IBM released Granite 4.2, a family of open-source reasoning models at 3B, 8B, and 30B parameters, under the Apache 2.0 license on August 25, featuring switchable thinking modes and trained via reinfor…
AptAI launched a centralized hub for discovering, testing, and deploying LLM adapters, enabling one-click serverless deployment with sub-millisecond execution overhead and zero cold starts. The platfo…
A fine-tuned model called ReViSQL-K2.6, developed by Tinker using reinforcement learning with verifiable rewards, exceeds the human benchmark of 92.96% on the BIRD text-to-SQL task, reaching 92.96% wh…
A developer's analysis of agent recovery systems reveals that a healthy service response does not guarantee task progress, citing an OpenHands incident where a stuck conversation remained responsive f…
A developer estimates that human code review costs $0.243 per changed line, 2.1 times the $0.114 model spend for an AI agent to produce an accepted changed line, based on a SmartBear/Cisco study, U.S.…
A controlled longitudinal study of 35 sequential releases of the Qwen Code CLI, holding the underlying LLM constant, found that agent harness evolution significantly impacts coding agent quality, with…
LangWatch, an open-source testing, evaluation, and observability platform for AI agents, offers simulation-based testing, adversarial red-teaming, and OpenTelemetry-native observability. The platform,…
OGX (Open GenAI Stack), an open-source AI application server and Python library, implements the APIs of OpenAI, Anthropic, and Google with pluggable backend providers, enabling developers to build age…
Eigent, an open source desktop application built on the CAMEL-AI multi-agent framework, enables users to deploy a multi-agent AI workforce locally, with agents that can access files, browser, and term…
LoopTroop, an open-source GUI for long AI coding tasks, has evolved its architecture to distinguish between discarding state that may belong to a failure and preserving state that is still plausibly v…
An engineer at Anthropic distills hard-won lessons from production AI agents including Claude Code, OpenHands, and SWE-agent into a practical field guide for building enterprise-ready agents. The guid…
BrassCoders finds that autonomous coding agents, such as Claude Code and SWE-agents, bypass human code review, allowing security vulnerabilities like hardcoded credentials to go undetected. The compan…
Traccia SDK, an open-source execution evidence layer for AI agents, has been released under the Apache 2.0 license. The tool records autonomous agent behavior to make it provable, reproducible, and au…
A developer's analysis of the AI-agent ecosystem reveals that model launches are the strongest acquisition event for agent tools, with openai/codex doubling claude-code's star pace within hours of GPT…
Workers ordered to use AI by their bosses are then penalized for it, a phenomenon researchers call the "AI penalty" that is costing employees promotions and raises, according to a Business Insider rep…
A study analyzing 1,794 valid coding trajectories from a dataset of 3,843, generated by seven leading models across three coding-agent scaffolds (OpenHands, MiniSWE, and Terminus2), found that coding …
A new study from Northeastern University professor Christoph Riedl finds that managers often devalue employees' work once they learn AI played a part, creating a 'AI penalty' where workers face conseq…
A developer argues that the future of LLM usage lies in loops rather than single prompts, citing that products like Claude Code, OpenAI Codex, and Cursor Agent already operate on iterative loops that …
NVIDIA tested its open-weight Nemotron family of models on real-world coding agent tasks and found a capability floor below which models cannot drive an agent loop at all. The smallest variant, Nano 1…
Loop Engineering is emerging as a replacement for traditional prompt engineering as AI agents increasingly rely on iterative feedback loops rather than isolated prompts. The intelligence of modern AI …