Design
Preset
Background
Text
Font
Size
Width
Account Pricing Friday, July 24, 2026

The Git Times

“The major problems of the world are the result of the difference between how nature works and the way people think.” — Gregory Bateson

AI Models
Claude Fable 5 $50/M —GPT-5.6 Luna $6/M —Gemini 3.1 Pro Preview $12/M —Grok 4.5 $6/M —DeepSeek V4 Pro $0.87/M —Qwen3.7 Max $4.42/M —Kimi K3 $15/M —
Full Markets →
Fresh on Hugging Face

Model Drops

The newest model releases builders are picking up right now.
Fresh on GitHub

Just Shipped

Significant new releases from the AI and dev-infra repos builders run on.

Repowise Cuts Agent Context Load by 96% with Precomputed Code Intelligence 🔗

Local, self-hosted tool indexes codebases once to eliminate redundant AI exploration and reduce token usage

repowise-dev/repowise · Python · 4.2k stars 4mo old · Latest: v0.34.0

Repowise transforms how AI agents interact with codebases by precomputing answers to common developer questions—such as function callers, change impact, and code health—so agents no longer need to rediscover the codebase on every task. Instead of repeatedly grepping, reading, and forgetting, agents query a persistent index that delivers context instantly. The system reduces token usage for loading a single commit’s context from 64,039 tokens down to 2,391—a 96% cut—and slashes overall context re-read by 41% across multi-step investigations.

Built in Python and designed for local use, Repowise runs entirely on the user’s machine, requiring no API key for initial indexing and keeping code private. It exposes ten task-shaped MCP tools compatible with Claude Code, Codex, Cursor, and VS Code, enabling agents to retrieve rich, multi-target context in a single call rather than chaining sequential queries. Recent updates include a VS Code UI bundle, optimized knowledge-graph enrichment, deterministic architecture diagrams, and a Present mode for guided documentation walkthroughs. The tool also provides defect-validated code health scores, change-risk scoring per PR, and dead code detection, offering teams a shared view of code quality without relying on cloud services.
The catch: Despite its performance gains, Repowise remains early-stage with 77 open issues and a narrow focus on MCP-enabled agents, raising questions about its usability for teams not invested in AI-assisted development workflows.

Why this leads today Repowise unifies AI-powered code health scoring, documentation, and architectural insights via MCP, delivering measurable gains in developer productivity and codebase maintainability.

Use Cases
  • AI agents reducing token use during codebase exploration
  • Teams enforcing code health via automated PR risk scoring
  • Developers navigating legacy code with precomputed impact maps

Source: repowise-dev/repowise — based on the README and release notes.

More on the Front Page

OpenWorker Delivers Finished Work, Not Just Chat, on Your Desktop 🔗

The open-source AI agent runs locally, integrates with 25+ tools, and requires user approval before acting.

andrewyng/openworker · Python · 2.3k stars 4d old

OpenWorker is a desktop AI agent that turns natural language prompts into completed tasks—drafting documents, triaging inboxes, updating calendars, or generating Slack replies with live data. Built in Python and powered by aisuite, it runs a local agent server that connects to your files, terminal, and over 25 connectors for apps like Jira, GitHub, and Slack.

Unlike chatbots that suggest next steps, OpenWorker executes multi-step workflows and delivers polished outputs, only proceeding after explicit user approval for consequential actions. It supports any model provider—OpenAI, Anthropic, Google, or local Ollama—keeping data on your machine unless you choose to share it via integrated services. The app auto-updates on macOS and offers unsigned Windows builds, with code signing in progress. Recent work includes v0.1.6, released three days ago, reflecting active development amid 70 open issues.

The catch: OpenWorker remains in beta with unsigned Windows builds triggering SmartScreen warnings, and its reliance on user approval per action may slow automation for users seeking fully hands-off operation.

Use Cases
  • Developers drafting release notes from Jira and GitHub activity
  • Managers triaging overflowing inboxes into priority actions
  • Analysts generating formatted reports from live spreadsheet data

Source: andrewyng/openworker — based on the README and release notes.

Omnigent unifies AI agents across devices with policy controls 🔗

Orchestrate Claude, Codex, Cursor and custom agents with real-time sync and sandboxing

omnigent-ai/omnigent · Python · 7.7k stars 1mo old

Omnigent is a Python-based meta-harness that lets developers orchestrate multiple AI agents—including Claude Code, Codex, Cursor, and custom agents—through a single interface. It enables real-time collaboration across terminal, browser, phone, and desktop apps, with shared session state and file sync.

Users can supervise agent teams, enforce policies like spend caps or approval gates, and run agents in cloud sandboxes via Modal, Daytona, or Kubernetes. The v0.6.0 release adds Slack integration, cross-platform desktop apps, and richer file viewing. The catch: With 676 open issues and rapid development, the framework may still lack stability for production-critical agent workflows despite its ambitious scope.

Use Cases
  • Developers debugging code with coordinated Claude and Codex agents
  • Teams reviewing AI-generated outputs via shared session transcripts
  • Engineers testing agent policies in isolated cloud sandboxes before deployment

Source: omnigent-ai/omnigent — based on the README and release notes.

VS Code 1.130.0 Ships with Enhanced Remote Container Debugging 🔗

Monthly update improves containerized workflow stability and breakpoint reliability

microsoft/vscode · TypeScript · 187.8k stars Est. 2015

Microsoft released Visual Studio Code 1.130.

0, refining remote development for containerized applications. The update stabilizes debugging sessions in Docker and Kubernetes environments, reducing disconnections during long-running processes. Breakpoint handling now better respects conditional logic in multi-container setups, a pain point noted in recent issues. Insiders builds receive daily feature previews, while the stable channel delivers monthly fixes. Despite its maturity, the project carries significant technical debt from its Electron foundation, contributing to memory usage concerns in large workspaces.
The catch: Persistent high RAM consumption during complex multi-root editing remains unresolved, limiting usability on constrained hardware.

Use Cases
  • Developers debugging microservices in Kubernetes
  • Teams standardizing on containerized build environments
  • Extension authors testing language server integrations

Source: microsoft/vscode — based on the README and release notes.

GitHub CLI v2.96.0 patches Codespace Jupyter security flaw 🔗

Release adds public repo asset downloads and broader coding agent detection

cli/cli · Go · 45.4k stars Est. 2019

GitHub CLI’s latest release fixes a critical vulnerability in gh codespace jupyter that could allow arbitrary command execution when connecting to malicious Codespaces. Users must update to v2.

96.0 immediately. The update also enables gh release download to fetch assets from public repositories without authentication, aligning behavior with gh extension install. Additionally, the tool now detects more third-party coding agents and supports antigravity-cli and antigravity2.0 in the gh skill command. These changes improve usability and security for terminal-based GitHub workflows.
The catch: Over 1,000 open issues suggest ongoing stability and feature maturity challenges despite frequent updates.

Use Cases
  • Developers managing PRs and issues directly from terminal
  • Automating GitHub Actions workflows with `gh` commands
  • Installing and updating agent skills for AI coding assistants

Source: cli/cli — based on the README and release notes.

Kero Brings Native Terminal Workspace to macOS with Swift and libghostty 🔗

Built for developers seeking split panes, Git integration, and project-based file trees

egoist/kero · Swift · 228 stars 6d old

Kero is a new native terminal workspace for macOS, written in Swift and powered by libghostty. It offers split panes, Git integration, and a file tree grouped by projects, aiming to streamline developer workflows.

The app downloads from kero.sh and is released under GPLv3. Active development is evident, with the last push just days ago and 22 forks indicating early interest. The catch: With 8 open issues and a very recent launch, long-term stability and feature maturity remain unproven for daily professional use.

Use Cases
  • Developers managing multiple project terminals
  • Teams preferring native macOS GUI over Electron apps
  • Users wanting Git status visible in terminal panes

Source: egoist/kero — based on the project README.

LLVM 22.1.8 release delivers updated compiler toolchain for developers 🔗

Latest point release includes prebuilt binaries for Linux, macOS, and Windows x64

llvm/llvm-project · LLVM · 39.4k stars Est. 2016

The LLVM Project released version 22.1.

8, providing prebuilt toolchains for Linux x86_64, Linux Arm64, macOS Apple Silicon, and Windows x64. This update maintains the modular compiler infrastructure used to build C, C++, and other languages via Clang, with core tools for bitcode optimization and object file generation. Developers rely on LLVM for performance-critical compilation workflows across platforms. The catch: Despite frequent commits, the project carries over 37,000 open issues, indicating ongoing maintenance challenges in a rapidly evolving compiler landscape.

Use Cases
  • Compiler engineers building C/C++ toolchains
  • Performance-focused developers optimizing release builds
  • Toolchain creators integrating custom language frontends

Source: llvm/llvm-project — based on the README and release notes.

NiceChunk Game Client Integrates Browser Play with Blockchain World Sync 🔗

Uses Chunk.js engine and modular SDK for decentralized game state and asset handling

nicechunk/game · JavaScript · 215 stars 4d old

The NiceChunk game client runs in the browser, syncing gameplay with blockchain state via the play/ directory’s runtime systems and chain synchronization. It relies on src/ for shared data and wallet support, while sdk/ handles Guardian decoding and chain accounts.

Assets and configs live in public/, with the core engine pinned as a submodule in chunk.js/. Setup requires git submodule update --init --recursive, followed by npm install, npm run check, and npm run build. The production game is hosted on the main NiceChunk site, which manages login and routing.
The catch: The client depends on external login and site routes, limiting standalone use or self-hosted deployment flexibility.

Use Cases
  • Developers build browser-based games with on-chain asset persistence
  • Teams integrate wallet login and Guardian identity verification into gameplay
  • Creators deploy modular game clients using shared SDKs and pinned engines

Source: nicechunk/game — based on the project README.

AI Agents Shift from Assistants to Orchestrated Workflows in Open Source 🔗

New frameworks enable multi-agent collaboration, skill sharing, and verifiable execution across domains like video, code, and enterprise automation

Trendai-agents
powerycy/goutoujunshijoeseesun/qiaomu-cut-skillRinDig/icm-architectomnigent-ai/omnigentdeepset-ai/haystack+28 more

Open source is rapidly evolving beyond standalone AI assistants toward orchestrated agent ecosystems where specialized skills are composed, shared, and governed. Projects like omnigent-ai/omnigent act as meta-harnesses, letting users swap between Claude Code, Codex, and custom agents while enforcing policies and sandboxing.

Similarly, stablyai/orca provides a TypeScript-based ADE for managing fleets of parallel agents, and agentlas-ai/Agentlas-OS offers a local-first orchestrator that spins up temporary agent hubs per task.

This pattern extends into domain-specific automation: joeseesun/qiaomu-cut-skill delivers a bilingual, brand-governed video director agent with verifiable rendering, while calesthio/OpenMontage positions itself as the first agentic video production system with 500+ skills. In enterprise settings, iflytek/skillhub enables self-hosted skill registries with RBAC and audit logs, and iflytek/astron-rpa packages agent-ready RPA tools for scalable automation. Even niche areas like hardware design are being touched—earthtojake/text-to-cad provides agent skills for CAD and robotics workflows.

The catch: While promising, this agent orchestration layer remains fragmented—many frameworks compete without interoperability standards, skill reuse is often limited to specific LLMs or runtimes, and verifiable execution claims are still early-stage, raising concerns about real-world reliability and vendor lock-in beneath open-source licenses.

Use Cases
  • Developers orchestrating multi-agent coding workflows across devices
  • Enterprises deploying governed, auditable AI skills internally
  • Creators automating end-to-end video production with AI agents

Open Source Agents Evolve Into Modular Skill Ecosystems 🔗

Projects now package LLM capabilities as reusable, composable skills — turning AI agents into customizable workflow engines.

Trendllm-tools
RinDig/icm-architectomnigent-ai/omnigentCookSleep/gpt_image_playgrounddeepset-ai/haystackMadsLorentzen/ai-job-search+18 more

A clear pattern is emerging in open source: LLM-powered tools are shifting from monolithic assistants to modular skill ecosystems. Repositories like omnigent-ai/omnigent and deepset-ai/haystack exemplify this by framing agents as orchestration layers where skills — discrete, reusable units of behavior — can be swapped, combined, and governed.

omnigent lets users harness Claude Code, Codex, or Cursor through a unified meta-harness, enforcing policies and enabling real-time collaboration without rewriting agent logic. Similarly, haystack provides explicit control over retrieval, routing, and generation, allowing developers to build production-grade RAG and agent workflows with composable pipelines.

This modularity extends to domain-specific skills. AgriciDaniel/claude-seo breaks SEO into 25 sub-skills and 18 sub-agents, while coreyhaines31/marketingskills packages CRO, copywriting, and analytics as deployable abilities for Claude Code. Even niche capabilities like video understanding (bradautomates/claude-video) or stock analysis (ZhuLinsen/daily_stock_analysis) are encapsulated as invokable skills. Frameworks such as microsoft/SkillOpt go further, optimizing and packaging natural-language skills as best_skill.md artifacts for frozen LLMs — suggesting a future where skills are versioned, shared, and traded like code libraries.

The catch: While promising, this skill-based paradigm risks fragmentation — incompatible skill formats, unclear versioning, and weak sandboxing could hinder reuse. Many projects remain prototypes, lacking governance or interoperability standards, making large-scale adoption uncertain despite rapid experimentation.

Use Cases
  • Developers orchestrate multi-agent workflows using swap-ready skills
  • Marketers deploy SEO and copywriting skills via Claude Code extensions
  • Researchers build customizable RAG pipelines with explicit skill routing

CLI Agents and Terminal Workspaces Reshape Dev Tooling 🔗

Open source unifies AI, code intelligence, and agent orchestration directly in the terminal

Trenddev-tools
egoist/kerowemake-services/wemake-python-styleguideomnigent-ai/omnigentcli/clirepowise-dev/repowise+17 more

A clear pattern is emerging in open-source dev tools: the terminal is becoming the central hub for AI-augmented development. Projects like egoist/kero offer native macOS terminal workspaces that streamline workflows, while agent-focused tools such as MoonshotAI/kimi-cli, can1357/oh-my-pi, and ogulcancelik/herdr act as multiplexers or interfaces for swapping between AI agents like Claude Code, Codex, and custom models without context switching.

Frameworks like omnigent-ai/omnigent and HKUDS/CLI-Anything go further, providing meta-harnesses to orchestrate, policy-enforce, and sandbox multiple agents in real time from any device. Code intelligence is also shifting left into the CLI: tirth8205/code-review-graph builds persistent, local-first code graphs to reduce AI context load during reviews, and repowise-dev/repowise delivers code health scores, auto-docs, and dead code detection via MCP. Even established tools like cli/cli and microsoft/vscode are adapting to this shift toward terminal-first experiences. This convergence signals a future where developers manage AI collaboration, code understanding, and execution entirely through composable, terminal-native interfaces.
The catch: Much of this remains experimental—agent orchestration lacks standardized protocols, context reduction techniques vary widely in effectiveness, and integrating multiple agents introduces latency and complexity that may outweigh benefits in smaller teams or legacy workflows.

Use Cases
  • Developers orchestrate multiple AI agents via terminal
  • Teams reduce AI context load with local code graphs
  • Enforce policies and sandbox AI agents in real time

Deep Cuts

OpenBrowser: Local Chromium Fingerprinting for Secure Multi-Profile Automation 🔗

Isolate environments with custom proxies, fingerprints, and RPA in a lightweight Electron-based tool

lyu0805/OpenBrowser · JavaScript · 301 stars

lyu0805/OpenBrowser is a local fingerprint browser built with JavaScript and Electron, designed to give developers fine-grained control over browser identity and behavior. It enables isolated Chromium environments where each profile can have unique fingerprints, proxy settings, and cookie stores — ideal for testing, scraping, or managing multiple accounts without cross-contamination.

With built-in RPA support, users can automate workflows across profiles using simple scripts. The project emphasizes privacy and flexibility, letting developers simulate real-user conditions at scale without relying on external services or complex infrastructure.
The catch: It's early-stage, primarily documented in Chinese, and lacks broad community testing or enterprise-grade stability.

Use Cases
  • QA engineers testing geo-specific web app behavior
  • Developers managing multiple social media accounts securely
  • Automation specialists running parallel RPA workflows with distinct fingerprints

Source: lyu0805/OpenBrowser — based on the project README.

AI Video Director Automates Editing with Verifiable Rendering 🔗

Combines FFmpeg, subtitles, and brand templating for reproducible AI-powered video workflows

joeseesun/qiaomu-cut-skill · JavaScript · 263 stars

joeseesun/qiaomu-cut-skill is an agent-native Skill for governed video production. It intelligently sources materials, generates bilingual subtitles, applies brand packaging, and renders outputs via FFmpeg—all with verifiable, reproducible results.

Built in JavaScript/Node.js, it treats video editing as a composable skill for AI agents, enabling consistent, audit-ready pipelines. Developers can plug it into larger agent systems to automate everything from social clips to localized explainers, trusting that each render follows predefined rules and source provenance.

Use Cases
  • AI agents generating multilingual marketing videos
  • Automating branded social content with subtitle overlays
  • Creating reproducible tutorial videos from structured data

Source: joeseesun/qiaomu-cut-skill — based on the project README.

Quick Hits

tau HuggingFace’s Tau is a minimalist Python coding agent ported from Pi, enabling lightweight, autonomous code generation and editing for developers seeking simplicity in AI-assisted programming. 1.9k
redis-poc Redis-PoC demonstrates critical RCE vulnerabilities in Redis versions 6.2.22 through 8.8.0, serving as a security research tool to validate and mitigate remote code execution risks. 356
VibeVoice Microsoft’s VibeVoice is an open-source frontier voice AI model enabling high-factory system for realistic, controllable speech synthesis for applications requiring natural, expressive audio generation. 50.5k
Beyond GitHub

The AI Wire

What builders are reading today — the headlines, papers, and announcements that aren't trending repos.

From the labs & arXiv

Netdata v2.10.4 bolsters macOS monitoring with hardware sensor collectors 🔗

Patch release adds GPU, thermal, and power metrics for Apple systems via new collectors

netdata/netdata · Go · 79.8k stars Est. 2013 · Latest: v2.10.4

Netdata’s latest patch release v2.10.4 introduces macOS-specific hardware sensor collectors, extending its real-time observability to Apple hardware.

The update adds GPU power, clock, and temperature monitoring, SMC and IOHID sensor data, fan speeds, power source status, thermal pressure, and NVMe SMART metrics. A powermetrics fallback via ndsudo ensures compatibility where direct hardware access is restricted, while a new cross-OS sensors function and shared temperature-histogram context improve consistency across platforms.
Beyond macOS, the release focuses on stability: the database engine now guards against corrupted or truncated data files with extent validation, SIGBUS protection, and improved journal error handling. Bounded-cardinality process grouping for the apps.plugin on macOS prevents metric explosion on systems with high process churn, preserving meaningful per-process monitoring.
These changes address long-standing gaps in Netdata’s Apple ecosystem support while reinforcing its core promise of zero-configuration, per-second insights with minimal overhead. The project remains committed to local, secure monitoring without central data aggregation, appealing to teams prioritizing data sovereignty.
The catch: macOS hardware monitoring relies on ndsudo for privileged sensor access, introducing a dependency and potential security consideration for locked-down environments.

Previously in The Times “covered” — Jul 20

Use Cases
  • DevOps teams monitoring Mac mini servers in edge deployments
  • SREs tracking GPU utilization on macOS-based ML workstations
  • Developers diagnosing thermal throttling on Apple Silicon laptops during builds

Source: netdata/netdata — based on the README and release notes.

More Stories

RAG Techniques repo adds visual book for code-free learning 🔗

Companion guide explains 22 RAG methods with diagrams before code

NirDiamant/RAG_Techniques · Jupyter Notebook · 28.8k stars Est. 2024

The NirDiamant/RAG_Techniques repository now includes RAG Made Simple, a visual companion book that teaches 22 production-grade RAG techniques through diagrams and plain-English analogies before requiring code. Built alongside the repo’s 42+ runnable Jupyter notebooks, it covers foundations, smarter retrieval, advanced architectures, and cutting-edge methods like Graph RAG and agentic RAG.

The book targets AI engineers and developers seeking to understand RAG fundamentals without boilerplate overload. Subscribers get early access and discounts on the book and related course.
The catch: The book’s code-free approach may not suit developers who learn best by implementing techniques directly in LangChain or LlamaIndex environments.

Use Cases
  • AI engineers learning RAG fundamentals visually
  • Teams evaluating retrieval methods for production deployment
  • Developers preparing for LangChain or LlamaIndex RAG projects

Source: NirDiamant/RAG_Techniques — based on the README and release notes.

Microsoft MCP Curriculum Adds Rust and TypeScript Examples for AI Workflows 🔗

Updated lessons now cover six languages with focus on modular, scalable service orchestration

microsoft/mcp-for-beginners · Jupyter Notebook · 16.8k stars Est. 2025

The microsoft/mcp-for-beginners project expanded its hands-on curriculum to include Rust and TypeScript implementations of Model Context Protocol patterns. Developers can now explore session setup, client-server communication, and security practices across .

NET, Java, JavaScript, Python, Rust, and TypeScript using Jupyter Notebooks. The update emphasizes practical techniques for building modular AI workflows from service definition to orchestration. Recent commits show active maintenance with the last push 24 days ago and ongoing issue triage.
The catch: The curriculum assumes foundational knowledge of asynchronous programming and API design, which may challenge beginners without backend experience.

Previously in The Times “covered” — Jul 17

Use Cases
  • Backend engineers learning MCP service implementation in Rust
  • Full-stack developers comparing MCP clients across six languages
  • AI practitioners building secure, scalable workflow orchestration pipelines

Source: microsoft/mcp-for-beginners — based on the project README.

Gemini CLI Adds GitHub Action Handler for Automated Egress Workflows 🔗

New caretaker-egress module enables Octokit integration for issue triage and code actions

google-gemini/gemini-cli · TypeScript · 106.1k stars Est. 2025

Google’s gemini-cli v0.52.

0 introduces a caretaker-egress feature that uses Octokit to handle GitHub Actions for automated issue egress. Developers can now trigger Gemini-powered workflows directly from PR events, enabling AI-assisted code reviews or triage of bugs or feature requests without leaving the terminal. The update also refactors CI file handling to reduce workspace noise and improves quota error messaging for shared projects. Built-in tools like Google Search grounding and file operations remain core to its terminal-first design. The catch: Heavy reliance on MCP for extensibility may complicate debugging for teams unfamiliar with Model Context Protocol standards.

Previously in The Times “covered” — Jul 20

Use Cases
  • Automate GitHub issue triage using Gemini AI in CI pipelines
  • Run shell commands with AI-assisted context from terminal prompts
  • Ground code generation in live Google Search results for accuracy

Source: google-gemini/gemini-cli — based on the README and release notes.

Quick Hits

yolov5 Ultralytics YOLOv5 delivers real-time object detection, segmentation, classification, and export in PyTorch — ideal for builders needing fast, production-ready vision AI. 57.7k
superpowers Obra/superpowers provides an agentic skills framework that turns developer intuition into repeatable, scalable workflows — no fluff, just working methodology. 260.3k
ComfyUI ComfyUI offers a modular, node-based GUI and API for diffusion models — giving builders full control over complex generative AI pipelines without code lock-in. 122.1k
h4cker The-Art-of-Hacking/h4cker curates thousands of ethical hacking resources — from DFIR to AI security — serving as a builder’s deep-reference library for offensive and defensive security. 28.6k
transformers Hugging Face Transformers is the unified framework for SOTA text, vision, audio, and multimodal models — enabling seamless inference and training across modalities. 162.9k

Autoware Universe expands autonomous driving toolkit with ROS2 planner upgrades 🔗

Latest release adds trajectory validation, lane departure checks, and CARLA interface refinements

autowarefoundation/autoware_universe · C++ · 1.7k stars Est. 2021 · Latest: 0.52.0

The Autoware Foundation’s universe repository released version 0.52.0, integrating new motion planning and safety modules for ROS2-based self-driving systems.

Key additions include a trajectory validator that filters uncrossable boundary departures and a behavior path side-shift node that enforces drivable area checks to prevent lane excursions. Perception saw updates to the multi-object tracker with polar coordinate association and image projection-based fusion now applied to CIE environments. The CARLA interface gained lightweight sensor configs, spectator follow scripting, and improved vehicle command gating for simulation workflows. Builders can leverage these updates for highway merging, urban navigation, and sensor fusion pipelines requiring precise route adherence. Code coverage metrics remain uneven across components, with several modules like Launch and Evaluator still marked TBD in the project’s transparency report.
The catch: Despite active development, nearly half the subsystem coverage data is unmeasured, raising concerns about reliability in safety-critical paths like control and perception where gaps could mask regressions.

Previously in The Times “covered” — Jul 21

Use Cases
  • Urban AV planners testing lane-keeping compliance
  • Simulation engineers refining CARLA-based sensor suites
  • Perception teams fusing LiDAR and camera data in ROS2 stacks

Source: autowarefoundation/autoware_universe — based on the README and release notes.

More Stories

Isaac Lab 3.0 Beta 2 Patch 1 updates for Isaac Sim 6.0.1 compatibility 🔗

Patch bumps h5py, fixes streaming crash, and aligns dependencies for Omniverse workflows

isaac-sim/IsaacLab · Python · 7.8k stars Est. 2022

Isaac Lab’s latest patch release updates compatibility with NVIDIA Isaac Sim 6.0.

1, addressing a streaming crash and refining Docker image setup. The patch bumps h5py to >=3.16.0 and cherry-picks a fix from Isaac Sim to stabilize Omniverse Kit extensions. These changes support NuRec workflows and improve reliability for robot learning pipelines using RL frameworks like RSL RL and SKRL. While the framework accelerates GPU-based simulation for sim-to-real transfer, its tight coupling to specific Isaac Sim versions creates version-locking risks for teams managing heterogeneous environments.
The catch: Isaac Lab requires exact Isaac Sim version alignment, complicating upgrades in shared or cloud-based infrastructures.

Previously in The Times “covered” — Jul 18

Use Cases
  • Train quadruped locomotion policies in simulation
  • Test multi-agent manipulation with vision sensors
  • Deploy RL workflows across local and cloud GPU clusters

Source: isaac-sim/IsaacLab — based on the README and release notes.

CADAM v0.3.0 Adds Agent Updates and Logo Redesign 🔗

Release focuses on AI generation tweaks and visual identity over core functionality

Adam-CAD/CADAM · TypeScript · 4.9k stars 10mo old

The latest CADAM release v0.3.

0 includes agent system updates and a redesigned logo, per recent commits. Core features remain browser-based text-to-CAD generation using WebAssembly, with parametric controls and STL/SCAD/DXF export. The app still runs live at adam.new/cadam, transforming prompts into OpenSCAD models. No new export formats or major AI model upgrades were added in this cycle. The catch: Open issues remain at 20, suggesting unresolved bugs or feature requests despite recent activity.

Previously in The Times “covered” — Jul 17

Use Cases
  • Engineers generate parametric parts from text prompts
  • Designers iterate models using interactive sliders
  • Teams export browser-made CAD for 3D printing workflows

Source: Adam-CAD/CADAM — based on the README and release notes.

Quick Hits

open-source-rover Build a functional 6-wheel Mars-inspired rover yourself with open-source plans from NASA JPL, perfect for hands-on robotics learning and prototyping. 9.5k
rosclaw Evolve AI agents in real-world robot bodies with rosclaw’s self-evolving runtime, enabling safe, adaptive physical intelligence through capability routing and skill growth. 170
nicegui Craft elegant, responsive web UIs entirely in Python with nicegui — no HTML/CSS/JS needed, just pure Python for fast, maintainable interfaces. 16.1k
carla Test and validate autonomous driving algorithms in a high-fidelity, open-source simulator that replicates complex urban and rural driving scenarios. 14.2k
openpilot Upgrade your car’s driver assistance with openpilot — an open-source OS that adds advanced autonomous features to 300+ supported vehicle models. 63.2k
icub-main Develop and control the iCub humanoid robot’s full sensorimotor capabilities using its main software stack, enabling advanced cognitive and physical AI research. 120

OpenCTI Platform Redesigns Connector Catalog as Faceted Marketplace in Latest Release 🔗

Version 7.260722.0 removes legacy editor, enables workflow resets, and adds configurable webhook deny lists

OpenCTI-Platform/opencti · TypeScript · 9.7k stars Est. 2018 · Latest: 7.260722.0

OpenCTI-Platform/opencti has released version 7.260722.0, introducing a redesigned connector catalog as a faceted marketplace experience to streamline integration discovery and configuration.

The update removes the legacy editor and its feature flags, simplifying the frontend while improving GraphQL playground usability with light mode support. Workflow enhancements now allow full reset capability and prevent publishing workflows that delete active statuses, reducing operational risk. Security and reliability are bolstered by improved Express HTTP configuration and a new configurable deny list for webhook URIs, helping teams block malicious or unintended endpoints. CSV export has also been refined for better data portability. Built on TypeScript with a STIX2-based knowledge schema, OpenCTI remains a dual-edition platform (Community and Enterprise) for managing technical and non-technical threat intelligence, linking observables to sources like MISP reports and MITRE ATT&CK frameworks.
The catch: Despite active development, the project carries 1,996 open issues, indicating significant technical debt or unresolved feature gaps that may affect stability in high-throughput or regulated environments.

Previously in The Times “covered” — Jul 23

Use Cases
  • Security analysts correlating TTPs with observed malware via STIX2
  • SOC teams automating threat data ingestion from MISP and TheHive
  • Enterprises enforcing password policies and workflow governance at scale

Source: OpenCTI-Platform/opencti — based on the README and release notes.

More Stories

PayloadsAllTheThings adds reverse proxy and PHP extract() vulnerability guides 🔗

Release 4.2 expands web app security cheatsheet with new exploitation techniques and bug fixes

swisskyrepo/PayloadsAllTheThings · Python · 79.4k stars Est. 2016

The project updated its knowledge base with dedicated sections for External Variable Modification (PHP extract() flaws) and Reverse Proxy Misconfigurations (Nginx focus). Command Injection gained worstfit and fullwidth bypass techniques; CSV Injection added Google Sheets exfiltration via IMPORTXML/IMPORTRANGE; File Inclusion integrated the lightyear tool for blind reads.

Numerous formatting issues and broken links were fixed. Despite active maintenance, the sheer breadth risks overwhelming newcomers seeking targeted, curated guidance.
The catch: The encyclopedic scope may dilute focus for builders needing precise, battle-tested vectors over exhaustive theory.

Use Cases
  • Red teamers crafting web app exploit chains
  • Bug bounty hunters testing for common vulnerabilities
  • Security teams validating internal pentest checklists

Source: swisskyrepo/PayloadsAllTheThings — based on the README and release notes.

Infisical adds Azure auto-rotation for app connection credentials 🔗

New feature enables automatic secret rotation for Azure services, reducing manual credential management overhead

Infisical/infisical · TypeScript · 28.3k stars Est. 2022

Infisical's latest release v0.162.

13 introduces automated rotation of Azure app connection credentials, allowing teams to periodically refresh secrets for services like Azure Storage and Key Vault without manual intervention. The update also includes LDAP credential rotation and improved GCP secret sync region configuration, expanding automated secret lifecycle management across cloud platforms. These enhancements reduce operational improvements target DevOps teams seeking to minimize credential-related security risks through automation. The catch: Despite active development with 663 open issues, the platform's self-hosted deployment complexity may challenge smaller teams lacking dedicated infrastructure expertise.

Previously in The Times “covered” — Jul 18

Use Cases
  • DevOps teams automating Azure credential rotation
  • Security engineers managing GCP secret replication policies
  • Platform teams implementing LDAP-based secret rotation workflows

Source: Infisical/infisical — based on the README and release notes.

Quick Hits

openzeppelin-contracts OpenZeppelin Contracts provides secure, audited Solidity libraries to build reliable smart contracts with proven security patterns. 27.2k
cve Trickest/cve automates collection and updating of the latest CVEs with proof-of-concept exploits for proactive vulnerability research. 8k
caldera Apache Caldera enables automated adversary emulation to test and improve enterprise defenses against real-world attack techniques. 7.1k
opennhp OpenNHP delivers a lightweight, cryptography-driven Zero Trust toolkit to secure infrastructure, apps, and data in AI-era environments. 13.8k
maigret Maigret aggregates detailed personal dossiers from 3000+ sites using just a username for deep OSINT investigations. 35.7k
hosts StevenBlack/hosts consolidates trusted hosts files and adds optional filters to block ads, malware, porn, and social tracking at the network level. 30.8k

Bun v1.3.14 sharpens TypeScript and JSX handling for faster dev cycles 🔗

The Rust-powered runtime drops Node.js compatibility layers to cut startup overhead

oven-sh/bun · Rust · 95k stars Est. 2021 · Latest: bun-v1.3.14

Bun v1.3.14 refines its all-in-one JavaScript toolkit with tighter TypeScript inference and JSX transformation, reducing the need for external compilers like Babel or ts-node in React projects.

Built on JavaScriptCore and written in Rust, the runtime executes bun run index.tsx without configuration, skipping the typical Node.js module resolution dance. The release streamlines bun test and bun install commands, claiming sub-second startup times for scripts that previously lagged in Node.js environments.
Benchmarks show Bun’s built-in test runner outperforms Jest by 2–3x in cold starts, while its package manager avoids the 1,000-file node_modules bloat by symlinking dependencies directly. Developers can still bunx cowsay 'Hello, world!' to execute packages globally, maintaining npm-like ergonomics. The upgrade path remains simple: bun upgrade pulls v1.3.14, with canary builds available via bun upgrade --canary.
The catch: Despite its speed, Bun’s Rust core and JavaScriptCore foundation mean native module support lags behind Node.js — developers relying on node-gyp compiled addons may hit compatibility walls requiring workarounds or abstention.

Previously in The Times “covered” — Jul 18

Use Cases
  • React devs skip Babel for JSX/TS
  • CLI tool authors cut startup latency
  • Teams replace Jest with bun test

Source: oven-sh/bun — based on the README and release notes.

More Stories

Ventoy 1.1.16 fixes UEFI secure boot and legacy firmware boot issues 🔗

Latest release resolves Kicksecure compatibility and option reset bugs

ventoy/Ventoy · C · 78.2k stars Est. 2020

Ventoy 1.1.

16 patches boot failures on older UEFI firmware with secure boot disabled and resolves a Kicksecure Linux boot issue. It also prevents the VTOY_WIN_UEFI_RES_LOCK option from resetting when entering VentoyPlugson. The tool continues to let users boot multiple ISO, WIM, IMG, VHD, or EFI files from a single USB drive without reformatting, supporting legacy BIOS, UEFI (x86, x86_64, ARM64, MIPS64EL), MBR, and GPT. Over 1,300 ISOs are tested, including Windows, Linux, and ChromeOS variants.
The catch: Active development shows 991 open issues, indicating ongoing stability and compatibility challenges across diverse firmware and OS combinations.

Use Cases
  • Sysadmins booting multiple rescue ISOs from one USB drive
  • Developers testing Linux distros without repeated USB reformatting
  • IT teams deploying Windows and WinPE images via unattended install

Source: ventoy/Ventoy — based on the README and release notes.

Codebase-memory-mcp v0.9.0 delivers first-class Windows support and indexing gains 🔗

Release adds crash-resistant indexing, memory safety fixes, and scalable 3D graph UI

DeusData/codebase-memory-mcp · C · 34.7k stars 4mo old

The v0.9.

0 release of codebase-memory-mcp brings first-class Windows support with non-ASCII path handling and native windows-arm64 builds. Indexing performance improved ~61% on kernel-scale repos via deterministic output and a crash supervisor that quarantines bad files instead of aborting. Memory safety gains include accurate RSS accounting and bounded parallel retention for low-RAM machines. The graph UI now supports up to 10M nodes with dead-code filtering, code previews, and GitHub deep-links. Extraction accuracy received broad fixes across C/C++, Python, Go, and TypeScript. The catch: Despite performance claims, real-world adoption beyond AI agent integrations remains unverified at scale.

Previously in The Times “covered” — Jul 18

Use Cases
  • AI agents indexing large monorepos for context-aware code suggestions
  • Developers visualizing call chains and HTTP routes in legacy systems
  • Teams enforcing codebase consistency via persistent knowledge graph queries

Source: DeusData/codebase-memory-mcp — based on the README and release notes.

RAGFlow Adds Snowball Stemmer for 16 Languages in v0.26.4 🔗

New tokenization feature improves multilingual document processing for AI agents

infiniflow/ragflow · Go · 85.9k stars Est. 2023

RAGFlow’s latest release introduces a language-aware Snowball stemmer supporting 16 languages, integrated into its tokenization pipeline with a new language parameter. Dutch language support was also added to the frontend.

The update resolves several bugs, including crashes from LM-Studio model names with '@' symbols and MCP server failures due to paginated API responses. Docling parser now correctly extracts mathematical formulas instead of dropping them. These changes enhance RAGFlow’s ability to process diverse, structured data for context-aware LLM applications.
The catch: Despite rapid feature growth, over 2,100 open issues suggest ongoing stability challenges in complex enterprise deployments.

Previously in The Times “covered” — Jul 18

Use Cases
  • Enterprises processing multilingual technical documentation
  • Developers building AI agents with domain-specific data sources
  • Teams integrating RAG with Confluence, Notion, or S3 workflows

Source: infiniflow/ragflow — based on the README and release notes.

Quick Hits

memos A self-hosted, Markdown-native note-taking tool for instant capture and full data ownership — ideal for builders who value speed and privacy. 61.8k
starship A lightning-fast, infinitely customizable shell prompt that enhances developer experience with minimal overhead and maximum flair. 59k
lazygit A simple, intuitive terminal UI that streamlines Git workflows — making version control faster and less error-prone for daily coding. 80.7k
traefik A dynamic, cloud-native reverse proxy that automates service discovery and routing — simplifying microservices deployment and observability. 64.1k
linux The foundational source code of the Linux kernel — the ultimate reference for understanding low-level systems, drivers, and OS architecture. 240.3k

HackRF Pro Gains Larger SPI Flash Access in Latest Firmware Update 🔗

Release v2026.01.3 resolves mixer frequency lock failures and expands storage for complex waveforms

greatscottgadgets/hackrf · C · 7.9k stars Est. 2012 · Latest: v2026.01.3

The greatscottgadgets/hackrf repository released v2026.01.3, addressing a critical mixer frequency lock failure that could disrupt signal stability during transmission.

This fix improves reliability for developers using HackRF in precision RF applications like wireless protocol testing or spectrum monitoring.
The update also adds software access to the larger SPI flash on HackRF Pro models, enabling users to store bigger firmware images or waveform libraries directly on the device. This expands capability for standalone operation without constant host dependency.
Documentation remains buildable via Sphinx, with PDF generation supported on Ubuntu through make latexpdf in the docs folder. Community support continues through GitHub issues labeled "technical support," with Great Scott Gadgets employees targeting two-week response times.
The catch: Despite ongoing development, the project maintains no formal support staff, and response times for non-technical-support issues remain undefined, leaving users reliant on community triage for urgent problems.

Previously in The Times “covered” — Jul 23

Use Cases
  • RF engineers testing wireless protocols
  • Hobbyists capturing and analyzing satellite signals
  • Researchers developing custom spectrum sensing tools

Source: greatscottgadgets/hackrf — based on the README and release notes.

More Stories

Mural plotter turns SVGs into wall drawings with JavaScript control 🔗

Recent commit shows ongoing maintenance despite 4.3-year project age

nikivanov/mural · JavaScript · 283 stars Est. 2022

Mural converts SVG files into coordinate commands for a low-cost wall plotter built with JavaScript. Users set pin distance in the UI, triggering automatic scaling and margin adjustments before the plotter draws line by line.

The project processes SVGs treating each unit as a millimeter, then uploads movement data to a microcontroller. Last updated just one day ago, the repo shows active maintenance with 19 open issues and 64 forks. The catch: reliance on browser-based UI and microcontroller upload may complicate standalone operation for field use.

Use Cases
  • DIY artists creating large-scale wall murals from digital designs
  • Educators teaching coordinate geometry through physical drawing
  • Makers prototyping pen plotter mechanics without CNC complexity

Source: nikivanov/mural — based on the project README.

Stack-chan v1.0.0 Unifies Firmware, Simulation, and Web UI for Kawaii Robots 🔗

Release integrates browser-based flashing, face editing, and Blockly programming with Moddable SDK 8.3.1

stack-chan/stack-chan · TypeScript · 1.6k stars Est. 2021

The stack-chan/stack-chan project released v1.0.

0, its first stable version after five years of development. This release unifies firmware for M5Stack CoreS3, Core2, and RT variants into a single workflow, featuring a browser-based UI for flashing, face customization, and visual programming via Blockly. A WebAssembly simulator now lets users test motions, expressions, and voice synthesis—including Stack-chan Voice, VOICEVOX, and ElevenLabs—directly in the browser. The build system standardizes on Moddable SDK 8.3.1 and Node.js 22 for reproducible CI. The catch: Open issues remain at 79, and migrating from pre-v1.0.0 requires rebuilding custom MODs due to breaking changes in firmware structure and context API.

Previously in The Times “covered” — Jul 18

Use Cases
  • Developers build expressive M5Stack robots with TypeScript
  • Educators teach robotics via browser-based Blockly programming
  • Hobbyists simulate and customize robot faces before hardware assembly

Source: stack-chan/stack-chan — based on the README and release notes.

Tulip CC adds web-based Python synth access for instant music creation 🔗

Users can now run Tulip Creative Computer in-browser without hardware purchase

shorepine/tulipcc · C · 950 stars Est. 2022

The Tulip Creative Computer project now offers a fully functional web version, letting users code synthesizers, graphics, and interactive scripts directly in a browser using MicroPython and LVGL. Built on the same ESP32-S3 firmware, the browser port replicates core features including real-time audio synthesis via AMY, touchscreen emulation, and MIDI-over-WebMIDI support.

This lowers the barrier to entry for musicians and makers wanting to experiment before committing to hardware. The AMYboard variant also runs online, expanding access to its advanced audio I/O capabilities. Development remains active, with the last commit just days ago and rolling OTA updates pushed to main.
The catch: Web versions lack hardware sensor integration and real-time deterministic timing critical for stage performance or physical interfacing.

Previously in The Times “covered” — Jul 18

Use Cases
  • Music educators teach synthesis concepts via browser
  • Prototypers test Tulip code before buying hardware
  • Developers debug AMYboard audio patches remotely

Source: shorepine/tulipcc — based on the README and release notes.

Quick Hits

p3a fabkury/p3a: A lightweight pixel art player running on ESP32-P4, enabling retro-style visuals on low-power microcontrollers without external displays. 87
AIOsense Schluggi/AIOsense: An ESPHome-powered all-in-one sensor suite combining temperature, humidity, pressure, and motion detection for easy smart home integration. 160
gdsfactory gdsfactory/gdsfactory: A Python library streamlining hardware design for photonics, quantum, MEMS, and PCBs with intuitive, code-driven layout generation and simulation-ready outputs. 989
axi pulp-platform/axi: Synthesizable SystemVerilog AXI IP modules with full verification infrastructure, enabling high-performance, reusable on-chip communication for SoC designs. 1.6k
librealsense realsenseai/librealsense: The open-source RealSense SDK providing cross-platform access to depth, color, and motion data from Intel RealSense cameras for robotics and vision apps. 8.9k

Noble Shader Pack Adds Pathtracing to Minecraft with Performance Focus 🔗

GLSL-based visual upgrade balances realism and frame rates for Optifine/Iris users on Windows/Linux

BelmuTM/Noble · GLSL · 168 stars Est. 2021

The Noble shader pack for Minecraft has integrated pathtracing techniques into its GLSL rendering pipeline, enhancing light simulation without requiring dedicated ray-tracing hardware. Built for Optifine or Iris on Minecraft 1.16+, it leverages custom shaders to simulate global illumination, soft shadows, and ambient occlusion using screen-space and temporal methods.

The project, last updated 24 days ago, maintains zero open issues and targets Nvidia GTX 950+ or AMD R7 370+ GPUs.

Noble avoids monetization despite offering paid variants on Patreon, remaining free and open source under GPLv3.0. Its README highlights hundreds of hours of developer effort, crediting ShaderLabs for GLSL guidance and a Discord community for testing. Showcases emphasize clean visuals over extreme realism, aiming for enjoyable, immersive gameplay rather than cinematic fidelity.

The catch: MacOS support is explicitly excluded, and performance gains depend heavily on GPU generation, potentially leaving older or integrated graphics users with uneven results despite meeting minimum specs.

Previously in The Times “covered” — Jul 17

Use Cases
  • Minecraft players seeking better lighting without RTX hardware
  • Builders testing custom shader behavior in modded environments
  • Developers studying GLSL path tracing approximations in games

Source: BelmuTM/Noble — based on the project README.

More Stories

Godot MCP Native Lets AI Edit Scenes and Scripts Directly 🔗

Plugin implements Model Context Protocol in GDScript with no external dependencies

yurineko73/Godot-MCP-Native · GDScript · 451 stars 2mo old

Godot MCP Native integrates AI assistants like Claude into the Godot Engine via the Model Context Protocol, enabling natural language control over scenes, scripts, nodes, and resources. Built entirely in GDScript, it runs natively without Node.

js or other dependencies, offering 155 tools for node manipulation, script editing, debugging, and project management. AI can read and modify projects in real time, applying suggestions directly in the editor. Installation is streamlined via the Asset Library or manual download. The project shows active maintenance with recent commits and new contributors, though 13 open issues indicate ongoing refinement.
The catch: Thirteen open questions remain about stability and edge-case handling in complex projects.

Use Cases
  • Developers refactor GDScript using AI voice commands
  • Designers adjust node properties via conversational prompts
  • Teams automate scene audits with AI-driven validation checks

Source: yurineko73/Godot-MCP-Native — based on the README and release notes.

Material Maker updates UI with panel layouts and shortcuts 🔗

Latest release adds workflow tweaks from community contributor williamchange

RodZill4/material-maker · GDScript · 5.7k stars Est. 2018

Material Maker 1.7 introduces menu options to load, save, and reset panel layouts, side panel toggling, and keyboard shortcuts for the quickbar.

It adds sorting to the add-node popup, improved graph view links for remote parameters, and file thumbnails in dialogs. The update also enables single-window mode and remembers display modes in file dialogs, all contributed by williamchange. Built on Godot Engine with GDScript, the tool remains focused on procedural texture authoring and 3D model painting via node-based graphs.
The catch: With 301 open issues and last commit a day ago, ongoing maintenance relies heavily on sporadic contributions despite the project’s ~8-year age.

Use Cases
  • Game artists create seamless textures using node graphs
  • Designers paint 3D models directly within Godot workflows
  • Developers synthesize materials via GLSL-powered procedural generation

Source: RodZill4/material-maker — based on the README and release notes.

Quick Hits

WickedEngine WickedEngine delivers high-performance 3D rendering with modern graphics features in C++ for real-time visual applications. 7.2k
learn-gdscript learn-gdscript teaches Godot’s GDScript from scratch in your browser — free, interactive, and ideal for beginners. 2.7k
shaders shaders provides a curated set of GLSL shaders optimized for Niri, enhancing desktop visuals with procedural effects. 185
Super-Mario-Bros.-Remastered-Public Super-Mario-Bros.-Remastered-Public recreates the classic with new levels, characters, modes, and a full level editor in GDScript. 2.8k
mpv-anime-build mpv-anime-build auto-detects anime content and applies AI upscaling (Anime4K, ArtCNN, etc.) for superior playback quality. 162
pyxel pyxel is a lightweight Rust-powered retro game engine designed for Python developers to create 8-bit style games. 17.6k
The Git Times AI Desk
Ask about today's stories — or hit “Ask about this” on any article to focus on one.

Unlock the Git Times AI desk to ask about today's stories and the AI model market.

Upgrade to Premium
Answers by the Git Times AI desk · verify before you ship