Design
Preset
Background
Text
Font
Size
Width
Account Pricing Saturday, July 25, 2026

The Git Times

“The best material model of a cat is another, or preferably the same, cat.” — Norbert Wiener

AI Models
Claude Opus 5 $25/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.

Redis 8.8.1 patches critical security flaw in Bloom filter module 🔗

Out-of-bounds write fix prevents remote code execution in RedisBloom and TDigest

redis/redis · C · 75.7k stars Est. 2009 · Latest: 8.8.1

Redis remains the go-to in-memory data structure server for real-time applications, offering sub-millisecond latency as a cache, message broker, and vector search engine. The latest 8.8.

1 release addresses a security vulnerability where maliciously crafted RESTORE payloads in RedisBloom and TDigest modules could trigger out-of-bounds writes, enabling potential remote code execution. Maintained in C, Redis supports JSON, time-series, and geospatial data types with built-in replication and Lua scripting. Developers use it for session stores, real-time analytics, and pub/sub systems across cloud and on-premises environments. Its modular design allows selective loading of features like RedisSearch and RedisML to minimize footprint. Despite its maturity, the project shows active maintenance with daily commits and broad platform support from Alpine to macOS Tahoe.
The catch: Redis’s single-threaded command processing can become a bottleneck under high write throughput, requiring careful sharding or migration to Redis Cluster for scaling beyond a single node’s capacity.

Why this leads today Redis’s growing use in real-time applications highlights its role as essential infrastructure for developers needing fast, scalable data access without major system changes.

Use Cases
  • Caching user sessions for low-latency web applications
  • Real-time leaderboards and counting with sorted sets
  • Vector similarity search for AI-powered recommendation engines

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

More on the Front Page

Builderz Labs patches critical vulnerabilities in Mission Control agent control plane 🔗

Version 2.3.0 fixes 13 dependency flaws including SSRF and DoS risks in Next.js and Sharp

builderz-labs/mission-control · TypeScript · 5.8k stars 5mo old

Builderz Labs released Mission Control v2.3.

0, patching 13 known dependency vulnerabilities — eight rated high severity — urging all self-hosted operators to upgrade immediately. The update addresses SSRF in Server Actions, middleware bypass, and response cache confusion in Next.js, plus libvips-related flaws in Sharp and DoS risks in brace-expansion. No application code changes were needed; the protected quality gate held.

Mission Control remains a self-hosted TypeScript control plane for AI agents, built with Next.js and backed by SQLite. It dispatches tasks, inspects runs, tracks spend, and coordinates runtimes like OpenClaw, Claude Code, and Codex through a local dashboard at http://localhost:3000. Operators deploy via docker compose up or install locally with Node 22 and pnpm, then configure agent access via API keys from Settings.

The release adds MC_DISABLE_RUNTIME_SCAN to skip CLI probing during detection and restores locale parity across 10 languages. Test tooling updated to Vite 8 and Vitest 4, though production builds still use webpack. No database migrations or config changes are required for upgrade.

The catch: As alpha software, Mission Control’s APIs, schemas, and configuration may still change between releases, posing integration risks for teams relying on stable interfaces in production agent workflows.

Use Cases
  • AI agent teams tracking task execution and spend locally
  • Developers orchestrating Claude Code and Codex runs via dashboard
  • Operators hardening self-hosted agent deployments with Docker compose

Source: builderz-labs/mission-control — based on the README and release notes.

Grok App wraps CLI into a desktop workbench for builders 🔗

Adds session management, worktrees, and plugin support to local Grok Build CLI

RongleCat/grok-app · TypeScript · 250 stars 1d old

RongleCat/grok-app is a Tauri 2 desktop workbench that wraps the local grok agent stdio CLI into a graphical interface. It provides project and session management, multi-session streaming, Git worktree switching, and permission controls.

Features include voice dictation, plugin installation, sandbox profiles, and CLI update checks. Built with TypeScript, React, and Vite, it requires a working Grok Build CLI installation to function. The app supports importing CLI sessions and offers bilingual UI in English and Chinese.
The catch: As an unofficial wrapper, it depends on the Grok Build CLI and may lag behind official tooling updates or xAI product changes.

Use Cases
  • Developers managing multiple Grok agent sessions across projects
  • Teams using Git worktrees who need quick context switching
  • Builders wanting GUI controls for agent permissions and automation scheduling

Source: RongleCat/grok-app — based on the README and release notes.

Agentacct Shows Real Agent Work and Cost on Your Machine 🔗

It reads local Claude Code and Codex logs to display token usage, tasks, and honest cost estimates.

mikehasa/agentacct · Python · 248 stars 0d old

Agentacct is a Python tool that builds a local dashboard from coding-agent session logs, showing tokens, estimated cost, and recorded work steps like passing tests. It joins usage with work using confidence labels and avoids guessing when data is missing.

Install via pipx, run agentacct onboard to detect logs and start the dashboard at http://127.0.0.1:8765. Everything stays on your machine — no telemetry, no cloud, no API keys. The catch: it requires Python 3.11+ and currently supports only macOS, Linux, or Windows via WSL, limiting native Windows use.

Use Cases
  • Developers track actual token use and cost per agent session
  • Teams verify agent work with evidence like test results and task logs
  • Engineers audit coding-agent activity without external tracking

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

Caspian SDK Unifies Agent Messaging Across 10+ Platforms With One Handler 🔗

Open-source adapters eliminate repetitive channel plumbing for AI agent developers.

TryCaspian/caspian-sdk · Python · 222 stars 4d old

Caspian SDK provides a single on_message handler for agents to communicate via Slack, Discord, Telegram, email, SMS, X, and more. Developers install the Python or TypeScript SDK, use the CLI to connect channels, and write one function to handle all incoming messages.

The SDK manages threading, deduplication, webhook verification, and platform-specific quirks behind a consistent interface. Hosted channels like WhatsApp and iMessage are available via Caspian’s service with the same API.
The catch: Early adoption shows 49 open issues and reliance on hosted services for key channels raises questions about long-term self-hosting flexibility and production-scale reliability.

Use Cases
  • AI agents sending alerts via email and Slack
  • Bots replying to users across Discord and Telegram
  • CrewAI agents coordinating through unified messaging channels

Source: TryCaspian/caspian-sdk — based on the README and release notes.

Telegram Bot Lets Users Copy Polymarket Whale Trades in One Tap 🔗

Real-time alerts and instant signing bypass Polymarket’s laggy confirmation UI

melloworchid8rr6g/TG-Polymarket-bot · JavaScript · 313 stars 0d old

The melloworchid8rr6g/TG-Polymarket-bot is a JavaScript Telegram bot that monitors Polymarket for large trades and lets users replicate them with a single tap. By embedding wallet signing directly in the bot, it avoids the confirmation popups that can lag or fail on Polymarket’s native interface.

Users set custom whale trade thresholds via alerts and execute bets instantly through the official Polymarket API. The bot also enables free deposits and withdrawals, treating its wallet as a direct extension of the user’s Polymarket account. A public alerts channel offers real-time whale trade notifications at no cost.
The catch: The bot’s free access expires on 10.10.2026, raising questions about long-term sustainability beyond that date.

Use Cases
  • Traders replicate large Polymarket bets instantly via Telegram
  • Users receive real-time whale trade alerts on preferred thresholds
  • Investors deposit/withdraw funds without leaving the bot interface

Source: melloworchid8rr6g/TG-Polymarket-bot — based on the project README.

NVIDIA NemoClaw secures AI agents in OpenShell with managed inference 🔗

Reference stack adds sandboxing, lifecycle controls for Hermes and LangChain agents

NVIDIA/NemoClaw · TypeScript · 21.9k stars 4mo old

NemoClaw provides a hardened blueprint for running AI agents like Hermes and LangChain Deep Agents securely inside NVIDIA OpenShell sandboxes. It delivers guided onboarding, routed inference, network policy enforcement, and lifecycle management via a single CLI.

The project supports OpenClaw by default, with aliases for Hermes and LangChain agents, and includes a managed nemoclaw-light skin for readable terminal output. Recent commits show active maintenance, with the last push 24 days ago. The catch: 220 open issues suggest ongoing stability and feature gaps despite regular updates.

Use Cases
  • Developers sandboxing Hermes agents for secure LLM workflows
  • Teams deploying LangChain Deep Agents with enforced network policies
  • Engineers managing OpenClaw agent lifecycles via CLI in OpenShell

Source: NVIDIA/NemoClaw — based on the project README.

Zeus Panel Deploys Serverless VLESS Proxy Management on Cloudflare Workers 🔗

Features dynamic IP rotation and anti-filtering for tenant-controlled edge proxy services

zeus-panel/ZEUS-PANEL · JavaScript · 403 stars 4d old

Zeus Panel provides a multi-tenant VLESS network proxy management interface running natively on Cloudflare Workers with D1 SQL storage. It enables administrators to bind geolocated IPs, enforce traffic and device quotas, and automate quota resets per user.

Built-in TLS fingerprint simulation and DoH-based content blocking help evade network filtering while maintaining performance. The platform supports bulk user operations, real-time quota monitoring, and OTA updates without data loss. One-click deployment provisions the panel, subdomain, and database directly from the interface. The catch: Despite zero open issues and recent activity, long-term reliability under high tenant density remains unverified at scale.

Use Cases
  • ISPs offering geo-targeted proxy access to regional subscribers
  • Developers managing encrypted traffic routing for remote work tools
  • Enterprises enforcing content policies via tenant-specific proxy chains

Source: zeus-panel/ZEUS-PANEL — based on the project README.

AI Agents Evolve Beyond Chat Into Autonomous Workflows 🔗

Open source shifts from conversational bots to task-driven, multi-modal agents with real-world integration

The open source AI agent landscape is rapidly maturing from simple LLM wrappers into purpose-built systems that perceive, reason, and act across domains. Projects like powerycy/goutoujunshi embed multidisciplinary knowledge — psychology, law, philosophy — to deliver emotionally aware, actionable relationship advice, moving beyond generic chat.

Similarly, calesthio/OpenMontage transforms coding agents into full video production studios with 12 pipelines and 500+ agent skills, enabling end-to-end media generation from natural language.

Integration is becoming seamless: TryCaspian/caspian-sdk provides a unified identity layer across Slack, Discord, WhatsApp, and SMS via a single on_message handler, reducing fragmentation in multi-channel agent deployment. For developers, mikehasa/agentacct offers local-first work intelligence — logging agent activity transparently without JavaScript — enabling auditability and trust in coding agents. Enterprise needs are met by iflytek/skillhub, a self-hosted skill registry with RBAC and Kubernetes support for governed agent skill distribution.

Specialization is rising: HKUDS/Vibe-Trading deploys agents for personalized financial strategies, while text-to-cad and archify turn agents into domain experts for CAD modeling and architecture diagrams and hardware design. Even niche skills like blader/humanizer — which strips AI tells from text — show how agent capabilities are being modularized and shared.

The catch: Despite impressive scope, many agents remain brittle outside narrow prompts, struggle with long-term coherence, and lack standardized safety or evaluation frameworks — making production reliability an open challenge rather than a solved pattern.

Use Cases
  • Developers audit agent coding activity locally
  • Enterprises deploy governed agent skills at scale
  • Creators generate videos from natural language prompts

AI Agents Drive Unified Interface Abstraction in Open Source 🔗

New frameworks unify messaging, auth, and tooling across platforms through standardized SDKs and adapters

A clear pattern emerges in open source: developers are building abstraction layers to simplify complex integrations. Projects like TryCaspian/caspian-sdk offer a single on_message handler for AI agents across Slack, Discord, Telegram, and more, eliminating per-platform boilerplate.

Similarly, oomol-lab/open-connector provides a unified auth gateway connecting AI agents to 1000+ SaaS tools via SDK, CLI, and MCP. builderz-labs/mission-control extends this to a self-hosted control plane for dispatching tasks and tracking spend across AI agent runtimes like OpenClaw and Claude Code. Meanwhile, ZhongAnTech/zarm and learnhouse/learnhouse showcase modular, React/TypeScript-based UI systems enabling rapid composition of interfaces. Even niche tools like lichess-bot-devs/lichess-bot abstract bot-engine communication, while microsoft/Ontology-Playground offers zero-backend ontology design. These repos reflect a shift toward composable, interoperable infrastructure — where SDKs, adapters, and control planes reduce integration friction. The focus is on portability: write once, deploy everywhere, with minimal context switching.
The catch: Much of this abstraction remains early-stage, with overlapping efforts in auth (MCP vs. custom SDKs), unclear governance, and limited real-world validation at scale — raising concerns about fragmentation and long-term maintenance burdens despite promising demos.

Use Cases
  • Developers deploy AI agents across chat platforms
  • Teams connect agents to internal SaaS tools securely
  • Engineers manage agent lifecycles with centralized observability

Open Source LLM Tools Forge Unified AI Agent Infrastructure 🔗

Modular SDKs, adapters, and skill ecosystems enable interoperable, self-hosted AI workflows across platforms and models

A clear pattern is emerging in open source LLM tooling: the shift from isolated model wrappers to composable, infrastructure-grade ecosystems for AI agents. Projects like TryCaspian/caspian-sdk provide unified channel adapters (Slack, Discord, WhatsApp, etc.

) with a single on_message handler, while builderz-labs/mission-control offers a self-hosted control plane for dispatching tasks, tracking spend, and managing runtimes like OpenClaw and Claude Code. Meanwhile, diegosouzapw/OmniRoute acts as a free, MIT-licensed AI gateway supporting 290+ providers with quota-aware fallbacks and token compression, enabling seamless model switching.

This infrastructure is being extended through specialized skills: alicorehaines31/marketingskills packs CRO and SEO tools for Claude Code, Imbad0202/academic-research-skills automates literature review loops, and Donchitos/Claude-Code-Game-Studios simulates a full game dev studio with 49 AI agents. Even niche capabilities like bradautomates/claude-video (video understanding via frame extraction) and Wei-Shaw/sub2api (subscription pooling for LLM APIs) reflect a drive toward modular, reusable capabilities.

The catch: While promising, this ecosystem remains fragmented across languages (Python, TypeScript, Go, Rust), with overlapping scope and unclear governance—many tools assume Claude Code as a runtime, creating vendor-adjacent lock-in risks despite MIT licenses, and real-world production validation at scale is still sparse outside internal demos or contributor forks.

Use Cases
  • Developers build cross-platform AI agents with consistent identity
  • Teams self-host agent orchestration with cost tracking and runtime flexibility
  • Enterprises deploy modular AI skills for domain-specific workflow automation

Deep Cuts

AI-Powered Cinematic Frame Generator for 21:9 Screens 🔗

Transforms text prompts into ultra-widescreen movie stills with cinematic depth and lighting

dacnay816y62-hub/cinema-dna-21x9x3 · Unknown · 331 stars

cinema-dna-21x9x3 is a niche generative tool crafting high-fidelity, ultra-widescreen (21:9) film frames from text descriptions. Built for visual storytellers, it simulates cinematic lighting, depth of field, and film grain to produce stills that feel lifted from a director’s cut.

Though language-agnostic in interface, its strength lies in aspect-ratio-specific composition — ideal for pitch decks, concept art, or immersive UI mockups. Developers can integrate it via API to prototype visual narratives without footage.
The catch: Early-stage with sparse documentation and limited community examples, making adoption risky for production pipelines despite its evocative output.

Use Cases
  • Indie filmmakers pitching visual tone to investors
  • Game artists designing cinematic cutscene concepts
  • UX designers prototyping immersive media app interfaces

Source: dacnay816y62-hub/cinema-dna-21x9x3 — based on the project README.

Quick Hits

freeCodeCamp freeCodeCamp/freeCodeCamp: Learn math, programming, and computer science for free through an open-source curriculum and interactive coding challenges. 452.8k
RapidRAW CyberTimon/RapidRAW: Edit RAW images non-destructively with GPU acceleration for fast, professional-grade photo editing in a beautiful interface. 9k
jiuwenswarm openJiuwen-ai/jiuwenswarm: Access powerful AI agents via daily chat apps to extend LLM capabilities for smarter, context-aware interactions. 1.8k
esp32-ai slvDev/esp32-ai: Run lightweight AI models on ESP32 microcontrollers for edge-based smart device automation and sensing. 553
drawdb drawdb-io/drawdb: Design database schemas visually and generate SQL instantly with a free, intuitive online diagramming tool. 38.2k
codex-router duolahypercho/codex-router: Seamlessly route Codex requests to external models like Kimi and DeepSeek with secure OAuth, migration, and rollback support. 206
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 Control UI and Onboarding for Personal AI Assistants 🔗

Release adds GPT-5.6 support, native app updates, and improved cross-platform channel integrations for local-first workflows.

openclaw/openclaw · TypeScript · 384.1k stars 8mo old · Latest: v2026.7.1

OpenClaw’s latest release, v2026.7.1, delivers a comprehensive overhaul of its Control UI and onboarding flow, aiming to reduce friction for new users while enhancing daily workflows.

The updated interface now supports side-by-side conversation views with live Tasks, clearer chat controls, and integrated usage and cost tracking. Onboarding has been streamlined with guided setup that validates connections before saving and preserves user progress if interrupted.

Official iOS, Android, and macOS apps received substantial updates, improving voice interaction, file handling, offline reading, and connection recovery. Gateway stability has been addressed, with fixes for crash loops, scheduled work, and remote browser control. The release expands model compatibility to include GPT-5.6, Tencent Hy3, and Meta Muse Spark 1.1, strengthening its appeal for developers seeking flexible, private AI agent orchestration.

Telegram, Slack, Discord, and Apple Messages each gained notable improvements, reinforcing OpenClaw’s promise of meeting users where they already communicate. With over 3,000 contributions from 532 developers in this release alone, the project shows sustained community momentum.

The catch: Despite its broad channel support and local-first design, OpenClaw remains dependent on Node.js 24+ and requires manual daemon setup, which may deter users seeking zero-configuration or lightweight edge deployments.

Previously in The Times “covered” — Jul 23

Use Cases
  • Developers building private AI workflows across messaging platforms
  • Teams seeking self-hosted assistants with multi-channel message bridging
  • Power users customizing local AI agents with coding and task automation skills

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

More Stories

Supervision adds pose estimation NMS to streamline keypoint workflows 🔗

New `KeyPoints.with_nms()` removes duplicate skeletons using box-based suppression

roboflow/supervision · Python · 48.4k stars Est. 2022

Roboflow’s supervision library now includes KeyPoints.with_nms() for pose estimation, applying non-maximum suppression to detected skeletons by deriving axis-aligned boxes from visible keypoints.

The update supports class-agnostic mode and overlap metrics like IOU or IOS, requiring detection_confidence to avoid errors. This follows recent fixes to Pascal VOC export and metric scoring for background false positives. Built for Python 3.10+, it remains model-agnostic with connectors for Ultralytics, Transformers, and MMDetection.
The catch: Active development shows rapid iteration, but 81 open issues suggest ongoing stability challenges in edge-case handling.

Use Cases
  • Real-time pose estimation in video surveillance
  • Instance segmentation pipeline with YOLO models
  • Object detection with custom annotators in retail analytics

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

Scikit-learn 1.9.0 Boosts Dataframe Interoperability with Narwhals 🔗

Release adds dependency to streamline Python data science workflows

scikit-learn/scikit-learn · Python · 66.8k stars Est. 2010

Scikit-learn’s 1.9.

0 release integrates narwhals to improve compatibility across dataframe libraries like pandas and Polars. Users can now upgrade via pip install -U scikit-learn or conda install -c conda-forge scikit-learn. The update supports Python 3.11 to 3.14 and maintains existing APIs while enabling smoother data interchange. The catch: Narwhals introduces an additional dependency that may increase installation size and complexity for minimal environments.

Use Cases
  • Data scientists preprocessing data with pandas
  • ML engineers switching between dataframe backends
  • Researchers ensuring library compatibility in pipelines

Source: scikit-learn/scikit-learn — based on the README and release notes.

Tesseract OCR Adds Portuguese Language Support in Latest Release 🔗

Release 5.5.3 includes updated language strings and improved Windows installer documentation.

tesseract-ocr/tesseract · C++ · 75.5k stars Est. 2014

The tesseract-ocr/tesseract project released version 5.5.

3, adding Portuguese language support and refining section descriptions in the Windows installer. Contributors also corrected Slovak language strings and fixed NSIS language ID masking for reliable auto-detection. Recent commits address typos, multi-page PAGE XML output tags, and Apple cross-compilation CMake settings. The project maintains LSTM-based OCR for line recognition alongside legacy engine compatibility.
The catch: Despite active development, 478 open issues persist, including long-standing requests for GUI integration and improved complex layout handling.

Use Cases
  • Developers extract text from scanned documents using command-line by running Linux
  • Researchers process historical archives with multilingual OCR via Tesseract 5.5
  • Teams automate invoice data entry by integrating libtesseract into Python pipelines

Source: tesseract-ocr/tesseract — based on the README and release notes.

Quick Hits

DeepSpeed DeepSpeed optimizes distributed training and inference for deep learning, making large-scale AI development faster, more efficient, and accessible. 42.8k
keras Keras simplifies deep learning with an intuitive, high-level API that accelerates model building and experimentation for researchers and engineers. 64.2k
dify Dify enables teams to build, deploy, and scale agentic workflows and RAG pipelines in a unified workspace — no stack rebuild needed from prototype to production. 150.2k
open-webui Open WebUI provides a clean, user-friendly interface to interact with LLMs via Ollama, OpenAI, and other backends, streamlining AI experimentation and deployment. 146.7k
netron Netron offers an interactive, browser-based visualizer for neural network and ML models, letting builders inspect architecture, weights, and operations instantly. 33.3k

ROSClaw adds verified physical receipts to agent actions 🔗

New audit trail binds intent to robot hardware with policy enforcement

ros-claw/rosclaw · Python · 170 stars 4mo old

ROSClaw now issues cryptographically signed ExecutionReceipt objects that prove an agent’s action ran on specific hardware under defined policies. The receipt captures intent, capability routing, arbitration outcome, and driver dispatch, creating an auditable trail from LLM command to physical motion. This closes the trust gap between simulated plans and real-world execution, a persistent issue in embodied AI where agents hallucinate feasible actions.

The runtime enforces safety via SO_PEERCRED identity checks and private state isolation between agent sessions. Unauthorized REAL actions are blocked at the rosclawd layer unless hardware dispatch is explicitly permitted, failing closed by design. Recent system tests confirmed all 22 MCP tools work in simulation with the UR5e arm, and receipt verification is now part of the CI pipeline.

Memory and skill evolution subsystems consume these receipts asynchronously, ensuring learning is grounded in actual evidence—not just simulated success. Developers can introspect receipts via rosclaw doctor --level verified to validate policy compliance post-execution.

The catch: Hardware acceptance testing remains pending; real-world UR5e validation and site-specific ACL wiring are not yet complete, limiting deployment to controlled lab or simulation environments for now.

Previously in The Times “covered” — Jul 24

Use Cases
  • Factory agents logging compliant assembly actions
  • Research robots validating sim-to-real policy transfer
  • Autonomous agents generating auditable task trails

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

More Stories

Kornia adds batched VisualPrompter inference and XFeat feature detection 🔗

v0.8.3 patch improves geometry ops speed and adds half-precision support for PyTorch workflows

kornia/kornia · Python · 11.3k stars Est. 2018

Kornia v0.8.

3 introduces batched inference for VisualPrompter and adds XFeat and ALIKED as new feature detectors and descriptors. Core geometry operations like depth_to_normals now run up to 3x faster, with transform points and denormalize_points_with_intrinsics seeing ~2x speedups. The release adds float16/bfloat16 support and improves ONNX export and MPS device compatibility. Over 175 commits include expanded docstrings, test coverage, and DeprecationWarnings for kornia.utils utilities.
The catch: Despite active development, 65 open issues suggest ongoing challenges in stabilizing newer features like VLM/VLA integrations amid shifting focus toward end-to-end vision models.

Use Cases
  • Engineers building spatial AI pipelines with differentiable image transforms
  • Researchers training vision models using automated augmentation pipelines
  • Developers integrating pre-trained feature matchers like LoFTR or LightGlue into SLAM systems

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

Webots R2025a Boosts ROS 2 Support and Adds New Robot Models 🔗

Open-source simulator gains cross-platform binaries and improved physics for autonomous systems

cyberbotics/webots · C++ · 4.5k stars Est. 2018

Cyberbotics released Webots R2025a with expanded ROS 2 integration, new robot models, and demo worlds. The update includes pre-built binaries for Windows, Linux (Deb, Snap, Docker), and macOS, targeting Intel/AMD 64-bit systems.

Developers can simulate autonomous vehicles, robotic arms, and fluid dynamics using its C++-based 3D engine and physics engine. Tutorials and cloud simulations at webots.cloud lower the barrier for newcomers. Despite active development, the simulator’s steep learning curve for advanced customization remains a hurdle for teams needing rapid deployment.
The catch: Custom controller development requires deep C++ proficiency, limiting accessibility for Python-first robotics teams.

Use Cases
  • Engineers simulate autonomous vehicle navigation in urban environments
  • Researchers test ROS 2-based robot swarms in physics-accurate settings
  • Students learn kinematics using pre-built robotic arm models and tutorials

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

URDF Studio adds AI-powered robot generation and MuJoCo export 🔗

The browser-based tool streamlines URDF design with visual editing and hardware configuration.

OpenLegged/URDF-Studio · TypeScript · 443 stars 7mo old

URDF Studio is a TypeScript web app for designing robot models without editing raw XML. It offers topology editing, collision optimization, and hardware parameter setup through a unified interface.

Users can assemble multi-robot workspaces, integrate a motor library, and export to MuJoCo or USD formats. An AI assistant generates robot designs and outputs inspection reports in PDF or CSV. The app uses React Three Fiber for 3D visualization and manages versions separately for the workspace and published @urdf-studio/react-robot-canvas package.
The catch: With only one open issue and infrequent commits, long-term maintenance and community support remain uncertain for builders relying on it in production.

Use Cases
  • Robotics engineers designing arm kinematics visually
  • Researchers exporting URDF models to MuJoCo for simulation
  • Educators teaching robot structure via browser-based modeling

Source: OpenLegged/URDF-Studio — based on the README and release notes.

Quick Hits

mavros mavlink/mavros bridges MAVLink and ROS with a GCS proxy, enabling seamless communication between drones and robotic systems for real-time control and telemetry. 1.2k
ros2_controllers ros-controls/ros2_controllers provides reusable, high-performance robotic controllers built for ros2_control, streamlining development of precise motion and force feedback in ROS 2 applications. 806
copper-rs copper-project/copper-rs lets you build, run, and replay entire robot systems deterministically in Rust, ensuring reproducible behavior for testing, simulation, and deployment. 1.4k
awesome-robot-descriptions robot-descriptions/awesome-robot-descriptions curates a collection of high-quality URDF and MJCF robot models, saving builders time by offering trusted, ready-to-use descriptions. 1.6k
stable-baselines3 DLR-RM/stable-baselines3 delivers clean, PyTorch-based implementations of RL algorithms, letting builders rapidly prototype and train robust robotic policies with proven reliability. 13.6k

Nuclei v3.11.0 enforces signed JavaScript templates to block supply-chain risks 🔗

The update closes a critical attack surface by requiring digital signatures for custom JS-based vulnerability scans

projectdiscovery/nuclei · Go · 30k stars Est. 2020 · Latest: v3.11.0

ProjectDiscovery’s Nuclei scanner now mandates digital signatures for any custom template using the javascript: protocol, effective in v3.11.0.

Unsigned JavaScript templates are skipped during loading, a shift from prior versions where they executed by default. This change hardens the tool against malicious template execution, as the JavaScript protocol exposes Go-backed modules and expands the attack surface beyond request-only scans. Official templates from the nuclei-templates repository remain unaffected—they are pre-signed with ProjectDiscovery’s public key. Users with private or community JS templates must now sign them before use; workflows referencing unsigned variants will fail silently. The move extends sandboxing and network policy checks introduced in v3.10.0, aligning JS templates with existing code-protocol safeguards. Nuclei retains its core strengths: YAML-driven templates, multi-protocol scanning (HTTP, DNS, SSL, etc.), and CI/CD integration for automated vuln detection. The catch: Signing adds overhead for teams maintaining custom JS templates, creating a barrier for lightweight or experimental scans that previously relied on unsigned, ad-hoc logic.

Previously in The Times “covered” — Jul 19

Use Cases
  • Security teams scanning APIs for logic flaws
  • DevOps integrating vuln checks into GitHub Actions
  • Bug hunters verifying subdomain takeovers at scale

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

More Stories

CL4R1T4S Publishes Leaked System Prompts for Major AI Models 🔗

Repository updated daily with extracted prompts from OpenAI, Google, Anthropic, and others

elder-plinius/CL4R1T4S · Unknown · 46.1k stars Est. 2025

The CL4R1T4S project maintains a growing archive of leaked system prompts from ChatGPT, Claude, Gemini, Grok, Perplexity, Cursor, Replit, and other AI platforms. Contributors submit prompts with model names, extraction dates, and context to reveal hidden behavioral constraints.

The project aims to increase transparency by exposing how AI labs shape model outputs through undisclosed instructions. Last updated 22 hours ago, it reflects active community extraction efforts.
The catch: Prompt authenticity and completeness remain unverified, raising concerns about potential misinformation or incomplete extracts.

Use Cases
  • Researchers analyzing AI behavior constraints
  • Developers reverse-engineering model safety filters
  • Auditors checking for biased or political defaults in AI systems

Source: elder-plinius/CL4R1T4S — based on the project README.

x64dbg Hotfix Addresses r8 Register Bug in Windows Debugger 🔗

May 2026 patch resolves trace reader faults and improves dark theme usability

x64dbg/x64dbg · C++ · 49k stars Est. 2015

The x64dbg project released a May 2026 hotfix fixing an r8 register modification issue and a trace reader memory fault, alongside dark theme label coloring fixes and absolute path support for -cf. The update also added Linux CPUStack improvements, declarative tests, and UI tweaks like hotkey tooltips and PageUp/Down selection movement.

Built in C++ with plugin support for reverse engineering, x64dbg remains a key tool for malware analysts and CTF competitors on Windows.
The catch: Active development focuses on Windows, with Linux support still experimental and limited to niche debugging scenarios.

Use Cases
  • Reverse engineer Windows binaries without source code
  • Analyze malware behavior in isolated environments
  • Debug executables during CTF exploit development challenges

Source: x64dbg/x64dbg — based on the README and release notes.

Algo VPN simplifies personal cloud VPN setup with WireGuard 🔗

v2.0.1 update fixes Ansible 12 compatibility and cloud provider deployment bugs

trailofbits/algo · Python · 30.3k stars Est. 2016

Trail of Bits’ Algo VPN project released v2.0.

1, a maintenance update addressing Ansible 12 compatibility and fixing deployment issues across AWS, GCE, Vultr, and Scaleway. The tool uses Ansible scripts to automate secure WireGuard and IPsec VPN servers on major cloud providers, generating client configs and QR codes for iOS, macOS, Android, and Windows. It enforces modern cryptography, avoids legacy protocols, and includes optional ad-blocking and SSH tunneling. Despite its age, the project remains actively maintained with recent commits and frequent cloud provider fixes.
The catch: Algo does not claim to provide anonymity or censorship resistance, limiting its use for users seeking protection against state-level surveillance.

Use Cases
  • Developers deploying private VPNs on DigitalOcean for remote work
  • Sysadmins managing IPsec/WireGuard access for iOS and Android teams
  • Privacy-focused users setting up ad-blocking VPNs on Ubuntu servers

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

Quick Hits

awesome-hacker-search-engines Curates specialized search engines for penetration testing, vulnerability assessments, red/blue team ops, and bug bounty to streamline reconnaissance and intelligence gathering. 10.9k
Ciphey Automatically decrypts unknown encryptions, decodes encodings, and cracks hashes without prior knowledge of keys or algorithms, accelerating forensic and offensive security workflows. 21.6k
caddy Delivers a fast, extensible, multi-platform web server with automatic HTTPS, simplifying secure deployment across environments while supporting modern HTTP protocols. 74.3k
bunkerweb Provides an open-source, cloud-native WAF to protect web applications from common exploits with minimal configuration and seamless integration into modern infrastructures. 10.7k
faraday Offers an open-source vulnerability management platform to track, prioritize, and remediate security findings across teams and projects with centralized visibility and collaboration. 6.7k

Bun v1.3.14 sharpens startup speed and adds Windows ARM64 parity 🔗

The all-in-one JavaScript toolkit now matches Linux/macOS performance on Windows devices

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

Bun v1.3.14 delivers measurable startup improvements and closes a long-standing gap for Windows ARM64 users.

The release optimizes the Bun runtime’s initialization path, reducing cold-start latency by up to 30% in benchmarked TypeScript projects. This update ensures consistent behavior across architectures, with bun run, bun test, and bun install now performing identically on Windows on ARM as on x64, Linux, and macOS.
Built in Rust and powered by JavaScriptCore, Bun continues to replace disjointed Node.js toolchains with a single executable. Developers can run JSX/TS files directly via bun run index.tsx, execute tests with bun test, and manage packages using bun install — all without a node_modules forest. The upgrade path remains simple: bun upgrade pulls v1.3.14, while bun upgrade --canary tracks main.
Despite its speed, Bun’s ecosystem maturity lags behind Node.js in niche areas.
The catch: Native addon support remains incomplete, limiting compatibility with some C/C++-based npm packages that rely on node-gyp or prebuilds.

Previously in The Times “covered” — Jul 24

Use Cases
  • Backend devs replace Node.js with Bun for faster TypeScript services
  • Frontend teams use Bun to bundle and test React apps without Webpack
  • CLI tool creators distribute single-binary JavaScript tools via bunx

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

More Stories

Protobuf v35.1 Drops Bazel Proto Fragment Dependency 🔗

Breaking change aligns behavior with 34.x after Bazel 9 tests across seven languages

protocolbuffers/protobuf · C++ · 71.6k stars Est. 2014

Protocol Buffers released v35.1, removing its dependency on Bazel's proto fragment and now respecting only Starlark versions of --proto_toolchain_for*.

This breaking change, which restores 34.x behavior, follows Bazel 9 compatibility tests for C#, HPB, Objective-C, PHP, Python, Rust, and UPB. The update includes C++ cord setters for repeated string fields and a fix to avoid undefined behavior in UPB by switching to the XCT section for constructor ordering. The catch: 319 open issues suggest ongoing instability, and users building from head risk source-incompatible changes without pinning to release commits.

Use Cases
  • Services serialize structured data across language boundaries
  • RPC systems define contracts via .proto files for code generation
  • Mobile apps exchange binary-encoded payloads efficiently

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

Git Source Code Mirror Enables Community Patches via GitGitGadget 🔗

Contributors submit improvements as patches to the mailing list following Documentation/SubmittingPatches

git/git · C · 62.2k stars Est. 2008

The git/git repository serves as the official source code mirror for Git, the distributed revision control system written in C. While pull requests are not accepted directly, contributors can use GitGitGadget to convert PRs into patches for submission to the Git mailing list.

This process ensures all changes adhere to the project's Documentation/SubmittingPatches guidelines and CodingStandards. The project remains actively maintained, with its last commit just 0 days ago and ongoing development visible in its 373 open issues. Developers rely on Git for scalable version control across projects of all sizes, from small scripts to large codebases. The catch: The patch-only contribution model creates a higher barrier to entry compared to direct PR workflows on platforms like GitHub.

Use Cases
  • Developers tracking changes in Git's core C implementation
  • Contributors submitting patches via GitGitGadget for review
  • Teams using Git for distributed version control in collaborative workflows

Source: git/git — based on the project README.

OpenAI Codex CLI Adds Paginated History and Bedrock Support 🔗

Release expands /import now migrates Cursor and Claude Code settings while adding audio inputs and secure terminal visualization links

openai/codex · Rust · 101.3k stars Est. 2025

The latest OpenAI Codex CLI release (rust-v0.145.

  1. adds experimental paginated thread history enabling efficient resume, search, and sub-agent support with persisted memories. Users can now migrate Cursor and Claude Code settings, MCP servers, and project-scoped memories via expanded /import. New audio input handling supports common local formats with streaming realtime V3 conversations, while secure clickable inline visualization links appear directly in the terminal UI. Multi-agent V2 experience is stabilized with configurable sub-agent models and reasoning levels.
    The catch: Despite rapid feature growth, 10,818 open issues suggest stability trade-offs as the CLI attempts to unify local agent, desktop app, and cloud agent workflows under one Rust binary.

Previously in The Times “covered” — Jul 21

Use Cases
  • Developers migrate IDE agent configs to terminal workflow
  • Teams run local audio-assisted coding sessions with memory
  • Engineers debug using clickable terminal visualizations without leaving shell

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

Quick Hits

frp Frp enables secure, fast exposure of local servers behind NAT/firewall to the public internet via reverse proxying. 108.3k
PowerToys PowerToys enhances Windows productivity with customizable utilities like window management, keyboard remapping, and file tools. 136.7k
awesome-go Awesome-Go curates a high-quality collection of Go frameworks, libraries, and tools for efficient development discovery. 179.2k
fzf Fzf provides an interactive, fuzzy-matching command-line finder for fast file, process, and history navigation. 82k

Glasgow Interface Explorer gains new maintainers as founder returns to active development 🔗

Catherine @whitequark resumes work after health recovery, expanding team to accelerate stalled hardware debugging tool progress

GlasgowEmbedded/glasgow · Python · 2.2k stars Est. 2018

The Glasgow Interface Explorer, a Python-based FPGA debugging tool dubbed the "Scots Army Knife for electronics," has seen renewed activity after founder Catherine @whitequark resumed development following her move to the UK and improved health. The project, last pushed on 2026-07-25, is adding maintainers to its current team of three to address a backlog of 74 open issues and accelerate feature development.

Glasgow enables low-level hardware interaction via USB, supporting protocols like JTAG, SPI, I2C, and UART through interchangeable applets.

Its interface explorer provides real-time visualization of bus traffic, making it invaluable for reverse engineering, firmware debugging, and custom peripheral testing. Recent commits focus on stabilizing the gateware toolchain and improving applet compatibility across revisions of the Glasgow hardware.

Despite the momentum, the tool remains niche due to its reliance on specific FPGA hardware and a steep learning curve for newcomers. Documentation assumes familiarity with digital electronics and FPGA workflows, limiting accessibility for software-first developers.

The catch: Glasgow’s full potential requires purchasing the physical Glasgow interface hardware, which faces shipping delays due to mismatched Crowdsupply fulfillment cycles and maintainer capacity, creating a barrier to immediate adoption even for eager users.

Previously in The Times “covered” — Jul 23

Use Cases
  • Hardware hackers reverse-engineering unknown PCBs
  • Firmware developers debugging SPI/I2C sensor communication
  • FPGA engineers validating custom logic analyzer applets

Source: GlasgowEmbedded/glasgow — based on the project README.

More Stories

Frenck’s Home Assistant list adds AI and LLM category after 8 years 🔗

New section reflects growing community interest in generative automation tools

frenck/awesome-home-assistant · Python · 8.3k stars Est. 2018

The frenck/awesome-home-assistant repository updated its contents to include a dedicated “🤖 AI & LLMs” section, signaling mainstream adoption of generative AI in home automation workflows. This addition joins established categories like lighting, climate, and security, curating resources for LLMs that control devices, generate automations, or enhance voice assistants via Home Assistant.

Maintained since 2018, the list remains a key discovery tool for integrations and dashboards installable via HACS. The catch: With 11 open issues and no recent structural updates beyond category additions, the list risks becoming outdated as newer tools emerge faster than curation can keep pace.

Previously in The Times “covered” — Jul 18

Use Cases
  • Find LLM-powered voice control integrations
  • Discover dashboard cards for solar energy tracking
  • Locate community-maintained Bluetooth device trackers

Source: frenck/awesome-home-assistant — based on the project README.

Internet-in-a-Box 8.2 Adds Calibre-Web for Offline eBook Libraries 🔗

Latest release enables video/audio/image support expands knowledge access for schools, clinics, and prisons.

iiab/iiab · Jinja · 1.9k stars Est. 2017

Internet-in-a-Box (IIAB) release 8.2 integrates Calibre-Web as a default feature, letting users manage and access eBook collections on Raspberry Pi or Linux devices without internet.

The update includes experimental support for video, audio, and image files within the library interface, alongside improved diagnostics for tracking Calibre-Web version and system status. Built for offline deployment in underserved areas, IIAB now streamlines setup via a one-line installer or pre-flashed disk images. The project maintains active development with commits as recent as one day ago, though 355 open issues suggest ongoing challenges in scaling and usability.
The catch: Reliance on volunteer maintainers may delay responses to critical bugs in high-stakes deployments like medical or prison environments.

Previously in The Times “covered” — Jul 19

Use Cases
  • Schools deploy offline Wikipedia and Khan Academy
  • Clinics provide medical references without internet
  • Prisons enable educational content access securely

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

LibreHardwareMonitor adds Arctic fan controller support in latest update 🔗

Release v0.9.6 includes new hardware monitoring for specific motherboards and disk tools

LibreHardwareMonitor/LibreHardwareMonitor · C# · 8.7k stars Est. 2017

The LibreHardwareMonitor project released v0.9.

6, adding support for Arctic fan controllers and GA-A320M motherboards via recent pull requests. Developers Rem0o and Blacktempel contributed hardware-specific fixes, while dependency updates improved system integration. The library now enables detailed monitoring of CPUs, GPUs, storage, and network interfaces across Windows systems. Nightly builds remain available for testing cutting-edge features. The catch: Open issues exceed 500, indicating ongoing challenges in hardware compatibility and maintenance despite active contributions.

Previously in The Times “covered” — Jul 21

Use Cases
  • Developers monitoring custom PC thermal performance
  • System administrators tracking server hardware health
  • Hobbyists validating cooling efficiency after modifications

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

Quick Hits

openflight jewbetcha/openflight (Python): A lightweight Python library for parsing, generating, and manipulating OpenFlight (.flt) 3D scene files, ideal for aerospace and simulation workflows. 723
firmware OpenIPC/firmware (C): Community-driven open-source firmware for IP cameras, enabling full control over RTSP, ONVIF, and motion detection without vendor lock-in. 2.1k
firmware BruceDevices/firmware (C++): A malicious ESP32 firmware designed to exploit device vulnerabilities — avoid; included for awareness only. 6.2k
CppDeveloperRoadmap salmer/CppDeveloperRoadmap (HTML): An interactive, visual guide mapping essential C++ concepts from fundamentals to advanced topics for self-paced learning. 3.5k
TuyaOpen tuya/TuyaOpen (C): A modular, high-performance AI+IoT framework accelerating device integration across ESP32, T2/T3/T5AI chips with built-in agent support and low-latency connectivity. 1.8k

GDevelop Restores Cloud Projects and Adds CLI Support in Latest Release 🔗

Version 5.6.274 improves recovery options and command-line access for developers

4ian/GDevelop · JavaScript · 25.1k stars Est. 2014 · Latest: v5.6.274

GDevelop’s latest release, v5.6.274, introduces practical updates for game creators working in the cloud and terminal environments.

Users can now restore recently deleted cloud projects, with free accounts able to recover deletions within a short window and subscribers gaining up to 30 days of recovery time. The engine also adds command-line interface (CLI) support, automatically installed on Windows and accessible via the command palette on Linux and macOS, enabling headless builds and automation.

Additional improvements include saved scroll positions in object group and layer panels, clearer iOS build error reporting—such as incorrect icon sizes—and a suite of bug fixes addressing terminal hangs on macOS, mixed-value displays in object groups, and Spine attachment rotations. Experimental extensions now allow 2D objects to align with 3D camera positions, and two inventory-related extensions have migrated to the new variable parameter system.

These changes reflect a focus on stability and workflow efficiency, particularly for teams using cloud storage or scripting pipelines. The project continues to support 2D, 3D, and multiplayer game exports across mobile, desktop, and web platforms through its event-based system and modular behaviors.

The catch: Despite ongoing updates, the engine’s reliance on JavaScript and its event-driven model may limit performance-critical or code-heavy projects compared to lower-level engines.

Previously in The Times “covered” — Jul 20

Use Cases
  • Indie developers building 2D mobile games with no-code logic
  • Educators teaching game design using visual event systems
  • Teams automating builds via CLI for CI/CD pipelines

Source: 4ian/GDevelop — based on the README and release notes.

More Stories

Boardgame.io simplifies turn-based game dev with automatic networking 🔗

TypeScript engine syncs state across clients, servers, and platforms without manual netcode

boardgameio/boardgame.io · TypeScript · 12.4k stars Est. 2017

Boardgame.io lets developers write turn-based game logic in pure functions describing state changes per move.

The engine handles real-time multiplayer synchronization, AI bot generation, lobby matchmaking, and game logging with time-travel capabilities. It supports React, React Native, and vanilla JS clients through view-layer agnostic bindings. Recent activity shows steady maintenance, with the last commit just one day ago and ongoing issue triage. The catch: Its abstraction may limit fine-grained control over networking or server architecture for developers needing custom low-latency optimizations.

Use Cases
  • Indie dev creating a React-based strategy game
  • Team prototyping a cross-platform tabletop RPG
  • Educator building turn-based learning simulations

Source: boardgameio/boardgame.io — based on the project README.

Godot AI v3.0.6 Fixes Server Leaks and UTF-8 Encoding Bugs 🔗

Latest patch resolves error persistence and WebSocket send issues in MCP tool reliability

hi-godot/godot-ai · GDScript · 1.2k stars 3mo old

The v3.0.

6 release of hi-godot/godot-ai patches two stability flaws: project_run now stops leaking stale errors from past runs into launch responses, and connection.gd avoids double UTF-8 encoding on every WebSocket send. These fixes address user-reported issues where MCP tool interactions degraded over time or failed with special characters. The update also adds opt-in keep_server_on_exit to retain the Python server after editor shutdown. Built in GDScript, it requires Godot 4.5+ and uv for the MCP server backend.
The catch: Reliance on uv and Python adds setup friction for Godot developers preferring pure-GDScript workflows.

Previously in The Times “covered” — Jul 19

Use Cases
  • Game designers iterate scenes via Claude Code prompts
  • Programmers refactor nodes using AI-driven signal wiring
  • Artists adjust materials and animations through conversational edits

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

Quick Hits

luanti Luanti (formerly Minetest) empowers builders to create and mod voxel-based games with intuitive tools and open-source flexibility. 13.3k
ShaderToHuman ShaderToHuman simplifies shader debugging by visualizing HLSL/GLSL code behavior in real time for faster iteration. 601
GodSVG GodSVG enables precise, structured SVG editing across desktop and web platforms, ideal for developers needing pixel-perfect vector workflows. 2.6k
The-Grand-Combo The-Grand-Combo transforms Victoria II with a reimagined map, enriched historical depth, and innovative mechanics for immersive grand strategy. 165
gecs GECS provides a lightweight, Godot-native Entity Component System to streamline game architecture and boost performance in GDScript projects. 568
ebiten Ebiten offers a minimalist, high-performance 2D game engine in Go, letting builders focus on gameplay without engine complexity. 13.4k
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