An engineer's benchmarking guide introduces a cache hit rate metric for measuring prefix caching effectiveness in LLM serving, implemented in the open-source tool llmperf-rs. The metric calculates the…
Moonshot AI released Kimi K3 in July 2026 as a 2.8T-parameter open-weight MoE model with a 1M token context window and native vision support, scoring 88.3 on Terminal-Bench 2.1, close to top closed mo…
Security firm Mitiga has identified a new backdoor attack technique dubbed 'PromptLogger' that poisons AI agent instruction files such as CLAUDE.md, AGENTS.md, and GEMINI.md to exfiltrate prompts, cre…
Vibrisse, a studio focused on eco-design and performance, warns that unconstrained use of generative AI tools like Cursor, Copilot, and Cline can silently erode a codebase's identity and introduce tec…
Anthropic released Claude Opus 5, DeepSeek finalized V4, and Moonshot AI published the open-weight Kimi K3 (2.8-trillion-parameter MoE model) during the biggest AI release week of 2026, fundamentally …
OpenLore, a deterministic, local-first memory and guardrails system for AI coding agents, launched with a one-command install that indexes a codebase into a navigable knowledge graph in seconds, requi…
Kimi's k3 model now runs on a single consumer GPU, achieving 113.83 tok/s on a 32 GB RTX 5090 after being shrunk to 28.8 GB from its original 48B size. The model works with coding agents like Claude C…
OpenCode, an open-source coding agent, can now run entirely locally using QVAC, an SDK and CLI from Tether that serves models on-device across GPU backends. The setup eliminates cloud dependencies, gi…
Cline, an AI coding agent formerly known as Claude Dev, can dramatically reduce implementation time from 5–10 minutes to about 45 seconds by using agent-based Tasks instead of chat, according to a tut…
VortxAI released emem, an open-source shared memory layer for multi-agent systems that stores signed, content-addressed facts outside any single model, enabling agents that share no vendor or trust to…
Claude 3.5 Sonnet is currently the hardest large language model to jailbreak due to its strong attention mechanism and resistance to persona-shift attacks, according to a Machine Learning Forum discus…
Harness Score, an open-source scanner that measures how well a code repository is set up for AI coding agents, has released version 1.5 with customizable scoring, multilingual documentation, and a Git…
Cline achieved an 88.8% SOTA score on Terminal-Bench 2.1 using recursive self-improvement with GPT-5.6-Sol as the leader model, costing $49.80—less than a tenth of Fable 5's $552 run. The single-promp…
A developer built an AI Skill that generates production-ready GitHub Actions workflows with zero configuration errors. The skill enforces 12 quality gates, 5 decision templates, and 3 mandatory securi…
A developer released Turo, an open-source token-saving proxy for CLI AI agents that reduces prompt token counts by up to 70% by stripping articles, prepositions, and repeated words while preserving co…
An AI developer built OmniRoute, an open-source AI gateway that provides a single OpenAI-compatible endpoint routing across 250+ providers. The gateway automatically selects the cheapest working provi…
JetBrains Air now supports GitHub Copilot, OpenCode, Pi, Cline, and other ACP-compatible agents, adds IntelliJ-powered navigation and diagnostics for Java and Kotlin, and enables local models via Olla…
LLM Gateway enables developers to use Moonshot's Kimi K3 model, which topped Arena's Frontend Code evaluation, with coding agents like Claude Code, Cursor, and Cline via a simple base-URL change. The …
Cursor costs $20 per month and GitHub Copilot costs $10 per month, both scoring 57% crowd approval, but solo developers on a tight budget have two free open-source alternatives: Cline and Aider, where…
Kimi has launched 'Kimi Everywhere', a kit that lets developers run any coding harness—including Claude Code, Codex, OpenCode, Cline, and Aider—on a single flat-fee Kimi Code subscription ($19+/mo). T…