultra, achieving the highest accuracy among open models, while completing more tasks at higher throughput and running at 10x lower inference cost per run than leading closed models.

Run Deep Agents Code on NVIDIA NemoClaw with deny-by-default networking, human approval, and audit logs for sensitive code modernization.

The article discusses how AI-generated UI often looks bad, and the author experiments with methods to improve it by evaluating design quality and using self-improvement loops, finding that showing the model its own output and injecting examples of good design from top apps significantly enhance results.

brAIn is an experimental AI agent framework that replaces the traditional chat loop with a NATS pub/sub bus architecture of long-lived daemon nodes. Nodes are reactive, only activating when relevant messages arrive, saving token consumption. Each node can have its own UI, supports distributed deployment, and features priority preemption and MCP client integration. The author demonstrates applications like ambient room agents, Slack listeners, and IoT controllers, and compares the architecture with existing tools such as LangGraph, AutoGen, and ROS 2.

Start with A is an open-source investment research platform integrating research, portfolio monitoring, and journaling into a disciplined workflow. It is self-hosted with BYOK AI support for Gemini, OpenAI, and Anthropic.

Roborock took its flagship robot vacuum and really improved it, while keeping the same price. Here's how.

Quilt is a CLI tool that tracks which AI coding agent wrote which lines in a shared Git checkout, enabling multiple agents to work simultaneously without git conflicts. It provides per-line attribution, symbol-level claims, and collision prevention to ensure clean per-agent commits.

The debate over open vs. closed AI has shifted from academic to urgent for security teams. U.S. policy contradictions, cost pressures, and the rise of competitive open-weight models like GLM-5.2 are driving enterprises toward Chinese open-source AI, with significant security implications.

The legal sector has seen the entrance of a host of AI vendors aiming to automate core legal functions.

Officials in Wyoming said a contractor for Meta flushed bacteria-contaminated water into public sewers during construction of a controversial AI datacenter, prompting Cheyenne water authorities to implement strict safety regulations for wastewater disposal. Meta said it is working to be a good neighbor and that drinking water supplies were not affected.