Discussion | Link

AI is transforming procurement from a reactive cost center to a strategic function, enabling dynamic sourcing, continuous risk monitoring, and bottom-up adoption despite inherent risk aversion.

The article explores the competitive landscape of open models in 2026, the key factors for their success (performance, provenance, license, tooling, finetunability), and analyzes Google's latest Gemma 4 series. It argues that success depends more on usability and ecosystem support than benchmark scores.

Google Research introduces a systematic evaluation framework that transforms established psychological questionnaires into situational judgment tests for LLMs. Testing 25 models reveals gaps in alignment: models deviate from human consensus in high-agreement scenarios and are overconfident when opinions are divided.

Enterprises lack reliable visibility, control, and accountability over the risks embedded in their third‑party networks, despite being legally and operationally responsible for them. This article explores how third-party risk management is transforming from a compliance issue to a strategic enterprise risk, leveraging AI for continuous monitoring, explainable analytics, and automated remediation.

AI benchmarks are facing saturation and measurement uncertainty. The famous METR chart shows rapid progress, but latest data has wide confidence intervals and the benchmark itself is nearing its limits. As AI tackles longer tasks, traditional methods fail to capture real-world complexity, widening the gap between measured and actual capabilities.

SuprSend AI is an AI-first platform for multi-channel notifications, designed to streamline and optimize notification delivery.

Two weeks of dogfooding Engram, Weaviate's memory product, in daily Claude Code sessions. This surfaced where a dedicated memory product adds value, and the specific mechanics that prevent integration with coding assistants from working well.

Connect any API to any AI agent.

ADeLe, developed by Microsoft Research in collaboration with Princeton and Universitat Politècnica de València, scores AI models and tasks across 18 core abilities, enabling prediction of performance on unseen tasks with ~88% accuracy. It reveals model strengths and weaknesses, providing explainable AI evaluation beyond traditional benchmarks.

GitHub Copilot CLI introduces /fleet, a slash command that orchestrates multiple AI sub-agents to work in parallel on different files. Learn how to write effective prompts, declare dependencies, and avoid common pitfalls.

Discussion | Link

OpenAI ships GPT-5.4 mini and nano with 400k context windows, faster but up to 4x pricier; Mistral open-sources Small 4 model; Meta’s Manus launches Mac agent; Nvidia unveils DLSS 5 and NeMo sandboxed runtime; plus safety and research updates.

Multimodal embeddings allow AI systems to search and reason across text, images, audio, and video in their native formats. This blog covers the key intuitions behind how this all works and walks through three practical implementations using Weaviate and Gemini.

AI capabilities far exceed common usage due to poor interfaces. Research shows chatbot interfaces impose cognitive load, offsetting productivity gains. The article explores specialized interfaces, personal agents like OpenClaw and Claude Cowork with Dispatch, and the emerging paradigm of interfaces generated on demand by AI.

An AI researcher used GitHub Copilot to build coding agents that automated repetitive analysis tasks, sharing key strategies: prompting conversational and verbose, frequent refactoring and documentation, and blaming process not agents. The team created 11 agents, 4 skills, and new workflows in under 3 days.

Adomate is an AI-driven tool that turns data into winning ads at scale, enabling efficient ad optimization across channels.

The Weaviate Managed .NET Client brings Entity Framework Core-like experience to C# developers, enabling attribute-driven schema, type-safe queries, and automatic migrations for vector databases.

This issue covers a diverse range of open models spanning OCR, RAG search, audio transcription, computer use, code editing, math theorem proving, and more. Models come from a broader set of builders including NVIDIA, Cohere, Sarvam, Mistral, and others, highlighting the industry's push for domain-specific, cost-effective models.

This issue explores Stanford professor Andy Hall's concept of 'political superintelligence', the challenges of robot drumming, Google's vision of a society of non-biological intelligences, Meta's self-improving hyperagents, and the new math benchmark HorizonMath.

Notion 3.4 introduces new dashboards, image generation, presentation mode, and smarter AI agents, turning the workspace into an all-in-one AI hub.

Ollama announces a preview release powered by Apple's MLX framework, delivering significant performance improvements on Apple Silicon, including NVFP4 support and enhanced caching.

Synci is an all-in-one AI soundtrack platform for content creators, offering a curated catalog of 40,000+ cinematic tracks, AI-powered sound effects and music generation, full licensing, and a Chrome extension for seamless workflow.

AsgardBench is a new benchmark that tests whether embodied AI agents can adjust their plans based on visual feedback. Built on AI2-THOR, it places agents in kitchen-like scenarios and requires them to dynamically modify action sequences by observing object states (e.g., whether a mug is clean). Tests show that visual input significantly boosts success rates, but current models still struggle with fine-grained visual distinctions, progress tracking, and timely plan updates.

Microsoft Research introduces GroundedPlanBench, a benchmark to evaluate vision-language models on joint action planning and spatial grounding for robot tasks. Their V2GP framework converts robot demonstration videos into training data, showing that grounded planning outperforms decoupled approaches.

Runway partners with Modal to power real-time inference for Runway Characters, a real-time video agent API that generates expressive digital personas from a single image. Modal's serverless platform enables low-latency, multi-GPU inference globally, allowing Runway to move from proof of concept to production in under 30 days.

"We cannot miss this moment because we are distracted by side quests," an exec said.

Attention Insight launches Visual Usability Checker Figma plugin, an AI-powered tool trained on millions of eye-tracking data points to validate design decisions before user testing.

Doppel, an AI-native cybersecurity platform, migrated its ML workflows to Modal to accelerate experimentation and simplify inference deployment. Training experiments became parallel, reducing feedback loops. Inference builds dropped from 30 minutes to under a minute, with automatic scaling for traffic spikes.

Learn how to integrate the Copilot SDK into a React Native app to generate AI-powered issue summaries, with production patterns for graceful degradation and caching.