Vannarot Roeung
Recent posts by Vannarot Roeung
Run Qwen3.6 2.5x Faster Locally With Unsloth NVFP4
Unsloth released NVFP4 quantized versions of Qwen3.6 that run up to 2.5x faster, with the 27B model fitting on a single 24GB GPU.
MuScriptor: Open Audio-to-MIDI Model for Full Mixes
Kyutai and Mirelo released MuScriptor, an open-weight model that transcribes a full song mix into separate, editable MIDI tracks, one per instrument.
aftr: Drive After Effects From Claude Code via MCP
aftr is an open-source MCP server that lets AI coding agents drive Adobe After Effects, apply one-call VFX presets, and self-correct renders through an autonomous loop.
Broll: An MCP Content Studio for Coding Agents
Broll is an open-source MCP server that gives coding agents like Claude Code and Codex a full content studio: generate, render, and publish video and carousels with your own keys.
ReviewFlow Brings GitLab Code Review Into Cursor
ReviewFlow is a new open-source extension that moves GitLab code review inside your editor, so you can draft comments and publish merge request reviews without leaving Cursor or VS Code.
Run GLM-5.2 (744B) Locally on 25GB RAM: Colibri
A developer released Colibri, a pure-C engine that runs GLM-5.2 (744B MoE) on a 25GB-RAM machine with no GPU by streaming experts from an NVMe SSD.
SWE-1.7: Cognition's Coding Model Nears Frontier
Cognition SWE-1.7 lands within a few points of GPT-5.5 and Claude Opus 4.8 on coding benchmarks, at a fraction of the cost. It is live in Devin at 1000 tokens per second.
Cursor 3.11: Side Chats and Agent Transcript Search
Cursor 3.11 introduces side chats, searchable agent transcripts, and programmable lifecycle hooks, turning the editor into a coordinated agent workspace.
NoimosAI Builds Self-Optimizing Sites With AI
NoimosAI opened its AI website builder to the public on July 10, generating SEO-focused sites from a chat prompt that keep optimizing over time.
Claude Code Desktop Gets a Sandboxed In-App Browser
Anthropic shipped a sandboxed, built-in browser inside the Claude Code desktop app on July 10, 2026, so the agent can read docs, click through live pages, and verify its own builds without your saved logins.
Drafted V2 Generates AI Home Plans in 2D and 3D
Drafted has released Drafted V2, a faster AI home-design tool that turns a list of rooms into complete 2D floor plans and interactive 3D models in seconds.
Microsoft Agent Framework for Go Enters Public Preview
Microsoft released a Go SDK for its Agent Framework on July 10, 2026, letting Go developers build tool-using, multi-agent workflows natively.
How to Build an Agentic Automation Pipeline on GitHub
An agentic automation pipeline watches your repository for an event, hands the change to a model with scoped tools, and returns a reviewable pull request. Build one from scratch on GitHub in under an hour.
OpenAI Discontinues ChatGPT Atlas Browser for New App
OpenAI is discontinuing ChatGPT Atlas, its standalone desktop browser, less than nine months after launch, folding its features into a unified ChatGPT app.
Grok 4.5 vs Opus 4.8 vs GPT-5.6: Coding Tested
Grok 4.5, Claude Opus 4.8, and GPT-5.6 Sol sit within a few points on the hardest agentic coding benchmarks. Here is which one to pick, and when.
GPT-5.6 Sol, Terra, Luna: OpenAI's New Model Tiers
OpenAI's GPT-5.6 ships as three tiers, Sol, Terra, and Luna, at different price and capability points. Here is how they compare and which to use.
LocalClip: Free Local-First AI Video Clipper for Mac
LocalClip is a new local-first AI video clipper for Mac that turns long recordings into vertical short-form clips entirely on your own machine.
LiteRT.js: Google's Browser AI Inference Runtime
Google released LiteRT.js on July 9, 2026, a JavaScript runtime that runs AI models directly in the browser with no server, up to 3x faster than other web runtimes.
Subagentmaxxing: Cheap AI Coders as Claude Subagents
Subagentmaxxing is a new open-source CLI that runs Codex, Cursor, and Grok as subagents under Claude Code, keeping flagship planning while cutting execution costs up to 25 times.
AlphaEvolve GA: Google's Code-Optimizing AI Agent
Google made AlphaEvolve generally available on July 9, 2026, moving its Gemini-powered algorithm-discovery agent onto the Gemini Enterprise Agent Platform for any Google Cloud customer.
GPT-5.6 Lands in Figma Make: Faster Design-to-Code
Figma added GPT-5.6 to Figma Make on July 9, 2026, improving first-pass design-to-code quality, self-healing fixes, and design fidelity.
ChatGPT Work: OpenAI's Agent That Builds Docs and Apps
OpenAI launched ChatGPT Work on July 9, 2026: a GPT-5.6 agent that runs for hours across your apps to build docs, dashboards, and web apps.
Meta Muse Spark 1.1: Meta's First Paid AI Coding API
Meta shipped its first paid AI model on July 9: Muse Spark 1.1, a paid agentic coding API at $1.25 per million input tokens with a $20 free credit.
AgentPrizm Gives Claude Code, Cursor Persistent Memory
AgentPrizm launched AgentMemory and AgentSkills, giving AI coding agents persistent, auditable memory and a marketplace of reusable, governed procedures over REST API and MCP.