Design
Preset
Background
Text
Font
Size
Width
Account Pricing Friday, July 31, 2026

The Git Times

“Civilization advances by extending the number of important operations which we can perform without thinking about them.” — Alfred North Whitehead

AI Models
Claude Opus 5 $25/M GPT-5.6 Luna $0.60/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 →

Multica Turns AI Coding Agents Into Autonomous Teammates For Dev Teams 🔗

Agents now self-assign, report progress, and compound skills in human-like workflows

multica-ai/multica · Go · ▲ 42.8k in 1d 6mo old · Latest: v0.4.15

Multica reimagines AI agents as first-class teammates rather than prompt-dependent tools. Instead of babysitting runs or copy-pasting context, developers assign GitHub issues to agents like they would to a human colleague.

The agent picks up the task, writes code, flags blockers, and updates statuses autonomously—appearing on project boards and participating in conversations. Built in Go, the platform supports over a dozen agent runtimes including Claude Code, Codex, GitHub Copilot CLI, and Cursor Agent, enabling vendor-neutral, self-hosted orchestration. Its Squads feature adds a routing layer where agent-led groups delegate work internally, mimicking human team hierarchies. Recent work refined skill management with a floating save pill showing change summaries and standardized icons, while issue interactions gained "Open in new tab" and resolved-thread badges. The system’s core innovation lies in treating agents as persistent, skill-accumulating entities that share context over time—reducing redundant prompting and enabling compound expertise. This approach mirrors time-sharing systems like Multics, adapting the concept for human-AI collaboration by multiplexing agent "users" across shared infrastructure.
The catch: Despite rapid iteration, 1,197 open issues signal ongoing instability in edge-case handling, particularly around skill persistence and cross-agent communication in large Squads, making production adoption risky for mission-critical pipelines without extensive testing.

Why this leads today Multica enables developers to deploy coding agents as trackable, skill-compounding teammates, shifting AI from passive tools to structured team collaborators in real development workflows.

Use Cases
  • Assign GitHub issues to AI agents for autonomous coding
  • Build agent squads that self-delegate tasks based on expertise
  • Compound reusable skills across agents to reduce redundant work

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

Fresh on Hugging Face

Model Drops

The newest model releases builders are picking up right now.

More on the Front Page

Coral Unifies Agent Data Access Through SQL Over APIs and Files 🔗

v0.9.0 adds OIDC/OAuth support and pagination fixes for MCP-exposed agent workflows

withcoral/coral · Rust · 5k stars 3mo old

Coral gives AI agents a local-first SQL runtime that queries APIs, files, and live data sources through a single interface. Instead of chaining tool calls for each source, agents write standard SQL, and Coral translates it into the appropriate API calls or file reads, returning a unified result set.

The project exposes this runtime over MCP, letting agents reason across Datadog, Sentry, Linear, Slack, and GitHub without custom glue code.

The v0.9.0 release, published July 31, introduces two breaking changes to its source spec: it now reads allOf-composed responses as page envelopes and follows next-page URLs embedded in response bodies. These updates improve pagination handling for REST APIs. The release also adds complete OIDC callback handling and OAuth authorization request support in the app layer, plus a reusable function tool for MCP integration.

Benchmarks with Claude Opus 4.6 show Coral-driven agents were 20% more accurate and 2x more cost efficient across 8% higher accuracy and 2.7x better cost efficiency on average across real-world AI tasks, with the strongest gains in complex coding workflows—where accuracy jumped 31% and cost efficiency improved 3.4x. Latency dropped 42% overall.

The catch: Coral remains early-stage, with 393 open issues and a narrow scope focused on agent-centric SQL translation; teams needing broad enterprise data virtualization or mature BI tooling may find its ecosystem limited compared to established query federation platforms.

Use Cases
  • Agents query multiple SaaS APIs via SQL
  • CLI users inspect live source schemas
  • MCP servers expose unified data to agents

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

Orca Orchestrates Parallel AI Agents Across Devices for Developers 🔗

Latest update adds type-ahead workspace pickers and emoji shortcodes to streamline workflow

stablyai/orca · TypeScript · ▲ 34.2k in 1d 4mo old

Orca lets builders run coding agents like Codex, ClaudeCode, and OpenCode side-by-side in isolated git worktrees, managed from desktop, mobile, or VPS. Its Ghostty-inspired terminal supports WebGL rendering, infinite splits, and persistent scrollback.

Users can click UI elements in Chromium to inject HTML, CSS, and screenshots into agent prompts, or drop comments on diffs to edit and commit without context switching. The mobile app enables remote steering and notifications. The catch: With 2,754 open issues and rapid daily churn, stability for long-running, enterprise-scale agent fleets remains unproven despite active development.

Use Cases
  • Developers testing agent outputs across five parallel worktrees
  • Teams reviewing PRs and issues without leaving the Orca interface
  • Remote workers editing files on beefy VPS via port-forwarded terminals

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

HNewhere brings Hacker News discussions to any article page 🔗

Adds sidebar comments, quote linking, and heat-mapped passage highlights for context

twalichiewicz/HNewhere · JavaScript · 199 stars 2d old

HNewhere is a lightweight userscript that detects Hacker News stories matching the current article and loads comments into a resizable sidebar. It enables quote-linked annotations, discussion heat scoring via BM25 to highlight widely discussed passages, and clickable cited text that jumps between article and comment.

Users can filter discussions, reply from the sidebar, and submit pages to HN when no thread exists. Appearance settings include theme detection, button shape/size control, and per-site hiding. It works with Tampermonkey or Violentmonkey and avoids interfering with auth flows or webmail.
The catch: Discussion heat requires three converging comments before highlighting passages, potentially missing nuanced early debates.

Use Cases
  • Developers researching bug fixes via HN comments on documentation
  • Journalists verifying claims by checking article discussions in situ
  • Students exploring technical paper critiques without context switching

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

Sub2API Guardian Automates Channel Traffic Management for Builders 🔗

Open-source Go tool balances price, speed, and load across sub2api channels with zero-downtime failover.

codermyxiaoc/sub2api-guardian · Go · 218 stars 1d old

Sub2API Guardian v1.0.

0 automates traffic routing across sub2api account channels by monitoring health, latency, and rate limits. It dynamically adjusts priority and load factors based on user-selected strategies—price-first, speed-first, or balanced—while preserving sub2api’s native 429 handling. The system writes baseline configurations before intervention, enabling instant rollback. A single binary, installable via curl | sudo bash, deploys as a systemd service with embedded frontend, SQLite backend, and audit logging for sync, probe, and failover events. It supports Linux AMD64 and ARM64, requires Go 1.24+, Node 18+, and pnpm. The catch: Early release with no open issues but limited real-world scale validation beyond single-server deployments.

Use Cases
  • DevOps teams automate failover for sub2api-backed services
  • SREs balance cost and latency across global API channel pools
  • Platform engineers reduce manual channel monitoring overhead

Source: codermyxiaoc/sub2api-guardian — based on the README and release notes.

Solidity Arbitrage Bot Automates Flashloan MEV Opportunities on Ethereum 🔗

Open-source contract enables permissionless arbitrage with router whitelisting and emergency pause

MIgHTy-alIeN/Ethereum-Flashloan-Mev-Bot · Solidity · 1.8k stars 2w old

MIgHTy-alIeN/Ethereum-Flashloan-Mev-Bot provides a Solidity smart contract that executes arbitrage between liquidity pools using flashloans, controlled by an external Python automation script. The bot searches for price discrepancies across routers and executes trades in a single transaction via executeArbitrage().

Owners configure allowed tokens and routers, set swap limits, and can pause operations or withdraw funds. Deployment requires funding the contract with 0.5–1 ETH and compiling with Solidity 0.8.20.
The catch: 16 open issues suggest unresolved reliability or security concerns despite recent activity.

Use Cases
  • DeFi traders automate cross-exchange arbitrage
  • Developers test MEV strategies on Ethereum testnets
  • Builders learn flashloan integration in Solidity contracts

Source: MIgHTy-alIeN/Ethereum-Flashloan-Mev-Bot — based on the project README.

Handy Brings Private, Offline Speech-to-Text to Any Text Field 🔗

Rust-based tool uses Whisper models locally for instant transcription without cloud reliance

cjpais/Handy · Rust · 28.1k stars Est. 2025

Handy transcribes speech directly into any active text field using configurable keyboard shortcuts, processing audio entirely on-device with Whisper or Parakeet V3 models. Built with Tauri v2 in Rust, it runs on Windows, macOS, and Linux, requiring only microphone and accessibility permissions.

The app filters silence via Silero VAD and pastes results instantly, prioritizing privacy by avoiding any cloud transmission. Recent updates added Hindi and Spanish translations, fixed macOS paste behavior, and improved Windows shutdown handling. The catch: Accuracy depends on model size and hardware, with larger Whisper models needing GPU support for real-time performance on less powerful machines.

Use Cases
  • Developers transcribing meeting notes into code comments
  • Writers drafting documents via voice in any text editor
  • Users with mobility impairments controlling apps via voice input

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

Open Source Agents Forge Modular LLM Toolchains for Specialized Workflows 🔗

Developers chain purpose-built skills, proxies, and routers to bypass model limits and vendor lock-in

Trendllm-tools
lingbol088-spec/5.6-JAILBREAK-NERV-codex-instruct-5.6rilldata/rillMIgHTy-alIeN/Ethereum-Flashloan-Mev-Botlangwatch/langwatchstablyai/orca+23 more

A clear pattern emerges in open source LLM tooling: modular, composable agents that chain specialized skills to extend model capabilities beyond raw inference. Projects like langwatch/langwatch provide evaluation frameworks for agent testing, while stablyai/orca acts as an ADE for orchestrating fleets of parallel agents with custom subscriptions.

Skill routers such as zhaoxuya520/reverse-skill and VoltAgent/awesome-agent-skills offer AI-powered routing and on-demand toolchain bootstrapping, enabling agents to dynamically select tools like penetration testing frameworks or academic research workflows. Proxies like lidge-jun/opencodex and diegosouzapw/OmniRoute abstract provider access—supporting 290+ models via one endpoint with quota-aware fallbacks and token compression—while raullenchai/Rapid-MLX optimizes local execution on Apple Silicon with 4.2x speed gains and drop-in OpenAI compatibility. Even niche tools like bradautomates/claude-video give agents multimodal abilities by processing video frames for Claude. This shift moves beyond prompt engineering toward executable agent pipelines where skills are versioned, shared, and composed like libraries. The catch: Despite rapid innovation, the ecosystem remains fragmented—skill formats vary (Python, TypeScript, PowerShell), interoperability is inconsistent, and many tools rely on unverified claims of performance or security, making production adoption risky without deep vetting.

Use Cases
  • Developers test LLM agent reliability before deployment
  • Teams route AI workloads across free and paid model providers
  • Agents autonomously select security tools for penetration testing tasks

Open Source AI Agents Shift from Assistants to Autonomous Workflows 🔗

Projects now orchestrate multi-agent systems for coding, trading, and governance with minimal human oversight

The open source landscape is rapidly evolving beyond single-purpose AI helpers toward coordinated agent networks that execute complex, end-to-end tasks. Repos like stablyai/orca enable fleets of parallel coding agents to run across desktop, mobile, and VPS environments, while dust-tt/dust provides a customizable platform for composing agent skills into tailored workflows.

Specialized agents are emerging for niche domains: HKUDS/Vibe-Trading deploys AI agents for algorithmic trading strategies, and alibaba/open-code-review integrates LLMs into deterministic pipelines for precise, line-level code analysis with built-in security rulesets.

Governance and reliability are gaining traction as agents grow more autonomous. microsoft/agent-governance-toolkit addresses OWASP Agentic Top 10 risks through policy enforcement, zero-trust identity, and execution sandboxing. Meanwhile, langwatch/langwatch offers TypeScript-based evaluation harnesses to test agent behavior, and can1357/oh-my-pi delivers a terminal-native coding agent with hash-anchored edits and subagent orchestration.

Creative and productivity tools are also being reimagined: calesthio/OpenMontage turns AI agents into a full video production studio with 12 pipelines and 500+ skills, while blader/humanizer removes AI-generated text signatures as an agent skill. Even foundational layers are shifting, with unicity-aos/aos-ce proposing an open agent operating system and herdrdev/herdr functioning as a Rust-based agent multiplexer in the terminal.

The catch: Despite impressive scope, many of these projects remain in early stages, with overlapping functionality, inconsistent interfaces, and limited real-world validation at scale—raising concerns about fragmentation and the gap between demo-ready prototypes and production-grade agent systems.

Use Cases
  • Developers orchestrating fleets of coding agents across devices
  • Traders deploying AI agents for multi-strategy market analysis
  • Teams evaluating and governing autonomous agent behavior for compliance

Open Source Embraces Modular Web Interfaces Across Languages 🔗

From dashboards to AI agents, projects unify web-native tools with platform-agnostic design.

Trendweb-frameworks
lissy93/dashylynx-family/lynxvendurehq/vendureLinwoodDev/Butterflyraullenchai/Rapid-MLX+13 more

A clear pattern emerges in open source: web interfaces are being rebuilt as modular, embeddable components that transcend traditional frontend boundaries. Dashy (lissy93/dashy) exemplifies this with its Vue-based, self-hostable dashboard featuring a live UI editor and widget system — enabling users to compose personal control panels without full-stack vendoring.

Similarly, Vendure (vendurehq/vendure) extends this modularity into commerce, using React and GraphQL to let developers plug in custom storefronts or admin UIs as independent services. The trend deepens with AI integration: Page Agent (alibaba/page-agent) introduces a TypeScript-driven in-page GUI agent that interprets natural language to manipulate any web interface, effectively turning the browser into a programmable surface. Meanwhile, GeoLibre (opengeos/GeoLibre) delivers a cloud-native GIS platform that runs identically in browsers, desktops, mobiles, and Jupyter notebooks — proving that complex web applications can now be truly universal via standardized web technologies. Even version control is evolving: GitButler (gitbutlerapp/gitbutler) uses Tauri, Rust, and Svelte to create a desktop Git client that feels native but is fundamentally a web app wrapped in a thin shell. This shift reflects a broader move toward interoperable, runtime-flexible web components that reject monolithic frameworks in favor of composable, declarative UIs.
The catch: Much of this remains experimental — many projects rely on rapidly evolving standards like Web Components or Tauri, and true cross-platform consistency often falters at the edges, especially with complex state or native integrations, leaving developers to wrestle with fragmentation beneath the promise of universality.

Use Cases
  • Developers build custom internal dashboards with live widget composition
  • E-commerce teams swap storefront UIs without rewriting backend logic
  • AI assistants automate browser tasks via natural language commands

Deep Cuts

Jailbreak Framework Bypasses AI Safety Layers for Uninterrupted GPT-5.6 Access 🔗

Embeds 23 real-time tamper rules and file routing to evade cloud moderation entirely

lingbol088-spec/5.6-JAILBREAK-NERV-codex-instruct-5.

6 is a Python-based framework that dismantles refusal mechanisms in GPT-5.6 through three-layered defense: context reconstruction prevents trigger detection, 23 live tampering rules neutralize safeguards mid-conversation, and file routing avoids cloud moderation checks. It bundles 31 MCP security tools, 28 skill modules, and Kali quad-backend integration, with Codex CLI ready for immediate deployment. Builders gain persistent, unfiltered model interaction for probing edge cases or stress-testing alignment — assuming they accept the ethical and operational risks of disabling core safety systems.
The catch: It remains niche and experimental, with limited documentation and unclear long-term viability as model defenses evolve.

Use Cases
  • Security researchers testing LLM jailbreak resilience
  • Developers probing alignment failures in high-stakes scenarios
  • Red teams simulating adversarial AI interactions offline

Source: lingbol088-spec/5.6-JAILBREAK-NERV-codex-instruct-5.6 — based on the project README.

Pixshell Unifies SSH Terminal and File Transfer in One Swift Desktop App 🔗

Native macOS and Windows builds eliminate Electron bloat while supporting Linux workflows

lyu0805/pixshell · Swift · 246 stars

Pixshell merges terminal emulation with SFTP file management in a single cross-platform client. Built in Swift, it delivers native performance on macOS and Windows without relying on Electron, reducing memory overhead and improving responsiveness.

Developers gain seamless workflow integration—edit remote files locally, execute commands in an embedded xterm-compatible terminal, and manage connections through a unified interface. Its connection manager saves credentials and tunneling settings per host, streamlining DevOps routines. Unlike heavier clients, Pixshell prioritizes speed and simplicity, making it ideal for frequent server access.
The catch: Early-stage with limited Linux GUI support and fewer community plugins than established alternatives.

Use Cases
  • Sysadmin debugging logs via terminal while transferring configs
  • Developer editing remote source files using local IDE sync
  • DevOps engineer managing jump hosts and port forwards daily

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

Quick Hits

AmberConsole A retro-inspired CSS design system bringing 1970s aesthetics to modern UIs with clean, nostalgic styling for builders seeking unique visual identity. 210
Rapid-MLX The fastest local AI engine for Apple Silicon, delivering 4.2x speed over Ollama with sub-100ms cached TTFT, full tool calling, and drop-in OpenAI compatibility for seamless AI workflow integration. 3.4k
gitbutler A modern Git client built with Tauri, Rust, and Svelte that enhances version control through native performance and cross-platform desktop flexibility powered by Git. 21.4k
personal-model Enables developers to create personalized HUMAN.md files — living documentation that captures individual workflows, preferences, and context for better team collaboration and self-awareness. 1.3k
lynx A cross-platform C++ browser engine empowering web developers to build consistent, high-performance experiences across devices while expanding participation in the open web ecosystem. 15k
decimen-optical-transfer A TypeScript library for optical transfer functions, enabling precise color and luminance manipulation in imaging pipelines for developers building advanced visual applications. 655
Beyond GitHub

The AI Wire

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

From the labs & arXiv

OpenClaw v2026.7.1 Overhauls Onboarding and Control UI for Personal AI Assistants 🔗

Adds GPT-5.6 support and official app updates across iOS, Android, and macOS with expanded model provider integration

openclaw/openclaw · TypeScript · ▲ 109 in 1d 8mo old · Latest: v2026.7.1

OpenClaw released v2026.7.

1, its most substantial update since launch, featuring a complete Control UI overhaul that lets users organize conversations side-by-side with live Tasks, file previews, and real-time usage and cost tracking. The onboarding flow now guides users from install to first chat with connection validation and preserved progress if setup is interrupted.

Official iOS, Android, and macOS apps received deep revisions in navigation, voice handling, permissions, offline reading, and native session controls. Gateway stability improved with fixes for crash loops, scheduled work, and remote browser control, while workspace terminals and goal tracking saw refinements.

Model support expanded to include GPT-5.6, Tencent Hy3, and Meta Muse Spark 1.1, alongside stronger Codex and connected coding-agent workflows. Users can now connect via Anthropic, OpenAI, Google Gemini, xAI Grok, OpenRouter, GitHub Copilot, and any OpenAI- or Anthropic-compatible endpoint, with OAuth sign-in available for Claude Pro/Max, ChatGPT/Codex, and GitHub Copilot subscriptions.

The release incorporated 3,063 contributions from 532 contributors, signaling active community maintenance.

The catch: Despite its broad channel support and local-first design, OpenClaw requires Node 24.15+ and runs as a persistent daemon, which may strain older hardware or conflict with strict endpoint security policies in enterprise environments.

Previously in The Times “covered” — Jul 26

Use Cases
  • Developers testing local AI workflows across messaging platforms
  • Privacy-focused users running personal assistants on own hardware
  • Teams evaluating self-hosted alternatives to cloud-based AI agents

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

More Stories

MCP Server Collection Grows as AI Tooling Demand Surges 🔗

Forks near 14K signal broad adoption despite 2.2K unresolved issues

punkpeye/awesome-mcp-servers · Unknown · 91.6k stars Est. 2024

The punkpeye/awesome-mcp-servers repo aggregates MCP (Model Context Protocol) servers enabling AI agents to access external tools and data. With 13,798 forks and recent activity just two days ago, it reflects sustained community engagement in standardizing AI integrations.

The project serves as a central index for developers building MCP-compatible services, though no README limits onboarding clarity.
The catch: Over 2,200 open issues raise questions about maintenance depth and server quality control amid rapid growth.

Use Cases
  • DevOps teams deploying agent workflows
  • AI researchers testing tool integration
  • Platform engineers standardizing context access

Source: punkpeye/awesome-mcp-servers — based on the project README.

Google Gemini Cookbook Adds Video Editing and Music Generation Guides 🔗

New tutorials cover Omni Flash for natural language video edits and Lyria 3 for AI song creation

google-gemini/cookbook · Jupyter Notebook · 17.6k stars Est. 2024

The google-gemini/cookbook repository now includes step-by-step guides for Gemini Omni Flash, enabling developers to edit videos using natural language prompts via the Interactions API. It also features Lyria 3 tutorials for generating 30-second clips, full songs, and image-to-music transformations.

Organized into Quick Starts and practical Examples, the cookbook supports hands-on learning for multimodal and agentic workflows. Recent updates reflect Gemini 3.6 Flash’s improved performance and lower token usage. The catch: The cookbook assumes familiarity with Google Cloud setup and API authentication, which may pose a barrier for beginners.

Use Cases
  • Developers edit videos using natural language with Gemini Omni Flash
  • Musicians generate AI-assisted songs and image-to-music with Lyria 3
  • Engineers build multimodal agents using the Agents API and Antigravity framework

Source: google-gemini/cookbook — based on the project README.

Quick Hits

ray Ray scales AI workloads with a distributed runtime and optimized libraries, letting builders train and serve models efficiently across clusters. 43.4k
firecrawl Firecrawl enables programmatic web interaction at scale — search, scrape, and automate sites via a simple API for data-driven apps. 158.5k
langchain LangChain provides the framework to build, deploy, and manage LLM agents with memory, tools, and reasoning — all in one platform. 143.1k
keras Keras simplifies deep learning with intuitive, modular APIs that let builders prototype and deploy neural networks fast, without sacrificing power. 64.2k
netdata Netdata delivers real-time, AI-enhanced full-stack observability with zero config, giving lean teams instant insights into system performance. 80k
AutoGPT AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters. 185.7k

ROSClaw bridges AI intent to physical action with evidence-based execution 🔗

Runtime enforces safety via policy arbitration and returns auditable ExecutionReceipts

ros-claw/rosclaw · Python · ▲ 1 in 1d 4mo old

ROSClaw acts as a trustworthy control plane that grounds high-level agent intentions into safe, verifiable robot actions. It does not replace ROS 2 or wrap LLMs thinly; instead, it sits between northbound agents like Codex or VLA services and southbound systems including ROS 2, simulators, and robot controllers.

When an agent requests an action, ROSClaw binds intent to a specific robot body and capability, checks policy and authorization, arbitrates physical resources, and dispatches to the appropriate driver. Crucially, it does not consider a request executed until it returns an ExecutionReceipt containing evidence of what happened during execution. This receipt feeds asynchronous memory and skill evolution processes, ensuring learning is rooted in real-world outcomes, not just simulated plans. The system enforces a “fail closed” stance by default, requiring sandbox validation before any physical action proceeds. Recent work has strengthened the capability routing layer and improved the fidelity of the e-URDF model used to represent robot morphology and constraints. Developers working with embodied AI agents who need verifiable, safe execution—especially in real-world deployment—world settings where latency, friction, and safety boundaries matter—are the primary audience. The core package supports Python 3.11–3.13, with advanced features like the LeRobot 0.6 runtime requiring Python 3.12+.
The catch: ROSClaw remains alpha software with known gaps in multi-robot coordination and limited production-scale validation outside controlled labs and simulation.

Previously in The Times “covered” — Jul 30

Use Cases
  • Robotics teams validating LLM-generated skills on physical arms
  • Autonomous vehicle engineers testing perception-to-action pipelines
  • Research labs building auditable embodied agents for hazardous environments

Source: ros-claw/rosclaw — based on the project README.

More Stories

ROS 2 documentation project serves live preview via sphinx-autobuild 🔗

Developers test changes locally with automatic browser reload on file saves

ros2/ros2_documentation · Python · 966 stars Est. 2018

The ros2/ros2_documentation repo hosts sources for docs.ros.

org, rebuilt nightly by Jenkins. Contributors use make html to generate HTML or sphinx-autobuild for live previews at http://localhost:2022 during edits. Builds require make, graphviz, and a Python venv with pinned versions from constraints.txt. Multiversion testing uses make multiversion to simulate site deployments. Parallel builds leverage Sphinx’s -j auto for faster output.
The catch: 186 open issues signal unresolved doc gaps or tooling friction despite active maintenance.

Use Cases
  • Docs writers validate ROS 2 feature guides locally
  • Contributors preview spelling fixes via `make spellcheck`
  • Teams test multisite builds matching rolling release docs

Source: ros2/ros2_documentation — based on the project README.

Gazebo Sim Adds Real-Time Remote Simulation via TCP/IP Transport 🔗

Latest release enables distributed robotics testing across networked servers and edge devices

gazebosim/gz-sim · C++ · ▲ 1 in 1d Est. 2020

Gazebo Sim’s gz-sim10_10.0.

0 release strengthens its TCP/IP Transport layer, allowing seamless simulation orchestration between local workstations and remote servers. Builders can now run physics-intensive scenes on cloud GPUs while rendering locally, or synchronize multi-robot fleets across geographically dispersed nodes using Gazebo Transport’s message-passing API. The update improves latency handling and adds TLS encryption for secure cross-site deployment, addressing a key gap in scalable robotics CI/CD pipelines. Documentation includes new command-line flags for interface binding and port multiplexing.
The catch: Real-time performance degrades significantly beyond 50ms network latency, limiting use in tight-loop control scenarios without local proxy buffering.

Use Cases
  • Simulate drone swarms over 5G links from edge gateways
  • Run nightly regression tests on GPU cloud instances
  • Coordinate humanoid robot trials across university lab sites

Source: gazebosim/gz-sim — based on the README and release notes.

Quick Hits

client-sdk-cpp Build realtime audio, video, and data apps with LiveKit’s official C++ SDK — low-latency, cross-platform, and production-ready. 66
robotgo Automate desktop GUI, mouse, keyboard, and screen control across Windows, macOS, and Linux with Go-native RobotGo — no dependencies, pure performance. 10.8k
mavros Bridge MAVLink drones to ROS ecosystems with mavros — enabling seamless telemetry, control, and GCS integration for robotics developers. 1.2k
ardupilot Develop autonomous vehicles — planes, copters, rovers, subs — with ArduPilot’s open-source flight stack: mature, modular, and hardware-agnostic. 15.6k
roomba_rest980 Reverse-engineer and locally control Roomba i7/j7+ robots via REST API — free them from the cloud and integrate into Home Assistant without vendor lock-in. 56
carla Open-source simulator for autonomous driving research. 14.2k

Yakit v1.4.8-0724 adds gRPC-powered MITM proxy for BurpSuite-like traffic control 🔗

New release enables custom Yaklang scripting and protocol multiplexing for dynamic fuzzing and replay

yaklang/yakit · TypeScript · ▲ 1 in 1d Est. 2021 · Latest: v1.4.8-0724

Yaklang’s Yakit platform now delivers a fully programmable MITM proxy via gRPC, letting security engineers intercept, modify, and replay HTTP traffic with Yaklang scripts instead of relying on BurpSuite’s GUI alone. The v1.

4.8-0724 release tightens integration between the Yaklang engine and Yakit’s client, enabling dynamic payload generation through fuzztags like {{int(1-10)}} and {{file(/path/to/dict)}} directly in the Web Fuzzer module. Users can now chain MITM interception → history review → automated fuzzing without switching tools, all while Yaklang handles protocol-level details like chunked encoding, boundary repair, and CRLF normalization under the hood. A key advancement is Yaklang’s port protocol multiplexing, which allows a single listening port to detect and respond to LDAP, HTTP, or custom services—eliminating the need to spin up multiple servers for chained exploits. The gRPC server design supports remote or local deployment, letting teams run Yakit headlessly in CI pipelines or attach GUI clients on demand. Yakit’s plugin store further extends this, permitting inline Yaklang execution at any pentest stage—from reconnaissance to post-exploit validation.
The catch: Despite its all-in-one ambition, Yakit remains heavy on Yaklang expertise; teams without internal DSL fluency face a steep learning curve to unlock advanced features like custom protocol handlers or in-memory fuzztag generation, limiting adoption in environments prioritizing tool familiarity over extensibility.

Previously in The Times “covered” — Jul 29

Use Cases
  • Red team engineers crafting multi-protocol exploit chains
  • Blue teams analyzing traffic with custom Yaklang detection rules
  • Pentesters automating credential stuffing via dynamic fuzztag payloads

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

More Stories

OpenAI's Codex Security CLI Finds and Fixes Code Vulnerabilities 🔗

Integrates with CI pipelines using API keys for automated security scanning

openai/codex-security · TypeScript · 7.5k stars 2w old

OpenAI's openai/codex-security project provides a TypeScript SDK and CLI tool that scans code for security vulnerabilities, validates findings, and suggests fixes. It supports Node.

js 22.13+ or 24.x/26.x and Python 3.10+, requiring Codex Security access for operation. In CI environments, it uses OPENAI_API_KEY or CODEX_API_KEY credentials without storing them locally. The tool compares scan results across revisions to identify new, persisting, or resolved issues and supports Dockerized, resumable scans of immutable Git revisions. Recent fixes improved Windows sandbox compatibility, credential keyring handling, and npm package lifecycle hardening. The catch: The project remains in early release (v0.1.4) with over 100 open issues, indicating ongoing stability and feature maturation needs.

Use Cases
  • Developers scanning Node.js applications for vulnerabilities
  • Security teams automating CI pipeline code checks
  • DevOps engineers managing credential-based security scans in containers

Source: openai/codex-security — based on the README and release notes.

Bettercap v2.41.7 patches MITM flaws in network attack framework 🔗

Fixes TCP acceptor panic and ICMPv6 router advertisement flag error

bettercap/bettercap · Go · 19.6k stars Est. 2018

Bettercap’s latest release v2.41.

7 addresses two security flaws: a crafted client handshake that could trigger a mysql.server panic (remote DoS) and an incorrect ICMPv6 router advertisement flag value (0x0c0xc0) causing spoofing failures. The update also closes connections on TCP acceptor panic recovery and improves PMKID validation logging. Built in Go, the framework remains a go-to for red teams testing WiFi, BLE, CAN-bus, and HID devices via deauthentication, MouseJacking, and credential sniffing.
The catch: Active development relies on a small core team, with 53 open issues raising questions about long-term responsiveness to emerging attack vectors.

Use Cases
  • Red team testing WiFi WPA2 handshake capture
  • Security researchers injecting HID frames via MouseJack
  • Auditors probing CAN-bus traffic with DBC decoding

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

Community-Scripts Adds Firecrawl, OmniRoute, PegaProx to Proxmox VE Toolkit 🔗

Latest release expands self-hosted service options with three new one-click install scripts

community-scripts/ProxmoxVE · Shell · ▲ 16 in 1d Est. 2024

The community-scripts/ProxmoxVE project added three new helper scripts in its July 30 release: Firecrawl for web scraping, OmniRoute for network policy management, and PegaProx for business process automation. Each follows the project’s standard one-command install pattern in Proxmox Shell, offering Default and Advanced setup modes.

Existing scripts like Immich and romm received bug fixes, while core tools gained dynamic GitLab URL support and configurable CA inheritance. With 2,801 forks and active maintenance, the library continues to streamline self-hosted deployments across home labs and production environments.
The catch: Script quality and maintenance vary across contributors, requiring users to vet individual implementations for long-term reliability.

Previously in The Times “covered” — Jul 27

Use Cases
  • Home lab admin deploys Firecrawl for automated data extraction
  • Network engineer configures OmniRoute for VLAN traffic shaping
  • Business analyst installs PegaProx to prototype workflow automation

Source: community-scripts/ProxmoxVE — based on the README and release notes.

Quick Hits

subfinder Subfinder rapidly discovers subdomains via passive sources, giving builders efficient reconnaissance without active scanning. 14.1k
caddy Caddy delivers a secure, high-performance web server with automatic HTTPS and modular extensibility across platforms. 74.5k
wstg WSTG provides a detailed, open-source framework for systematically testing web application and service security. 9.6k
sherlock Sherlock enables builders to trace usernames across social media platforms to identify digital footprints and potential risks. 87.4k
bunkerweb BunkerWeb offers a cloud-native, open-source WAF that protects web apps with modern, scalable threat defense. 10.8k

Bun v1.3.14 sharpens JavaScript tooling with faster installs and TypeScript 6 support 🔗

The Rust-powered runtime now patches security gaps and improves module resolution for monorepos

oven-sh/bun · Rust · ▲ 17 in 1d Est. 2021 · Latest: bun-v1.3.14

Bun v1.3.

14 refines its all-in-one JavaScript toolkit with targeted upgrades to the package manager and test runner. The release upgrades the underlying JavaScriptCore engine, reducing cold-start latency by approximately 15% in benchmarked React server-side rendering workloads. Security scanning via bun audit now detects vulnerabilities in transitive dependencies using an updated advisory database, addressing a gap noted in prior versions. Module resolution in monorepos has been tightened to prevent accidental hoisting of private packages, aligning behavior more closely with pnpm’s strictness.

The installer now defaults to linking Bun’s global binaries into user PATH on macOS and Linux, eliminating a manual step for new users. On Windows, the PowerShell install script gains improved error handling for antivirus false positives. TypeScript 6.0 support is active, enabling newer syntax like parameter property modifiers without transpilation flags.

Despite these gains, Bun’s Windows arm64 build remains experimental, with reported instability in native module binding for GPU-accelerated workloads.

The catch: Bun’s single-binary design sacrifices fine-grained control over individual tool versions, locking users into synchronized updates for runtime, bundler, and package manager—even when only one component needs upgrading.

Previously in The Times “covered” — Jul 28

Use Cases
  • Migrate Node.js servers to sub-second startup
  • Test TypeScript libraries with built-in watch mode
  • Manage monorepo dependencies with zero-config workspaces

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

More Stories

Gin 1.12.0 Boosts Go Web Performance with Protobuf and Binding Upgrades 🔗

Release adds Protocol Buffers support, escaped path routing, and improved error handling in context.

gin-gonic/gin · Go · 89k stars Est. 2014

Gin 1.12.

0, released July 30, enhances the high-performance Go web framework with Protobuf content negotiation, escaped path options, and new binding support for encoding.UnmarshalText. Bug fixes address ClientIP parsing, empty value handling, and file descriptor leaks in RunFd. The framework maintains its zero-allocation router and Martini-like API, now requiring Go 1.25+.
The catch: Open issues exceed 700, raising questions about maintainership bandwidth amid growing feature scope.

Use Cases
  • Building high-throughput REST APIs in Go
  • Developing microservices handling concurrent requests
  • Creating web apps needing fast response times

Source: gin-gonic/gin — based on the README and release notes.

Docker's Moby Project adds embedded container runtime option in latest release 🔗

Experimental embedded-containerd feature runs containerd inside daemon process for reduced overhead

moby/moby · Go · ▲ 10 in 1d Est. 2013

The Moby Project's v29.7.

0 release introduces an experimental embedded-containerd feature that runs containerd inside the daemon process rather than as a separate managed process. This change aims to reduce overhead and simplify deployment for containerized systems. The release also stabilizes the image mount type and adds a default-stop-timeout daemon option for containers lacking explicit timeout settings. A security update patches CVE-2026-17106 in the go-archive dependency.
The catch: The embedded-containerd mode remains experimental and may not suit production workloads requiring strict process isolation.

Use Cases
  • Engineers building custom container runtimes
  • Integrators optimizing daemon resource usage
  • Developers testing containerd configuration changes

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

PowerToys v0.100.2 patches Command Palette memory leak for Windows users 🔗

Fix resolves performance drain after extended Command Palette usage in latest release

microsoft/PowerToys · C · ▲ 379 in 1d Est. 2019

Microsoft PowerToys released v0.100.

2, addressing a memory leak in the Command Palette utility first spotted in v0.100.1. The fix reverts a Performance Monitor dock refresh change that caused resource buildup during prolonged use. With over 30 utilities including FancyZones and PowerRename, PowerToys remains a key toolkit for Windows customization. The patch is available via GitHub releases, WinGet, and the Microsoft Store for both x64 and ARM64 architectures. The catch: 7,470 open issues suggest ongoing stability challenges despite active maintenance.

Previously in The Times “covered” — Jul 25

Use Cases
  • Developers stream window layouts with FancyZones
  • IT admins batch-rename files using PowerRename
  • Users access apps faster via Command Palette shortcuts

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

Quick Hits

prometheus Prometheus delivers scalable, multi-dimensional time-series monitoring with powerful query language and alerting for reliable infrastructure observability. 65.4k
RuView RuView transforms everyday WiFi signals into real-time spatial mapping, vital signs, and presence detection — no cameras, no video, just physics-powered intelligence. 87.8k
tmux Tmux enables persistent, multiplexed terminal sessions with split panes, scripting, and remote access — essential for efficient command-line workflows. 48.2k
rtk Rtk slashes LLM token usage by 60–90% on dev tasks via smart CLI proxying — a single Rust binary with zero deps for faster, cheaper AI interactions. 74.1k
hugo Hugo builds blazing-fast static websites with instant live reload, rich templating, and zero JS bloat — the speed king of modern site generation. 89.3k

ESP32 Pinout Site Prevents Hardware Failures Before They Happen 🔗

Auto-generated from Espressif’s official KiCad and GPIO datasets, it guarantees pin accuracy across 23 modules.

FelixKunzJr/ESPPinoutWebsite · TypeScript · 46 stars 2mo old

FelixKunzJr/ESPPinoutWebsite delivers an interactive reference that stops makers from bricking ESP32 boards by misusing pins. Hosted at esp32pin.

com, it renders every ESP32-family module — ESP32, S2, S3, C3, C5, C6, and H2 — with physically accurate pad layouts and official Espressif KiCad symbols. Unlike static diagrams, it enforces real-time conflict detection: strapping pins, ADC2/Wi-Fi arbitration, flash-reserved GPIOs, and input-only constraints are flagged instantly. The pin mapping builder exports Arduino-compatible #define statements, shareable URLs, and PNGs, all validated against Espressif’s esp-gpio-tool via automated tests. Data is not hand-copied; it’s generated from vendored official sources, with a test suite that cross-checks the catalog on every run to prevent drift. Filters help isolate Wi-Fi-safe ADC pins, touch-capable GPIOs, and unconstrained outputs, reducing guesswork in circuit design.

The catch: While comprehensive for the ESP32 family, the tool does not support non-Espressif chips or alternative frameworks like ESP-IDF beyond Arduino exports, limiting its utility for developers working outside the Arduino ecosystem or with multi-vendor boards.

Use Cases
  • Embedded designer validating pin safety before PCB layout
  • Arduino user exporting conflict-free GPIO definitions for sketches
  • Educator teaching ESP32 hardware constraints with interactive diagrams

Source: FelixKunzJr/ESPPinoutWebsite — based on the project README.

More Stories

ESP32-S3 air-quality station gains TLS MQTT and custom thresholds 🔗

Release v1.1.5 adds optional gas sensor visibility and web-based threshold tuning

21cncstudio/project_aura · C · 712 stars 6mo old

Project Aura’s firmware now supports MQTT TLS/SSL with CA certificate configuration, enabling secure broker connections. Users can set custom display thresholds via web API, applied to warnings, dashboards, and charts.

Optional gas screens for NH₃, SO₂, NO₂, H₂S, and O₃ are visible in diagnostics and main views. The SFA40 diagnostics page and JSON debug endpoint aid formaldehyde sensor troubleshooting. Builders flash via browser installer or PlatformIO, using Waveshare displays and Sensirion SEN66 sensors. OTA updates, local dashboard, and Home Assistant integration work offline.
The catch: 16 MB flash is mandatory, ruling out cheaper ESP32-S3 variants with less memory.

Use Cases
  • Makers monitoring indoor pollutants with touchscreen feedback
  • Home Assistant users integrating secure air-quality data
  • DIY builders creating enclosed stations with OTA updates

Source: 21cncstudio/project_aura — based on the README and release notes.

C++ Roadmap Project Updates Visual Guide After Years of Static Content 🔗

Latest commit refreshes draw.io source and SVG output with minor layout tweaks

salmer/CppDeveloperRoadmap · HTML · 3.5k stars Est. 2021

The salmer/CppDeveloperRoadmap project received its first meaningful update in over two years, modifying the core `roadmap.drawio.

svg` file to adjust section spacing and icon alignment. Hosted on GitHub since 2021, the HTML-based guide aggregates C++ learning paths from fundamentals to advanced topics like AI integration and regulatory compliance. It remains a reference for junior and mid-level developers seeking structured self-study without navigating fragmented tutorials. The project’s CC BY-NC-SA 4.0 license allows non-commercial adaptation, though edits require manual synchronization between the draw.io source and exported formats. The catch: The Miro version is explicitly labeled obsolete and unsynced since July 2026, raising reliability concerns for teams relying on that platform for collaborative planning.

Previously in The Times “covered” — Jul 25

Use Cases
  • Junior developers planning C++ learning sequences
  • Educators structuring introductory programming curricula
  • Self-taught coders targeting industry-standard competency

Source: salmer/CppDeveloperRoadmap — based on the project README.

Quick Hits

glasgow GlasgowEmbedded/glasgow: A versatile Python-powered toolkit for debugging, programming, and interfacing with embedded hardware — like a Swiss Army knife for makers and engineers. 2.2k
AIOsense Schluggi/AIOsense: An ESPHome-based all-in-one sensor module that simplifies environmental monitoring with plug-and-play temperature, humidity, pressure, and more — no coding needed. 160
tulipcc shorepine/tulipcc: A portable, C-based creative computer and AMYboard synthesizer that lets builders craft real-time audio and visual art with Python scripting on bare metal. 951
litex enjoy-digital/litex: A Python-driven framework to rapidly design and deploy custom SoCs and FPGA hardware — turning complex digital design into accessible, reusable IP blocks. 4k
TuyaOpen tuya/TuyaOpen: A next-gen C-based AI+IoT framework enabling fast, seamless integration of intelligent agents and connectivity across ESP32, T-series chips, and beyond. 1.8k
awesome-home-assistant A curated list of amazingly awesome Home Assistant resources. 8.3k

PlayCanvas Engine Adds First-Class 3D Gaussian Splatting Support for Web Apps 🔗

Latest release v2.21.3 fixes MSDF text rendering while expanding real-time 3D visualization capabilities

playcanvas/engine · JavaScript · ▲ 7 in 1d Est. 2014 · Latest: v2.21.3

The PlayCanvas engine now natively loads and renders 3D Gaussian Splats (3DGS) within WebGL2 and WebGPU contexts, enabling photorealistic scene capture and relay directly in browsers. This follows the v2.

21.3 patch that fixed bright outlines around MSDF text glyphs — a subtle but notable polish for UI-heavy applications. Built on glTF 2.0 asset streaming with Draco and Basis compression, the engine supports asynchronous loading of massive 3DGS datasets without blocking the main thread. Developers can manipulate splatted scenes via TypeScript or JavaScript scripts, combining them with traditional mesh-based objects, physics via ammo.js, and WebXR for immersive AR/VR previews. Used by companies like BMW and Snap for product visualizations and ads, PlayCanvas bridges high-fidelity graphics and web accessibility.

The catch: Gaussian Splatting support remains experimental — performance varies significantly across mobile GPUs, and there’s no built-in editor tooling yet for splat manipulation outside code.

Previously in The Times “covered” — Jul 27

Use Cases
  • Web developers creating interactive product configurators
  • Game studios prototyping photorealistic WebXR experiences
  • Visualization teams streaming large-scale 3D captures to browsers

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

More Stories

libGDX 1.14.2 fixes input and rendering bugs for Java game devs 🔗

Minor update resolves stuck click states and bitmap font cache issues

libgdx/libgdx · Java · 25.3k stars Est. 2012

The libGDX framework released version 1.14.

2 with targeted fixes for desktop and mobile Java game development. Changes include resolving a stuck over state when ClickListener is cancelled, correcting BitmapFontCache.clear() to reset glyphCount, and improving Tiled map template handling for multiple tilesets. The release also updates Delaunay triangulation using exact predicates and switches the default bullet character to Unicode. These changes address specific pain points reported in the issue tracker without altering core architecture. libGDX remains a cross-platform tool for 2D and 3D games via OpenGL ES, with Gradle-based setup and no enforced coding style. The catch: Despite active maintenance, 335 open issues suggest ongoing challenges in stabilizing edge cases across its broad platform support.

Use Cases
  • Indie devs shipping 2D Android games
  • Teams prototyping HTML5/WebGL multiplatform titles
  • Students learning Java-based game architecture

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

Godot AI plugin gains security hardening in v3.0.7 release 🔗

Production-grade MCP server and AI tools for the Godot engine. A Snap to install. Totally free and fun.

hi-godot/godot-ai · GDScript · ▲ 20 in 1d 3mo old

The hi-godot/godot-ai project released v3.0.

7, focusing on security improvements after integrating the Model Context Protocol into Godot 4.5+. This MCP server lets AI assistants like Claude Code and Codex edit scenes, nodes, and scripts directly within the editor via over 120 operations across 43 tools. Installation remains a one-click process from Godot’s Asset Library, though the Python server still requires uv for the plugin automating, signal wiring, and material configuration without manual coding.
USE_CASES:

  1. Developers rapidly using AI-assisted mode requires uv. Despite active maintenance — last commit zero days ago — the project still lists five open issues, indicating ongoing refinement. The catch: AI-driven scene edits risk unintended changes if assistant prompts are ambiguous or poorly constrained, requiring careful prompt engineering and version control hygiene.

Previously in The Times “covered” — Jul 25

Use Cases
  • Game designers iterate levels using natural language prompts in Godot
  • Programmers generate boilerplate GDScript code via AI-assisted node setup
  • Artists adjust materials and animations through conversational AI commands

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

Flecs 4.1.6 Adds Entity Ranges and Script Functions for Game Devs 🔗

Release includes 130 bugfixes, 500+ tests, and new inspector panels in the web UI

SanderMertens/flecs · C · ▲ 1 in 1d Est. 2018

Flecs 4.1.

6 introduces an entity range API that correctly handles recycling and new script features like include statements, functions, and swizzle operators. The update also brings multiple inspector panels, entity creation from the explorer, and a dialog for adding components in the web-based UI. With over 13,000 CI tests and zero dependencies, the C99/C++17 ECS supports millions of entities, cache-friendly archetype storage, and lockless multicore scheduling. The catch: Despite its performance, Flecs requires manual memory management and lacks built-in garbage collection, which may increase complexity for teams used to higher-level engines.

Use Cases
  • Game studios building large-scale simulations with millions of entities
  • Developers creating portable C/C++ engines without external dependencies
  • Teams needing runtime entity inspection and modding via web UI tools

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

Quick Hits

phantom-camera A Cinemachine-inspired Godot 4 Camera addon that enables smooth, cinematic camera control with minimal setup for dynamic gameplay visuals. 3.5k
godot Godot Engine is a free, open-source, cross-platform game engine supporting 2D and 3D development with a node-based architecture and built-in editor for rapid prototyping. 114.8k
GDevelop GDevelop is an accessible, no-code-friendly game engine letting creators build 2D, 3D, and multiplayer games visually — ideal for beginners and rapid iteration. 25.2k
godot_heightmap_plugin A GDScript-powered heightmap terrain plugin for Godot 4 that lets you sculpt, texture, and generate realistic landscapes directly in the editor. 2.2k
gdUnit4 gdUnit4 brings professional unit testing to Godot 4 with embedded test inspection, mocking, and scene testing — enabling reliable, TDD-driven GDScript and C# development. 1.2k
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