Design
Preset
Background
Text
Font
Size
Width
Account Pricing Thursday, July 23, 2026

The Git Times

“The purpose of a system is what it does.” — Stafford Beer

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.

Syncthing Adds Proxy Support and Session Controls in v2.1 🔗

New grouping and block indexing options improve usability and efficiency for peer-to-peer file sync

syncthing/syncthing · Go · 86.8k stars Est. 2013 · Latest: v2.1.2

Syncthing, the open-source peer-to-peer file synchronization tool written in Go, released version 2.1 with meaningful usability and configuration enhancements. The update introduces device and folder grouping in the GUI via a new group attribute, helping users manage complex sync topologies.

It now supports HTTP and HTTPS proxies with CONNECT tunneling, expanding network flexibility beyond existing SOCKS proxy support. For environments where storage efficiency trumps minimal transfer size, block indexing can be disabled per folder using the blockIndexing attribute, reducing database overhead. GUI session duration is now fully configurable—admins can extend, shorten, or set login cookies to infinite via sessionCookieDurationS and adjust the cookie path with sessionCookiePath. These changes reflect Syncthing’s core goals of security, automation, and individual control, reinforcing its role as a self-hosted alternative to centralized cloud sync. The project remains actively maintained, with the last commit just two days ago and a steady stream of fixes addressing platform-specific quirks, connection reporting, and filesystem resilience. Syncthing continues to prioritize data integrity and end-to-end encryption without relying on third-party servers.
The catch: While Syncthing excels at direct device-to-device sync, it lacks native conflict resolution for concurrent edits, requiring users to manage versioning externally when files are modified simultaneously across nodes.

Why this leads today Syncthing's rising adoption reflects a shift toward decentralized, self-hosted file synchronization as users seek privacy-resilient alternatives to cloud-dependent services.

Use Cases
  • Developers syncing code and configs across personal laptops and workstations
  • Teams sharing large media files over local or encrypted peer connections
  • Individuals backing up sensitive documents to home servers without cloud reliance

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

More on the Front Page

Single HTML File Replaces PowerPoint With Built-In Editor and Collaboration 🔗

Bento Slides v1.0.7 delivers a self-contained, offline-first presentation tool that saves and morphs within one portable file.

nyblnet/bento · TypeScript · 744 stars 5d old

Bento Slides turns a single HTML file into a full-featured presentation suite — viewer, editor, and presenter — that runs in any browser without installation. The file contains its own data, fonts, and editor logic, rewriting itself on save via the File System Access API.

Users can edit, present, and share decks that carry their software with them, eliminating dependency on cloud platforms or logins. Morph animations now support editable morph IDs, letting users pair elements across slides without altering their base IDs, while stable identities for comments and collaboration remain intact. Version 1.0.7 lifts the 2× scale cap, enabling crisp 8K fullscreen rendering of vector and text content on high-resolution displays. In-place updates now retain the file handle, allowing silent overwrites after initial approval. Live collaboration uses end-to-end encryption with keys stored in the file, enabling peer-to-peer editing where anyone who opens a copy joins the session, with offline edits merging precisely. The project avoids binary lock-in by storing slide data as a readable JSON block at the file’s start.

The catch: Despite its self-contained design, Bento relies on modern browser APIs like File System Access, which may not work uniformly across all browsers or environments, particularly in locked-down corporate settings.

Use Cases
  • Presenters share decks via email that recipients can view and edit immediately
  • Teams collaborate on slides offline with end-to-end encryption and merge-on-reconnect
  • Developers embed or modify presentations as code using the visible JSON data block

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

MCP Server Collection Grows Despite Rising Maintenance Burden 🔗

Forks surge past 13k as open issues climb to 2,713 amid steady commits

punkpeye/awesome-mcp-servers · Unknown · 91.2k stars Est. 2024

The punkpeye/awesome-mcp-servers repository aggregates Model Context Protocol implementations for AI agent tooling. Recent activity shows consistent updates with the last commit just zero days ago, indicating ongoing curation.

Forks have surpassed 13,444, reflecting broad community duplication and adaptation. However, open issues now total 2,713, suggesting growing maintenance strain as contributors report bugs and feature requests faster than they are resolved. The catch: sustaining quality control becomes challenging when issue volume outpaces maintainer capacity in a rapidly evolving MCP ecosystem.

Use Cases
  • Developers discover MCP servers for custom AI agent integrations
  • Teams fork the list to build internal approved tooling inventories
  • Researchers compare MCP implementations across different language runtimes

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

Go CLI automates Grok account registration with OAuth and CPA JSON output 🔗

One-command tool handles registration, proxy clearance, and token export for builders

Charles-0509/Grok-Register · Go · 238 stars 2d old

Charles-0509/Grok-Register is a Go-based CLI that streamlines Grok account creation by combining registration, OAuth flow, and CPA-ready JSON generation in a single command. It uses tls-client with Chrome fingerprinting to bypass Cloudflare, supports Turnstile in offscreen mode, and includes WARP/Docker clearance automation.

The tool exports SSO tokens to tokens.txt and waits for Management API upload before exiting. Builders can install via scripts/install.sh on Linux/macOS with interactive setup for proxy, threads, and data paths.
The catch: The project shows recent activity but has four open issues and relies on Docker and Playwright, which may complicate minimal or air-gapped deployments.

Use Cases
  • Developers automate Grok account creation for testing
  • Teams batch-register accounts with custom thread counts
  • Engineers export SSO tokens for CPA gateway integration

Source: Charles-0509/Grok-Register — based on the project README.

Sticker Forge lets builders craft tactile WebGL stickers with real-time peel physics 🔗

Combines rich text, image uploads, and depth-aware shadowing for interactive die-cut designs

CatsJuice/sticker-forge · JavaScript · 251 stars 2d old

Sticker Forge turns text or uploaded artwork into interactive WebGL stickers with realistic peel physics. Users drag die-cut edges inward to curl the sticker and reveal a satin back surface with dynamic shadowing.

The project delivers an interactive builder in app/ and a framework-independent Web Component in lib/ via ES module or IIFE bundles. Builds output to public/embed/ with inlined peel sound and TypeScript declarations. GitHub Pages deploys to https://sticker.oooo.so on pushes to main.
The catch: Only two open issues and recent commits suggest early maturity; real-world performance at scale remains untested.

Use Cases
  • Designers create custom die-cut stickers for web apps
  • Developers embed interactive sticker makers in SaaS platforms
  • Artists prototype tactile UI elements with real-time physics feedback

Source: CatsJuice/sticker-forge — based on the project README.

Design Judge Skills Streamlines Award Submissions with Evidence-Based Agent Modules 🔗

Python toolkit verifies entries against official rules using 22k+ award records for transparent scoring.

SeanJ1ang/design-judge-skills · Python · 267 stars 5d old

SeanJ1ang/design-judge-skills breaks design award preparation into five independent Python skills: searching verified winners, evaluating designs with evidence scales, matching awards, drafting submissions, and checking readiness. It pulls from 22,125 aggregated observations of iF, Red Dot, IDEA, and other award entries to ground assessments in official sources.

Each module—like design-award-search or design-submission-check—can run solo or chain via $design-award-pipeline, requiring real-time re-verification of deadlines, fees, and rules from award websites. Built for Claude Code, Codex, and Hermes agents, it enforces fact/inference separation and avoids simulating private judging.
The catch: Reliance on live official pages means submissions could fail if award sites change structure or block automated checks during peak seasons.

Use Cases
  • Designers verify award eligibility using official criteria
  • Teams draft submission text with source-attributed claims
  • Agents auto-check entry packages against current rules

Source: SeanJ1ang/design-judge-skills — based on the project README.

Python Bot Automates Polymarket Endcycle Sniper Strategy for Small Gains 🔗

Scans, filters, and executes trades on near-resolution markets with risk controls

AdelanSupport/adelan-support · Python · 309 stars 2d old

AdelanSupport/adelan-support is a Python bot that automates the endcycle sniper strategy on Polymarket, targeting markets resolving within 0–7 days where favored outcomes trade at $0.88–$0.

  1. It scans for opportunities, scores them by conviction, and places limit orders via CLOB V2 with built-in risk management including position sizing and daily loss caps. All activity logs to SQLite, and dry-run mode is enabled by default to prevent accidental real trades. The project is freshly created with no forks or open issues, indicating early-stage development. The catch: As a brand-new tool with no community feedback or audit history, its reliability and edge in live markets remain unproven.
Use Cases
  • Traders automate Polymarket endcycle sniper strategy
  • Developers study prediction market arbitrage bots
  • Researchers test low-edge, high-conviction trading models

Source: AdelanSupport/adelan-support — based on the project README.

AI Agents Evolve Beyond Chat into Specialized Workflow Orchestrators 🔗

Open source projects now build agent systems that automate complex tasks across video, code, design, and enterprise operations

Trendai-agents
powerycy/goutoujunshipireel/pireelSeanJ1ang/design-judge-skillsKlaatAI/klaatcodejoeseesun/qiaomu-cut-skill+33 more

The open source AI agent landscape is shifting from conversational interfaces to purpose-built workflow automation systems. Projects like pireel/pireel and calesthio/OpenMontage demonstrate agent-driven video production pipelines that handle storyboarding, captioning, and rendering through natural language commands.

In software development, KlaatAI/klaatcode and can1357/oh-my-pi deploy smart model routing to optimize coding tasks by selecting the most efficient AI model per sub-task, reducing costs and latency. Design and evaluation workflows are being automated via SeanJ1ang/design-judge-skills, which uses evidence-based agent skills for award submissions and critique generation. Enterprise readiness is advancing through iflytek/skillhub, a self-hosted skill registry with RBAC and audit logs for governing agent capabilities at scale. Orchestration layers like lobehub/lobehub and stablyai/orca manage fleets of parallel agents, enabling 24/7 operation through scheduling and reporting. Even niche domains are seeing agentification: earthtojake/text-to-cad provides agent skills for hardware design, while kangarooking/cangjie-skill distills books and podcasts into executable agent skills. This pattern reveals a move toward modular, interoperable agent systems where discrete skills are composed into end-to-end workflows — turning AI from a chatbot into a programmable workforce.

The catch: Despite rapid innovation, the ecosystem remains fragmented across incompatible agent protocols, skill formats, and orchestration runtimes, with many systems still requiring tight coupling to specific LLMs or lacking standardized evaluation, making true interoperability and enterprise-grade reliability elusive for now.

Use Cases
  • Developers automate code reviews using terminal agents with smart model routing
  • Marketing teams generate multilingual video campaigns via browser-based AI editors
  • Enterprises deploy governed agent skills for internal tool automation and compliance

Open Source Shifts to Modular LLM Skills and Agent Orchestration 🔗

Projects decompose AI capabilities into reusable, interoperable skills for flexible agent composition

Trendllm-tools
SeanJ1ang/design-judge-skillsKlaatAI/klaatcodelightly-ai/lightly-trainelizaOS/elizaOtterMind/Chat2DB+20 more

A clear pattern is emerging in open source: the decomposition of LLM capabilities into discrete, reusable skills and agent orchestration layers. Rather than monolithic AI apps, projects now focus on modular components that can be mixed, matched, and routed intelligently.

KlaatAI/klaatcode exemplifies this with smart model routing — selecting Claude, GPT, Gemini, or DeepSeek per task to cut costs 10x. Similarly, Microsoft/SkillOpt trains and exports reusable natural-language skills as best_skill.md artifacts for frozen LLMs. Agent frameworks like lobehub/lobehub and elizaOS/eliza provide operating systems for hiring, scheduling, and managing AI agent teams, while deer-flow/llm-space offers a desktop app to prototype and inspect agent harnesses step-by-step. Domain-specific skills are proliferating: AgriciDaniel/claude-seo delivers 25 SEO sub-skills with 18 sub-agents, Mukul975/Anthropic-Cybersecurity-Skills maps 754 skills to frameworks like MITRE ATT&CK, and Wei-Shaw/sub2api unifies LLM API access via a shared proxy. Even niche tools like bradautomates/claude-video extend Claude with video understanding via frame extraction and transcription. This shift enables composable AI — where agents are assembled from certified skills rather than built from scratch.
The catch: Despite rapid innovation, the ecosystem remains fragmented — skill formats, agent harnesses, and routing logic lack standardization, making cross-project reuse uncertain and integration brittle outside tightly coupled setups.

Use Cases
  • Developers compose custom AI agents from reusable skills
  • Teams route LLM tasks to optimal models for cost efficiency
  • Organizations deploy validatedruncybersecurity agents using standardized skill frameworks

Open Source Data Infrastructure Converges Around AI, Observability, and Semantic Standards 🔗

Projects unify database tooling, analytics, and metadata exchange to enable smarter, composable data workflows

Trenddata-infra
ClickHouse/ClickBenchOtterMind/Chat2DBbaserow/baserowgrafana/grafanaRobbyant/lingbot-map+15 more

Open source data infrastructure is rapidly evolving beyond isolated tools toward integrated, AI-augmented stacks. ClickHouse/ClickBench provides standardized benchmarks pushing analytical database performance, while OtterMind/Chat2DB delivers an AI-powered SQL client that interprets natural language into queries across MySQL, PostgreSQL, ClickHouse, and more.

Baserow extends this with a no-code, AI-driven platform for building databases and automations, complete with GDPR and SOC 2 compliance, positioning itself as a self-hostable Airtable alternative. Observability remains central, with Grafana enabling unified visualization of metrics, logs, and traces from Prometheus, Loki, and InfluxDB. Meanwhile, Apache Ossie emerges as a critical effort to standardize semantic metadata exchange across AI, BI, and analytics platforms—offering a vendor-neutral “single source of truth” for data meaning. On the AI front, Microsoft’s Qlib demonstrates how quant research platforms now integrate RL and supervised learning with automated R&D via RD-Agent, while Ray scales AI workloads through its distributed compute engine. Together, these projects signal a shift: data infrastructure is no longer just about storage or querying—it’s about intelligent, interoperable layers where AI enhances usability, observability ensures insight, and semantic standards enable trust enables real-time analytics, AI, and BI tools can truly interoperate.
The catch: Much of this integration remains aspirational—semantic standards like Ossie lack broad adoption, AI-driven tools often rely on brittle LLM prompts rather than deep data understanding, and composability is hindered by inconsistent APIs and conflicting licensing models, raising doubts about real-world interoperability beyond demo environments.

Use Cases
  • Data teams benchmark analytical DB performance
  • Developers query databases using natural language
  • Enterprises build compliant, AI-enhanced internal tools

Deep Cuts

OpenBrowser: Local Fingerprint Browser for Secure Multi-Env Testing 🔗

Isolate environments with custom fingerprints, proxies, and RPA automation in Electron

lyu0805/OpenBrowser · JavaScript · 252 stars

lyu0805/OpenBrowser is a Chromium-based Electron app that lets developers create isolated browser profiles with unique fingerprints, proxy routing, and session persistence — all locally. It’s designed for testing anti-fingerprinting defenses, managing multiple accounts securely, or automating workflows without leaving traces.

With built-in RPA support, you can script interactions across environments while maintaining distinct browser identities. Ideal for privacy-conscious devs, security researchers, or QA teams needing reproducible, spoofable browsing contexts. The catch: It’s early-stage, primarily documented in Chinese, and lacks broad community validation despite its promising feature set.

Use Cases
  • QA engineers testing site behavior across varied browser fingerprints
  • Developers isolating dev/staging/prod environments with unique proxies
  • Researchers analyzing anti-fingerprinting techniques in controlled settings

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

Quick Hits

lobehub LobeHub orchestrates your AI agents like a Chief Operator — hiring, scheduling, and reporting 24/7 to automate complex workflows without manual oversight. 80.7k
pireel Pireel lets AI agents edit talking-head videos in-browser with storyboarding, graphics, captions, and WebCodecs export — no backend needed. 579
hyperliquid-tracker Hyperliquid Tracker alerts you instantly to large trades on Hyperliquid, giving builders real-time market signal detection for DeFi strategies. 504
eliza ElizaOS provides an open-source agentic operating system to build, deploy, and manage autonomous AI agents with modular, scalable infrastructure. 18.8k
PM-Sniper PM-Sniper is a Python tool for precision project management — predicting delays and optimizing resource allocation using lightweight ML models. 313
klaatcode KlaatCode is a terminal-based AI coding agent that intelligently routes tasks to the best model (Claude, GPT, etc.) for 10x cost efficiency and Claude Code-grade accuracy. 251
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 Rewrites Onboarding and Control UI for Local AI Assistants 🔗

Major release adds GPT-5.6 support, native app upgrades, and daemon-first setup across macOS, Linux, and Windows

openclaw/openclaw · TypeScript · 383.9k stars 8mo old · Latest: v2026.7.1

OpenClaw’s latest release overhauls its onboarding flow and control interface, making the personal AI assistant faster to set up and easier to use daily. The openclaw onboard CLI now guides users through gateway, workspace, and channel configuration with connection validation and session recovery, preserving progress if interrupted. Once running, the redesigned Control UI displays live tasks, file transfers, usage costs, and Gateway health alongside conversations in WhatsApp, Telegram, Slack, Discord, iMessage, and 20+ other platforms.

Under the hood, v2026.7.1 expands model support to include GPT-5.6, Tencent Hy3, and Meta Muse Spark 1.1, while improving Codex integration for connected coding workflows. Official iOS, Android, and macOS apps received updates to voice input, file handling, offline queuing, and native session controls. The Gateway daemon—installed via launchd, systemd, or Windows Service—now runs more stably, with fixes to crash loops and remote browser control.

Built in TypeScript and designed for single-user, self-hosted operation, OpenClaw emphasizes data ownership by running entirely on user devices. It requires Node 24.15+ and works with npm, pnpm, or bun.

The catch: Despite its broad channel support, OpenClaw remains a single-user tool with no multi-user collaboration features, limiting its usefulness in team or shared environments where assistants need to coordinate across people.

Previously in The Times “covered” — Jul 22

Use Cases
  • Developers running local AI assistants on personal devices
  • Power users syncing AI across WhatsApp, Slack, and iMessage
  • Privacy-focused builders avoiding cloud-based agent services

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

More Stories

Anthropic's Claude Cookbooks Add SQL Query Integration for Developers 🔗

Latest update expands tool use recipes with database connectivity examples

anthropics/claude-cookbooks · Jupyter Notebook · 49.4k stars Est. 2023

The anthropics/claude-cookbooks repository recently added a SQL query integration guide under Tool Use and Integration, showing how to connect Claude to databases for natural language querying. Builders can now copy-paste Python snippets that orchestrate Claude with SQL execution, extending its reasoning with structured data.

The cookbook remains a go-to resource for practical Claude API patterns, updated as recently as yesterday with 5,857 forks indicating active community use. The catch: Most examples assume Python proficiency and a Claude API key, limiting immediate adoption for teams using other languages or restricted API access.

Use Cases
  • Developers adding natural language database queries to applications
  • Teams prototyping retrieval-augmented generation with external knowledge
  • Engineers learning tool use patterns for AI agent development

Source: anthropics/claude-cookbooks — based on the project README.

GitHub Spec Kit v0.13.4 adds Factory Droid CLI integration for AI workflows 🔗

Release improves spec-of-specs documentation and fixes catalog validation bugs

github/spec-kit · Python · 123.4k stars 11mo old

GitHub’s spec-kit project released v0.13.

4, adding Factory Droid CLI integration to streamline AI-driven spec-driven development.
documentsspecscript type in installation guides

  • validates cached catalog shapes before return to prevent integration errors
  • rejects non-list catalogs in bundle manifests with clearer error messages

The catch: Open issues remain high at 299, suggesting stability challenges despite frequent commits.

Previously in The Times “covered” — Jul 17

Use Cases
  • Developers define AI-agent-ready software specs before coding
  • Teams standardize PRDs using executable specifications across projects
  • Organizations extend spec workflows with custom presets and bundles

Source: github/spec-kit — based on the README and release notes.

Microsoft Qlib Integrates RD-Agent for Automated Quant Research 🔗

LLM-driven agents now mine factors and optimize models within the AI investment platform

microsoft/qlib · Python · 46.5k stars Est. 2020

Qlib, Microsoft’s Python-based quant investment framework, now incorporates RD-Agent to automate factor mining and model optimization using LLM-powered autonomous agents. The integration supports end-to-end R&D workflows, from data processing to strategy deployment, with recent updates adding Parquet support and improved MLflow configuration.

Developers can leverage pre-built demos for quant factor mining and model optimization via linked YouTube tutorials.
The catch: Despite active development, 465 open issues suggest ongoing stability and usability challenges for production adoption.

Use Cases
  • Quant researchers automate factor discovery
  • Financial engineers optimize ML trading models
  • Data scientists backtest strategies with curated datasets

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

Quick Hits

ray Ray scales AI workloads with a distributed runtime and integrated ML libraries, letting builders train and deploy models efficiently across clusters. 43.3k
AutoGPT AutoGPT provides modular, open-source tools to create autonomous AI agents, empowering builders to focus on logic, not infrastructure. 185.6k
hermes-agent Hermes-Agent adapts and learns alongside users, evolving its behavior over time to become a personalized, long-term AI companion. 219.2k
prompts.chat Prompts.chat lets builders discover, share, and self-host curated AI prompts privately — no vendor lock-in, full control over your prompt library. 166.2k
ai-agents-for-beginners AI Agents for Beginners offers 18 hands-on Jupyter lessons to teach core concepts of building, testing, and deploying AI agents from scratch. 70.2k

Gazebo Sim Jetty Release Boosts ROS2 Integration for Real-World Robot Testing 🔗

Latest version adds native ROS2 Humble support and improved sensor fidelity for autonomous systems

gazebosim/gz-sim · C++ · 1.4k stars Est. 2020 · Latest: gz-sim10_10.0.0

Gazebo Sim’s Jetty release (gz-sim10_10.0.0) delivers native ROS2 Humble integration, eliminating the need for middleware bridges when testing autonomous robots.

Developers can now directly publish sensor data and subscribe to control topics via Gazebo Transport’s ROS2 interface, streamlining workflows for navigation and manipulation stacks. The update enhances Gazebo Sensors with higher-fidelity noise models for IMU, GPS, and LiDAR, critical for validating perception algorithms in noisy real-world conditions. Rendering upgrades include OGRE v2 support for dynamic shadows and physically based rendering, improving visual fidelity for camera-based tasks. Users gain access to updated simulation models like TurtleBot4 and enhanced SDF tooling for custom robot and environment design. The project maintains cross-platform support via Homebrew, Windows, and Ubuntu Noble, with command-line tools enabling headless CI/CD pipelines for regression testing.
The catch: Despite active development, 607 open issues signal ongoing stability challenges, particularly in complex multi-robot simulations where physics engine synchronization and transport latency remain pain points for large-scale deployments.

Previously in The Times “covered” — Jul 16

Use Cases
  • Test autonomous navigation algorithms with realistic sensor noise
  • Validate ROS2-based manipulation pipelines in dynamic environments
  • Run headless regression tests for robot firmware in CI/CD workflows

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

More Stories

Rerun 0.34.1 fixes live-stream crashes and codec errors in Rust SDK 🔗

Patch resolves stack overflow and improves MP4 handling for robotics data streams

rerun-io/rerun · Rust · 11.2k stars Est. 2022

The latest Rerun release patches a stack overflow in live-streamed data processing and improves error reporting for unsupported MP4 codecs. Built in Rust with Python, C++, and Rust SDKs, it ingests multimodal robotics data—images, point clouds, and video—then visualizes, queries, and streams it directly to training pipelines.

Teams use it to sync sensor logs, run CV pipelines live, and avoid export delays by streaming data straight into ML workflows.
The catch: Open issues remain high at 1,292, suggesting ongoing stability challenges despite recent fixes.

Previously in The Times “covered” — Jul 17

Use Cases
  • Robotics teams ingest multimodal sensor logs for synchronized analysis
  • Developers run SLAM and hand-tracking pipelines as live table edits
  • ML engineers stream training data directly from raw robot logs without export jobs

Source: rerun-io/rerun — based on the README and release notes.

MuJoCo 3.10.0 adds Unity plug-in for real-time robotics simulation 🔗

Latest release expands integration with game engines for faster prototyping workflows

google-deepmind/mujoco · C++ · 14.3k stars Est. 2021

Google DeepMind released MuJoCo 3.10.

0, introducing an official Unity plug-in that enables real-time physics simulation within the game engine. The update streamlines workflows for roboticists and animators by letting users test articulated models directly in Unity scenes without exporting data. MuJoCo’s core remains a high-performance C++ physics engine with multithreaded rollout capabilities and OpenGL visualization, now bridged to Unity via the new plug-in. Documentation includes setup guides for Linux, Windows, and macOS, with Python bindings retained for research scripting. The catch: The Unity plug-in requires a paid Unity Pro license for commercial use, limiting accessibility for indie developers and open-source robotics teams.

Previously in The Times “covered” — Jul 17

Use Cases
  • Roboticists testing control algorithms in simulated environments
  • Animators creating physics-based character interactions
  • Researchers prototyping RL agents with real-time rendering

Source: google-deepmind/mujoco — based on the README and release notes.

Spatial Math Python Fixes Critical Bug in Angle-Vector Conversion 🔗

Release 1.1.16 patches `angvec2r` function after migration to RAI Institute stewardship

rai-opensource/spatialmath-python · Python · 638 stars Est. 2020

The spatialmath-python library released version 1.1.

16, fixing a breaking bug in the angvec2r function that converts angle-axis representations to rotation matrices. This patch follows the project’s migration to the RAI Institute copyright and updates to organizational links. Used in robotics and vision for pose representation via SE3, SO3, and quaternion classes, the tool remains actively maintained with a commit just one day ago. Despite its utility in 2D/3D transformations, the library’s narrow focus on spatial math may limit appeal for developers needing broader geometric computation or physics integration.
The catch: Its specialized scope means it doesn’t include general-purpose math or graphics rendering, requiring pairing with other tools for full robotics pipelines.

Use Cases
  • Robotics engineers converting sensor data to robot poses
  • Graphics developers animating 3D object orientations
  • Vision researchers estimating camera extraneous parameters

Source: rai-opensource/spatialmath-python — based on the README and release notes.

Quick Hits

glim GLIM provides a flexible, extensible C++ framework for high-precision 3D point cloud localization and mapping, ideal for robotics and autonomous systems needing robust spatial awareness. 1.7k
BehaviorTree.CPP BehaviorTree.CPP offers a battle-tested, batteries-included C++ library for building modular, reactive AI behaviors in robotics and games with minimal boilerplate. 4.1k
client-sdk-cpp LiveKit’s official C++ client SDK enables developers to build low-latency, real-time audio, video, and data applications using WebRTC-native protocols with full native performance. 66
spatio_temporal_voxel_layer This voxel layer modernizes navigation maps by leveraging modern 3D graphics techniques to deliver dynamic, high-fidelity environmental representations for robotic path planning. 846
ardupilot ArduPilot delivers a mature, open-source C++ flight stack supporting planes, copters, rovers, and subs — trusted globally for reliable, customizable autonomous vehicle control. 15.5k

OpenCTI revamps connector catalog into marketplace for threat intel workflows 🔗

Latest release removes legacy editor, adds workflow reset and webhook deny lists

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

OpenCTI’s 7.260722.0 release refactors its connector catalog into a faceted marketplace experience, streamlining how security teams discover and configure integrations with tools like MISP and MITRE ATT&CK.

The update removes the legacy frontend editor and its feature flags, replacing it with a cleaner interface while introducing workflow enhancements: users can now fully reset workflows, prevent publishing those that delete active statuses, and restore prior versions. Webhook security improves with configurable URI deny lists, and GraphQL Playground gains light mode support. CSV exports are refined, and admins can enforce password changes. Built on TypeScript with a STIX2-based schema, OpenCTI structures technical and non-technical threat data—TTPs, observables, attribution—linking each to sources like reports or MISP events, then infers new relations to surface actionable knowledge.

The catch: Despite active development, over 2,000 open issues suggest ongoing stability and usability challenges that may impact enterprise adoption at scale.

Previously in The Times “covered” — Jul 20

Use Cases
  • Security analysts correlating TTPs with observables using MITRE ATT&CK
  • Incident responders exporting STIX2 bundles for cross-tool sharing
  • Teams automating threat data ingestion via MISP and TheHive connectors

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

More Stories

Community scripts add OS guard to prevent Proxmox container update failures 🔗

New core feature blocks mismatched OS updates that could break containers post-install

community-scripts/ProxmoxVE · Shell · 29k stars Est. 2024

The community-scripts/ProxmoxVE project released a core update on July 22, 2026, adding an OS mismatch guard to prevent container updates from failing due to version conflicts. This feature, implemented in pull request #15948, checks host and container OS compatibility before applying updates, reducing post-install troubleshooting.

The release also updated OPNSense to FreeBSD 15 and 26.7, fixed bugs in InvoiceNinja, Anytype, and Nametag scripts, and added a storage share helper tool. Scripts still require root access and internet during install, with default setups completing in under five minutes.
The catch: Advanced mode demands manual tuning of storage backends and networking, which may overwhelm users seeking zero-configuration deployments.

Previously in The Times “covered” — Jul 19

Use Cases
  • Homelab admins deploy Home Assistant containers via single command
  • Self-hosters update OPNSense firewalls with automated FreeBSD base bump
  • Developers test Jellyfin media servers with pre-allocated VM resources

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

ProjectDiscovery Subfinder Drops Dead Facebook Source in v2.14.0 🔗

Release removes non-functional Meta CT integration after API discontinuation, streamlining passive subdomain enumeration.

projectdiscovery/subfinder · Go · 14.1k stars Est. 2018

The Go-based subdomain enumeration tool subfinder v2.14.

0 removes the defunct Facebook (Meta CT) source following its API shutdown, eliminating related errors in automation and CI pipelines. This passive-focused utility pulls subdomains from curated public sources like crtsh and GitHub, offering JSON/file/stdout output and stdin/stdout workflow integration. Recent fixes also resolve Netlas download caps, transport bugs across Shodan and IntelX, and rate-limit nil-map panics. The catch: Its strictly passive model limits discovery of newly created or hidden subdomains requiring active probing.

Use Cases
  • Bug bounty hunters enumerating target domains passively
  • Security teams automating subdomain discovery in CI/CD
  • OSINT researchers gathering infrastructure data from public sources

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

Quick Hits

dirsearch Dirsearch efficiently brute-forces web paths and directories to uncover hidden endpoints and potential attack surfaces. 14.5k
sherlock Sherlock rapidly identifies social media profiles linked to a username across hundreds of platforms for OSINT investigations. 87k
trivy Trivy comprehensively scans containers, Kubernetes, code, and cloud environments for vulnerabilities, misconfigurations, secrets, and SBOMs. 37k
nginx NGINX delivers high-performance, scalable web serving, reverse proxying, and load balancing for modern web infrastructure. 31.2k
bbot bbot recursively scans the internet to discover assets, services, and vulnerabilities for proactive security reconnaissance. 10.2k
yakit Yakit unifies reconnaissance, scanning, exploitation, and reporting in an all-in-one TypeScript-powered cybersecurity platform. 7.6k

UV 0.11.31 Adds Workspace Cross-Referencing and Malware Checks for Python Devs 🔗

Latest release enables shared dependency sources across workspaces and introduces optional security auditing for installed packages.

astral-sh/uv · Rust · 87.8k stars Est. 2023 · Latest: 0.11.31

Astral’s uv 0.11.31 enhances project scalability by allowing workspace members to reference packages in other workspaces via path, eliminating duplication in monorepo setups.

A new .venv file option now points to centralized virtual environments, streamlining team workflows. Security-conscious teams gain opt-in malware scanning via audit.malware-check and audit.malware-check-url settings, integrating lightweight threat detection into dependency resolution. Performance improves with a fix for quadratic slowdowns when resolving transitive version conflicts, a long-standing bottleneck in large dependency trees. The update also updates Windows timezone data to IANA 2026c and refines error handling for unsupported uv pip compile flags.venv-clifor, and venvuv uv venvtool for Python projects, replacingpip, poetry, and py. Despite its speed and breadth, uv’s unified approach may overspecialize for teams invested in existing toolchains like Poetry or Hatch, where migration costs outweigh incremental gains.
The catch: While uv consolidates many Python tooling functions, its all-in-one design can complicate gradual adoption in enterprises with entrenched, workflow-specific dependencies on tools like pipx or virtualenvwrapper.

Previously in The Times “covered” — Jul 22

Use Cases
  • Manage monorepo Python projects with shared dependencies across teams
  • Automate secure dependency scanning in CI pipelines for compliance
  • Replace multiple Python tooling commands with a single, faster workflow

Source: astral-sh/uv — based on the README and release notes.

More Stories

Sway Compiler Patch Fixes Help Links and Publishing Workflow 🔗

v0.71.2 resolves documentation and release tooling issues in the Fuel blockchain language

FuelLabs/sway · Rust · 61.6k stars Est. 2021

The Sway compiler received a patch release v0.71.

2 that fixes broken help links requiring compiler version awareness and corrects crates.io publishing. The update addresses two specific pull requests: one adjusting help output behavior and another ensuring proper version bumping and artifact distribution. Sway, a Rust-inspired language for Fuel smart contracts, continues active development with daily commits and over 5,000 forks.
The catch: Open issues remain high at 923, indicating ongoing stability and usability challenges despite regular updates.

Use Cases
  • Develop Fuel smart contracts using Rust-like syntax
  • Build decentralized applications on the Fuel blockchain
  • Contribute to an open-source blockchain compiler toolchain

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

Moby Project patches critical Docker Engine security flaws in latest release 🔗

Fixes address command injection, panic triggers, and filesystem bypasses in container tooling

moby/moby · Go · 71.9k stars Est. 2013

The Moby Project released docker-v29.6.

2, patching five CVEs in Buildkit including Git source checkout command injection (CVE-2026-15793) and WCOW cache mount NTFS junction bypass (CVE-2026-15788). Updates also include containerd static binaries to v2.2. The project remains a modular toolkit for assembling container systems, offering swappable components like build tools, registry, and runtime. Despite recent activity, it carries 3,840 open issues and targets developers, not end users, with minimal UX focus.
The catch: Its developer-first design means limited out-of-the-box usability for teams needing polished, end-user tooling.

Use Cases
  • Engineers building custom container runtimes
  • Integrators swapping container orchestration components
  • Developers experimenting with modular buildkit frontends

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

Hugo v0.164.0 Adds Dark/Light Syntax Highlighting and Crypto Template Functions 🔗

Release fixes performance regression affecting large sites from v0.128.0

gohugoio/hugo · Go · 89.1k stars Est. 2013

Hugo v0.164.

0 introduces dark/light style pairs for the Chroma highlighter, new template functions encoding.HexEncode, encoding.HexDecode, and crypto.Hash, plus Pandoc citation support and RST syntax highlighting. The release also patches a performance regression impacting large sites. Builders gain finer control over code styling and cryptographic operations in templates. The catch: Open issues remain at 236, indicating ongoing maintenance demands despite recent fixes.

Use Cases
  • Developers build documentation sites fast
  • Designers theme blogs with custom syntax highlights
  • Teams deploy static sites via Hugo Deploy

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

Quick Hits

codex-auth A Zig CLI tool to seamlessly switch and manage multiple Codex accounts for streamlined AI workflow control. 2.4k
lede The C-based source tree for Lean’s LEDE, enabling customizable, lightweight router firmware for embedded networking devices. 31.5k
prometheus A Go-powered monitoring system and time series database that collects, stores, and queries metrics at scale for observability. 65.3k
RuView Rust-based RuView uses WiFi signals to detect presence, vital signs, and spatial layout in real time — no cameras or video needed. 84.5k
tmux The C source code for tmux, a terminal multiplexer that lets users manage multiple sessions, windows, and panes efficiently. 48k

HackRF Pro Gains Larger SPI Flash Access in Firmware Update 🔗

v2026.01.3 fixes mixer lock failures and expands storage for custom firmware on HackRF Pro hardware.

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

Great Scott Gadgets released v2026.01.3 of the HackRF firmware, resolving mixer frequency lock failures that disrupted stable signal reception under certain conditions.

The update also enables access to the larger SPI flash chip on HackRF Pro models, allowing developers to store bigger bitstreams or custom firmware directly on the device. This improves flexibility for field-deployed SDR applications where reprogramming via USB isn’t practical. The HackRF platform remains a low-cost, open-source software-defined radio covering 1 MHz to 6 GHz, with hardware designs and software hosted in the C-based greatscottgadgets/hackrf repository. Documentation is maintained in the docs folder and can be built locally with Sphinx, while community support relies on GitHub issues and Discord, with technical support tickets typically addressed within two weeks. The project sees steady activity, with the last commit just zero days ago and 80 open issues indicating ongoing maintenance.
The catch: Despite the hardware’s wide frequency range, achieving optimal performance often requires external filters and amplifiers, as the stock HackRF Pro front-end lacks sufficient selectivity for congested RF environments.

Previously in The Times “covered” — Jul 17

Use Cases
  • Field engineers capturing wideband signals for spectrum monitoring
  • Researchers prototyping custom wireless protocols in lab settings
  • Educators teaching SDR concepts with affordable, open hardware

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

More Stories

Automotive Skills Suite Adds 152 Claude Tools for Standardized Engineering Outputs 🔗

Builder-reviewer pairs automate deliverables from ISO 26262 to AUTOSAR with xlsx traceability

jherrodthomas/automotive-skills-suite · Unknown · 2.3k stars 2mo old

The jherrodthomas/automotive-skills-suite provides 152 installable .skill files for Claude, split evenly between builder and confirmation reviewer roles.

Each pair generates standardized Excel deliverables across automotive engineering lifecycles, from functional safety (ISO 26262) to cybersecurity (ISO/SAE 21434) and quality (AIAG-VDA). Skills consume upstream outputs as stable xlsx contracts, enabling traceable V&V, calibration, and MBSE workflows. The chain-enforced format ensures downstream consistency in artifacts like FMEAs, test plans, and control plans.
The catch: Despite recent activity, five open issues remain unresolved, raising questions about long-term maintenance and edge-case handling in complex toolchains.

Previously in The Times “covered” — Jul 17

Use Cases
  • Systems engineer generates ISO 26262 safety case with automated reviewer feedback
  • Quality lead produces APQP/PPAP documents with built-in confirmation validation
  • AUTOSAR developer exports SWC compositions and RTE configurations via paired skills

Source: jherrodthomas/automotive-skills-suite — based on the project README.

Rust proxy bridges Android Auto wireless to car USB head units 🔗

Uses io_uring and MITM mode to enable EV routing and display tweaks

aa-proxy/aa-proxy-rs · Rust · 434 stars Est. 2024

The aa-proxy-rs project lets builders connect a wireless Android phone to a USB car head unit for Android Auto, using Rust and io_uring for efficient I/O. It supports automatic reconnection, bandwidth monitoring, and an embedded web UI for control.

Recent v0.22.0 adds protocol version override and Vehicle Energy Forecast publishing over WebSocket, plus album art injection with configurable placement. MITM mode allows tweaks like removing tap restrictions, enabling video in motion, and per-vehicle SDR UI overrides.
The catch: 18 open issues and reliance on niche Linux kernels may deter builders seeking stable, turnkey solutions.

Use Cases
  • DIYers enabling wireless Android Auto in older cars
  • Developers testing Android Auto modifications safely
  • EV owners adding Google Maps EV routing via proxy

Source: aa-proxy/aa-proxy-rs — based on the README and release notes.

Quick Hits

project_aura 21cncstudio/project_aura (C): A compact ESP32-S3 air-quality station with real-time LVGL UI, MQTT streaming, and seamless Home Assistant integration for smart home monitoring. 708
litex enjoy-digital/litex (Python): LiteX simplifies FPGA-based hardware design by letting builders create custom SoCs and peripherals with intuitive Python scripts and reusable IP cores. 4k
node-feature-discovery kubernetes-sigs/node-feature-discovery (Go): Automatically detects and advertises hardware features (CPU, GPU, NICs) of Kubernetes nodes to enable intelligent workload scheduling and resource optimization. 1.1k
glasgow GlasgowEmbedded/glasgow (Python): A versatile open-source hardware toolkit that acts as a “Scots Army Knife” for electronics — combining logic analyzer, JTAG, UART, SPI, and more in one USB device. 2.2k
espectre francescopace/espectre (Python): ESPectre uses Wi-Fi CSI spectre analysis to detect motion through walls, offering privacy-preserving sensing with Home Assistant integration for smart automation. 8.8k
Embedded-Engineering-Roadmap m3y54m/Embedded-Engineering-Roadmap (Unknown): A curated, structured learning path for embedded systems engineers, guiding beginners to experts through essential topics, tools, and hands-on projects. 12.4k

PlayCanvas Engine Adds Gaussian Splatting Support and XR Improvements in v2.21.0 🔗

Latest release includes breaking changes to mesh handling and introduces volumetric fog, KTX2 cubemaps, and two-hand XR manipulation.

playcanvas/engine · JavaScript · 16.3k stars Est. 2014 · Latest: v2.21.0

The PlayCanvas Engine updated to v2.21.0 with native support for the KHR_gaussian_splatting glTF extension, enabling direct loading and rendering of 3D Gaussian Splats without external converters.

Developers can now import .splat files via the engine’s pluggable resource parser system, streamlining workflows for photorealistic 3D content.

Volumetric fog rendering was added to CameraFrame, enhancing atmospheric depth in scenes, while KTX2 cubemap support improves texture efficiency and loading performance. XR capabilities expanded with the new XrManipulation script, allowing two-hand drag, rotate, and scale interactions in immersive environments, alongside refined teleport visuals using ballistic arcs and ring indicators.

Breaking changes include removal of MeshInstance parameter pass flags and renaming of gsplat script classes to align with core GSplat casing, requiring updates to existing projects using custom mesh instancing or splat scripts.

The catch: Despite active development, the engine’s WebGPU path lags behind WebGPU feature parity, and advanced rendering techniques like ray tracing remain unsupported, limiting high-end visual fidelity on compatible hardware.

Previously in The Times “covered” — Jul 17

Use Cases
  • Build browser-based 3D product configurators with real-time lighting
  • Create WebXR training simulations with hand-tracked object manipulation
  • Deploy glTF-native Gaussian Splats for immersive cultural heritage visualizations

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

More Stories

Nakama Adds Samsung Galaxy Store IAP Support in v3.40.0 🔗

Release fixes leaderboard scheduler race conditions and group admin deletion bugs

heroiclabs/nakama · Go · 12.9k stars Est. 2017

Nakama v3.40.

0 adds Samsung Galaxy Store purchase validation with new runtime hooks and config options. It improves Apple JWS receipt validation and introduces a console setting to disable display name search. Fixes resolve orphaned group edges on admin deletion and race conditions in leaderboard/tournament schedulers that could block hook triggers. The release requires nakama-common v1.47.0 for Go runtime users.
The catch: Open issues remain at 122, indicating ongoing maintenance demands despite recent activity.

Use Cases
  • Game devs adding cross-platform IAP validation
  • Studios scaling realtime multiplayer with matchmaking
  • Apps requiring persistent chat and social graphs

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

Godot 4.7.1 stabilizes engine with bug fixes for cross-platform devs 🔗

Maintenance release improves usability without breaking existing projects

godotengine/godot · C++ · 114.5k stars Est. 2014

Godot Engine’s 4.7.

1 update resolves stability and usability issues across its 2D and 3D toolchain, ensuring smoother exports to desktop, mobile, and web platforms. The MIT-licensed C++ engine maintains backward compatibility, letting teams adopt the patch without reworking assets or scripts. Active community contributions continue to refine the editor and export templates, with 18,570 open issues indicating ongoing refinement. The catch: Despite frequent commits, the high volume of unresolved issues suggests lingering edge cases in complex 3D workflows or niche platform exports that may require manual debugging.

Previously in The Times “covered” — Jul 19

Use Cases
  • Indie studios shipping 2D pixel games to Steam and itch.io
  • Mobile developers prototyping touch-based puzzles for iOS and Android
  • Educational teams teaching game design using a free, open-source engine

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

Flecs v4.1.6 adds data-oriented hierarchies and scripted entity creation 🔗

Release includes 130 bugfixes, 500+ tests, and new entity range API for safe recycling

SanderMertens/flecs · C · 8.5k stars Est. 2018

The latest Flecs release introduces data-oriented hierarchies, enabling efficient parent-child entity relationships without runtime overhead. A new entity range API ensures safe handling of recycled IDs during iteration.

Scripting now supports include statements, functions, and direct component assignment via expressions like Rgb: color(x, z).rgb. The explorer UI gained multi-panel inspection, entity creation, and component dialogs. Built on a zero-dependency C99 core with a type-safe C++17 API, Flecs processes millions of entities per frame using cache-friendly archetype storage.
The catch: Despite 13,000+ CI tests, the project’s single-maintainer model raises long-term sustainability concerns for enterprise adoption.

Use Cases
  • Game developers building large-scale simulations with millions of entities
  • Engine creators needing lockless multithreaded systems and prefab support
  • Teams requiring runtime reflection, JSON serialization, and unit-annotated components

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

Quick Hits

bevy Bevy empowers builders with a refreshingly simple, data-driven Rust game engine that prioritizes modularity and performance for rapid iteration. 47.3k
SpacetimeDB SpacetimeDB lets builders develop real-time applications at light speed by treating databases as first-class, queryable state machines. 24.8k
gdUnit4 gdUnit4 gives Godot 4 builders an embedded, feature-rich unit testing framework with TDD support, scene testing, and mocking for GDScript and C#. 1.2k
bgfx bgfx offers builders a cross-platform, graphics API-agnostic rendering library that lets them bring their own engine while leveraging high-performance GPU access. 17.3k
Alpha-Piscium Alpha-Piscium delivers builders a high-quality, realistic Minecraft shaderpack in GLSL that enhances lighting, shadows, and atmosphere without sacrificing performance. 149
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