Google published androidx.a2ui and androidx.a2ui.compose as new Jetpack artifact groups containing a 1.0.0-alpha01 edition of its official A2UI engine for Compose, alongside androidx.compose.material3…
An Anthropic engineer stated that production code written by Claude should meet a higher bar than human-written code, citing lint rules, tests, Claude-driven end-to-end tests, Claude-powered fuzzers r…
The "Thursdays with Koog" series concluded after 11 weeks with a post examining Koog's near-complete focus on hosted models and its limits for local, on-device inference. The post notes Koog's only on…
Knosh 0.4.0, an agent harness from CommonsWare, now supports the Model Context Protocol (MCP) via a Koog-supplied bridge, allowing developers to configure remote or local MCP servers in its knosh.json…
Mark Murphy reports that mlx-serve, an Ollama-like inference server with a GUI, delivers a 4x speed increase for running Qwen 3.8 on his 64GB M2 Ultra Mac Studio compared to his previous Ollama setup,…
Knosh 0.3.0, a one-shot coding agent built on the Koog framework, implements a permission system that allows agents to deny specific tools by their unique identifiers in frontmatter, with filesystem-m…
Qwen 3.8, a 27B-parameter open-weights model from Alibaba, delivers coding output in the Claude Haiku-to-Sonnet range but is 15-20 times slower than Claude Sonnet, according to tests by a developer on…
Knosh 0.3.0, an open-source tool, now uses Koog's AIAgent with configurable temperature and maxIterations parameters. Temperature controls the randomness of token selection in LLMs, while maxIteration…
Koog's AIAgent delegates its work to an AIAgentStrategy interface, with the default singleRunStrategy() factory function implementing a one-shot graph pattern that calls the LLM, executes tools, and r…
Meta's open-weights Muse-Glimmer model, run locally via Ollama on a 64GB M2 Ultra Mac Studio, took more than twice as long as Qwen 3.6 to complete a documentation-review prompt and produced a small fr…
Local models such as Qwen 3.6 and Gemma 4 are close to being useful for Kotlin Multiplatform code generation but still fall short for many developers, according to Mark Murphy's experiments. Murphy su…
Knosh 0.3.0, an open-source coding agent built on Koog's AIAgent framework, now displays a Braille-dot loading animation that advances with each tool call and status event, improving user feedback dur…
A developer argues that Model Context Protocol (MCP) tools are safer and more auditable than shell access for coding agents, citing Simon Willison's warning that giving an agent a shell with internet …
Koog's Knosh 0.3.0 offers a class-based approach for defining tools, where developers extend Tool or SimpleTool and use LLMDescription annotations to guide the LLM, as demonstrated by the CreateDirect…
Mark Murphy's 'Thursdays with Koog' column explains that in the AIAgent framework, the user's instruction is only a 'user message' within a larger prompt that includes a system prompt and conversation…
Knosh, an open-source coding agent, uses Koog's AIAgent instead of PromptExecutor because PromptExecutor only executes one-shot prompts without handling tool calls or multi-turn conversations. Mark Mu…
Security expert Thomas Ptacek argues that LLM agents are uniquely suited for vulnerability research, and Anthropic researcher Nicholas Carlini demonstrated that a simple prompt to Claude Opus can iden…
Mark Murphy's Knosh coding agent maps model strings from Markdown frontmatter to Koog's LLMProvider and LLMClient objects, supporting providers including Ollama, Anthropic, Mistral, and OpenAI. The ma…