DeepSeek V4 Flash on a Single AMD MI300X
A production configuration for running DeepSeek-V4-Flash-0731 on a single AMD MI300X GPU achieves 168.6 tok/s median single-stream decode and 542 tok/s aggregate across 8 concurrent streams, with the …
A production configuration for running DeepSeek-V4-Flash-0731 on a single AMD MI300X GPU achieves 168.6 tok/s median single-stream decode and 542 tok/s aggregate across 8 concurrent streams, with the …
Homebench, a new open-source command-line tool, benchmarks local large language models for speed, memory, and quality, presenting results in a live terminal leaderboard. The tool, installable via pip,…
Debarshi Basak released Clawx, an open-source package manager that treats agent-driven tasks as installable packages, where each package is a Markdown file with YAML metadata that becomes the system p…
Flare Redact, a new open-source JavaScript library, provides scoped secret and PII redaction for AI agents, detecting and masking sensitive data in logs, prompts, and text across 24 languages with zer…
Dynatrace engineer shares a bash script that installs dtctl, configures authenticated contexts from environment secrets, and installs the dtctl AI-assistant skill for Claude Code environments. The scr…
Ninjahawk released bongochat, a 125.8-million-parameter language model trained end to end with nanochat on a single RTX 5070 (12GB) GPU in about 2.5 hours with no cloud spend, and it ranks #1 on the G…
RagLeap has released RagLeap Core, an MIT-licensed, open-source RAG engine that powers its self-hosted AI business manager, allowing users to run a document-grounded chat system on their own infrastru…
Shrivara has released TokenBar, a free macOS menu bar app that tracks daily AI usage across Claude, Codex, OpenCode, and pi, displaying spend, tokens, and cache hit rates with odometer-style updates. …
DBoard, an open-source admin panel builder created by developer Haimanot25, uses AI to generate admin panels, dashboards, forms, and queries from natural language prompts for existing databases, suppo…
Mountain Owl released Bubo, a self-hosted AI code-review tool that reviews GitLab MRs and GitHub PRs using a bring-your-own-LLM approach, posting only actionable findings as inline threads. The tool s…
Cckeep, a new open-source tool by developer Kamihork, prevents Claude Code's Remote Control sessions from silently disconnecting by re-arming them within 31 seconds, addressing a known issue where ses…
9Router has released 9router-Python, a Python SDK and CLI for its AI gateway that routes requests to 35 models across providers including Gemini, NVIDIA, OpenRouter, Groq, Ollama, Tavily, and Exa. The…
Hedgehog, a new AI engineering workflow from developer skyf0xx, enforces test-driven development and opinionated architecture to keep AI-generated codebases structured as they grow. The tool, installe…
A developer has published a bypass for the Ghost Font protection mechanism on Mixfont, using a non-machine-learning attack that achieves a 0.9077 IoU mask overlap and reads 'GHOST'. The same attack ag…
Unsloth, an open-source AI startup, released Unsloth Studio (Beta), a platform that lets users run and train text, audio, embedding, and vision models locally on Windows, Linux, and macOS, with suppor…
A 2026 comparison of 10 AI website and app builders, maintained by WeInc, evaluates tools including Lovable, Bolt.new, v0, Replit Agent, 10Web, Hostinger AI, Framer AI, Durable, and Base44, focusing o…
A developer released 'write-normal', a Claude Code skill that strips recognizable signs of AI writing from prose, such as fake wisdom, rhetorical devices, inflated vocabulary, filler, sycophancy, and …
A developer released how.nvim, a Neovim plugin that answers questions about a user's configuration using the official Codex CLI, requiring Neovim 0.10+ and a ChatGPT or OpenAI API key. The plugin insp…
Robinduckett released Leclaude, a Windows shell extension that displays a robot badge on folders containing Claude Code session history, available for Windows 10 and 11 on x64 and ARM64. The tool, bui…
A new Claude Skill called vc-teardown, available for download from GitHub, stress-tests startup ideas by researching incumbents, attacking the idea with investor-style objections, and rebuilding it in…