Design
Preset
Background
Text
Font
Size
Width
Account Pricing

The Git Times

AI Models
Claude Opus 5 $25/M GPT-5.6 Luna $0.60/M Gemini 3.1 Pro Preview $12/M Grok 4.5 $6/M DeepSeek V4 Pro $1.26/M Qwen3.8 Max $6/M Kimi K3 $15/M
Full Markets →

LinearAbiltyCastingThreeJS Turns Skillshot VFX Into a Playable Three.js Sandbox 🔗

Developers explore hand-written GLSL and Vite to prototype League of Legends-style ability effects with line and ground-targeted casts.

achrefelouafi/LinearAbiltyCastingThreeJS · JavaScript · ▲ 122 in 1d 5d old

achrefelouafi/LinearAbiltyCastingThreeJS is a skillshot VFX sandbox built with Three.js, Vite, and custom GLSL shaders that lets developers prototype and tweak ability effects in real time.

The project implements five distinct abilities — Frost Lance (Q), Storm Lance (E), Cinder Fall (R), Nova Beam (F), and Voltaic Snare (V) — each with unique visual behaviors: line casts that arm on keypress, follow mouse aim, and fire on click; and a far cast (V) that uses a thick-boundary cursor circle to preview area of effect before commitment. Effects include fracturing ice fronts, branching lightning, raymarched lava trails, spiraling light columns, and electric leashes that scorch the terrain.

Built with Vite for fast iteration and Three.js for rendering, the sandbox emphasizes low-level control through hand-written GLSL, avoiding abstraction layers to give developers direct access to shader logic for customization. The immediate visual feedback loop makes it ideal for learning advanced rendering techniques or prototyping game abilities without engine overhead.

The catch: The project lacks documentation, configuration options, or export mechanisms, limiting its use beyond local experimentation and raising questions about scalability or integration into larger codebases.

Use Cases
  • Game developers prototyping ability VFX
  • Graphics programmers learning GLSL and Three.js
  • Educators teaching real-time shader effects
  • Indie creators testing skillshot mechanics
  • Open-source contributors expanding effect library

Source: achrefelouafi/LinearAbiltyCastingThreeJS — based on the project README.

More on the Front Page

ARM64 Android port brings PlayStation 3 emulation to mobile devices 🔗

Leverages RPCS3 upstream with Vulkan fixes and dynamic thread scheduling for Adreno GPUs

ARMSX2/ARMSX3 · C++ · 243 stars 5d old

ARMSX3 ports the RPCS3 PlayStation 3 emulator to Android ARM64 devices, enabling PS3 gaming on smartphones and tablets. Built from the latest RPCS3 code, it includes recent ARM64 improvements and requires the Android NDK r27+, CMake 3.

30+, and JDK 17 to compile. The core library, built via CMake and Ninja, outputs an unstripped 1.3 GB binary that must be stripped and manually copied into the app’s JNI libs directory before assembling the release APK with Gradle.

Release 0.5 delivers significant performance gains: Spider-Man: Web of Shadows now runs at ~37 ms per frame on Snapdragon 8 Gen 2, down from 120 ms, after fixing a Vulkan queue drain that previously consumed 14.6 ms per frame. Occlusion query hangs are now resolved by auto-abandoning stale queries after a few seconds. The thread scheduler no longer pins SPU threads to fixed cores, instead delegating core assignment to Android’s scheduler for better big.LITTLE utilization. Sonic Unleashed and God of War III demo now boot successfully.

The catch: The project demands manual build steps — including cloning submodules and two external repos, stripping binaries, and copying libraries — and provides no prebuilt APKs, limiting accessibility for users unwilling to compile from source.

Use Cases
  • Developers test PS3 game performance on ARM64 Android
  • Enthusiasts play PS3 titles on compatible smartphones
  • Contributors improve emulator efficiency on mobile GPUs

Source: ARMSX2/ARMSX3 — based on the README and release notes.

PPT Master adds 26 new brand template workspaces for native PowerPoint generation 🔗

Release v4.5.0 expands style options with composition grammar and visualization catalogs

hugohe3/ppt-master · Python · ▲ 45k in 1d 8mo old

Hugo He’s PPT Master now includes 26 new template workspaces—fifteen brand identities like McKinsey, NVIDIA, and Huawei, plus eleven style variants—enabling AI to generate native .pptx files with accurate shapes, transitions, and data-backed charts.

The engine uses Kimi K3 to parse PDFs, DOCX, or web content, structure narratives, and apply composition grammar instead of preset shapes. Users can inject custom templates and export narrated decks from speaker notes.
The catch: Eight open issues remain, including edge cases in complex animation rendering and limited support for non-Latin fonts in auto-generated text boxes.

Use Cases
  • Consultants generate firm-branded decks from whitepapers
  • Educators turn lecture notes into animated student presentations
  • Executives auto-create earnings summaries with live-chart visualizations

Source: hugohe3/ppt-master — based on the README and release notes.

Claude Code plugin rewrites AI responses in plain English locally 🔗

Uses Ollama to display simplified text without altering Claude's original output

gvzdv/claudish-to-english · Shell · 467 stars 1d old

The gvzdv/claudish-to-english plugin hooks into Claude Code to show a plain-English rewrite of each assistant message, powered by a local LLM via Ollama. It operates display-only: Claude's reasoning and saved transcripts retain the original text, while only what appears on screen is transformed.

An optional second hook can rewrite Markdown files into plain English on write or edit, opt-in and off by default. Every hook fails open—if Ollama is down, times out, or the model isn't ready, users see Claude's original text with a one-time notice explaining the skip. The plugin shells out to a local model and requires pre-warming; the default is gemma4:26b-mlx, but users can set CLAUDISH_MODEL to any pulled Ollama tag. After Anthropic review, it will be available in the community marketplace.
The catch: As a working prototype, it depends on a local Ollama setup and may skip rewrites silently if the model isn't loaded or responsive, requiring manual warm-up and configuration.

Use Cases
  • Developers simplify complex AI explanations for faster comprehension
  • Teams convert technical Markdown docs into plain language for stakeholders
  • Non-experts access Claude's reasoning without jargon overload

Source: gvzdv/claudish-to-english — based on the README and release notes.

Open-source YouTube Music client delivers ad-free streaming without subscription 🔗

Windows desktop app enables background playback and local playlists via reverse-engineered API

sohaibdevv/youtube-music · TypeScript · 837 stars 0d old

sohaibdevv/youtube-music is a TypeScript-based Windows application that streams YouTube Music audio without ads, login, or tracking. It supports search, background playback, media keys, and custom playlists saved locally.

The app runs from a password-protected ZIP (ytm4all) and requires no installation or registry changes. Built on the reverse-engineered YouTube Music API, it streams audio like a browser — no downloading or DRM bypass. Licensed MIT, it’s a lightweight alternative for users seeking free, uninterrupted music playback. The catch: As a recent v1.0.0 release with zero forks and no issue history, long-term stability and API resilience remain untested.

Use Cases
  • Developers testing audio streaming apps on Windows
  • Users wanting ad-free music without YouTube Premium
  • Builders needing background playback during coding sessions

Source: sohaibdevv/youtube-music — based on the README and release notes.

WeKnora Adds Revision History to Self-Maintaining Wiki Mode 🔗

Agents now track markdown edits, diffs, and rollbacks in enterprise knowledge bases

Tencent/WeKnora · Go · 19.7k stars Est. 2025

Tencent’s WeKnora project enhanced its autonomous Wiki Mode with page-level revision history, allowing agents to diff, revert, and manually edit generated markdown pages. The update supports tracking changes in self-maintaining knowledge bases built from raw documents via Feishu, Notion, or Yuque sync.

Users can now review and rollback edits directly in the interface, improving trust in AI-curated content. The feature arrives alongside fixes for vector store cleanup and Feishu DOCX block-level ingestion. The catch: Open issues exceed 500, raising questions about project maintainability despite active development.

Use Cases
  • Engineers syncing Notion docs for AI-powered internal Q&A
  • Teams building self-updating technical wikis from PDF and Excel sources
  • Enterprises deploying multi-tenant RAG agents with scoped API access

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

Astryx Design System Lets AI and Humans Share UI Tooling 🔗

Built on React 19+ and StyleX, it exports components for direct composition without build plugins

facebook/astryx · TypeScript · 11.9k stars 7mo old

Astryx, Meta’s internal design system now open source, delivers 150+ accessible components with brand-level theming, dark mode, and a CLI. It uses StyleX for styling but lets teams override with Tailwind, CSS modules, or plain CSS—no wrapping required.

Components are exportable at any level, and swizzle eject exposes full source for deep customization. The system is designed so people and AI assistants use identical tooling, APIs, and docs.
The catch: Requires React 19 or later, creating a hard upgrade path for teams on older React versions.

Use Cases
  • Design teams building branded UIs without forking component source
  • AI agents generating consistent UI using typed React components
  • Engineers customizing themes via CSS property overrides only

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

Telegram Bot Lets Traders Copy Polymarket Whale Moves Instantly 🔗

Free public alerts channel delivers real-time whale trade signals with one-tap execution

mothparkzo6249/TG-Polymarket-bot · JavaScript · 1k stars 2w old

The mothparkzo6249/TG-Polymarket-bot is a JavaScript Telegram bot that monitors Polymarket’s live trade feed for large whale transactions and enables users to replicate them with a single tap. By embedding wallet signing directly into the bot, it bypasses Polymarket’s UI delays and confirmation popups, ensuring instant, glitch-free bet execution.

Users can set custom alert thresholds for trade size and probability, and manage funds via the bot or Polymarket website using the same balance. The bot encrypts keys behind a user-defined PIN and offers a free public alerts channel for broader market insight.
The catch: Access to the bot’s full features expires on October 10, 2026, after which functionality may be restricted or require payment.

Use Cases
  • Retail traders copy whale Polymarket bets in real time
  • Analysts receive automated alerts on large market movements
  • Users execute trades without navigating Polymarket’s confirmation UI

Source: mothparkzo6249/TG-Polymarket-bot — based on the project README.

Open Source Shifts Toward Self-Improving AI Agent Ecosystems 🔗

Modular skills, persistent memory, and agent-native tooling redefine autonomous workflows

Trendai-agents
SaladDay/pi-from-scratchayi-ai/nie-grassroots-logicfacebook/astryxvm0-ai/vm0hugohe3/ppt-masterTencent/WeKnoracortexkit/magic-contextPrimeIntellect-ai/prime-agentcloudflare/cloudflare-osTencentCloud/TencentDB-Agent-Memorygoogle/skillsifixai-ai/iFixAiayghri/i-have-adhdbojieli/ai-agent-bookcorsairdev/corsairtrycompai/crmtt-a1i/archifycan1357/oh-my-pilangchain-ai/openwikiesengine/DeepSeek-Reasonixopenchamber/openchamberrocketride-org/rocketride-serverget-bb/bbhuangruiteng/loopxalibaba/open-code-reviewagentplugins/agent-plugins-specalirezarezvani/claude-skillsyc-software/qmblader/humanizerHKUDS/Vibe-Tradinglemony-ai/cascadeflowcalesthio/OpenMontagePan-Chera/Multi-Agent-CADmagnitudedev/magnitudevectorize-io/hindsightNVIDIA/cosmos

Open source is converging on AI agents as composable, self-improving systems rather than monolithic models. Projects like PrimeIntellect-ai/prime-agent demonstrate recursive self-enhancement in coding workflows, while Tencent/WeKnora and TencentCloud/TencentDB-Agent-Memory show how raw documents evolve into governed, reusable agent memories—chat logs, skills, wikis, and code graphs—shared across frameworks.

Memory innovation is central: cortexkit/magic-context enables unbounded, self-managing context for long-session agents, and vectorize-io/hindsight introduces learning memory that adapts over time.

Tooling is rapidly agent-first. facebook/astryx provides a fully customizable, agent-ready design system, and corsairdev/corsair offers an integration layer for agent extensions, formalized by agentplugins/agent-plugins-spec. Skills marketplaces are emerging: alirezarezvani/claude-skills bundles hundreds of agent-ready abilities across coding agents, and google/skills delivers product-specific agent capabilities. Use cases span domains: hugohe3/ppt-master generates native PowerPoint from documents, HKUDS/Vibe-Trading acts as a personal trading agent, and calesthio/OpenMontage turns agents into video studios with 100+ tools.

The catch: Despite rapid innovation, the ecosystem remains fragmented—competing memory formats, skill interfaces, and agent loops hinder interoperability, and many self-improvement claims lack rigorous validation in noisy, real-world tasks, leaving durability and trust unproven at scale.

Use Cases
  • Developers build self-improving coding agents
  • Teams deploy shared agent memory systems
  • Enterprises create agent-native CRM and design tools

Open Source Unifies LLM Tooling Around Standardized Agent Interfaces 🔗

Projects converge on common protocols to let AI agents work across frameworks and models seamlessly

Trendllm-tools
SaladDay/pi-from-scratchshadcn-ui/chatbot-template0xnyn/airshipvm0-ai/vm0mozilla-ai/any-llmTencent/WeKnoraChevey339/kelivoNVIDIA-NeMo/Gymconfident-ai/deepevaldiegosouzapw/OmniRoutezhaoxuya520/reverse-skillvirgiliojr94/book-to-skillcathrynlavery/diagram-designTencentCloud/TencentDB-Agent-MemoryZhuLinsen/daily_stock_analysisMadsLorentzen/ai-job-searchVincentwei1021/video-shotcraftrocketride-org/rocketride-serverhuangruiteng/loopxalibaba/open-code-reviewalirezarezvani/claude-skillsdifferent-ai/openworkbradautomates/claude-videocoreyhaines31/marketingskillsAlishahryar1/free-claude-codetashfeenahmed/freellmapiAgriciDaniel/claude-seothebuggeddev/anatomyzhamao-robot/zhamao-framework

A clear pattern is emerging in open source LLM tooling: the rise of standardized interfaces that let AI agents operate independently of specific frameworks or providers. Rather than building monolithic, locked-in assistants, developers are creating modular skills, adapters, and gateways that plug into any compatible agent runtime.

This shift is visible in projects like OmniRoute, which offers a single endpoint to access 290+ LLM providers with automatic fallback and token compression, and mozilla-ai/any-llm, which provides a unified Python interface to communicate with any LLM backend. Agent skill ecosystems are also coalescing—alirezarezvani/claude-skills offers hundreds of portable skills for Claude Code, Codex, and other agents, while virgiliojr94/book-to-skill turns technical PDFs into usable agent skills on demand. Workflow orchestration is following suit: rocketride-org/rocketride-server provides an IDE-integrated pipeline engine with support for dozens of models and vector databases, and huangruiteng/loopx enables durable, quota-aware agent loops across Claude Code, Codex, and similar tools. Even domain-specific tools like TencentCloud/TencentDB-Agent-Memory are designing shared memory layers—chat memory, skill libraries, code graphs—that can be governed and reused across agent teams. The underlying movement is toward interoperability: agents as composable units, not siloed products.
The catch: While promising, this fragmentation of interfaces risks creating new silos—each project often defines its own "standard," and true plug-and-play compatibility remains limited outside tightly coupled ecosystems like Claude Code. Many tools are still experimental, lack long-term maintenance plans, or depend on volatile AI provider APIs, making production adoption uncertain despite rapid innovation.

Use Cases
  • Developers build portable AI skills usable across Claude Code, Codex, and Cursor
  • Teams deploy unified LLM gateways to access 290+ providers with automatic failover
  • Engineers orchestrate multi-model AI workflows from IDEs with durable agent state management

Open Source Data Infrastructure Shifts Toward Modular, Multimodal Pipelines 🔗

Projects now combine AI, streaming, and domain-specific tooling for real-time, composable data workflows

Trenddata-infra
sohaibdevv/youtube-musicFlaminis/Dalaranhugohe3/ppt-masterapache/camelgraphframes/graphframeselastic/kibanaytsaurus/ytsauruslinuxfoundation/crowd.devbrightdata/clitalivia-group/taliviaZhuLinsen/daily_stock_analysistt-a1i/archifymalisper/pgrustcoreyhaines31/marketingskillsmiuuyy/codex-chatgpt-webhpcaitech/ColossalAIh2oai/h2o-3FFmpeg/FFmpeg

A clear pattern is emerging in open source data infrastructure: a move toward modular, multimodal systems that tightly integrate AI, real-time processing, and domain-specific interfaces. Rather than monolithic stacks, new projects emphasize composability and specialized function.

sohaibdevv/youtube-music demonstrates lightweight, API-driven clients for niche data sources, bypassing traditional auth and subscription models. Flaminis/Dalaran extends this with Rust-based, ROS 2-native visualization for multimodal time-series, enabling robotic and sensor data exploration directly from .rrd files. Meanwhile, hugohe3/ppt-master shows how AI can transform raw documents or topics into native, animated PowerPoint decks with data-backed charts — turning unstructured input into formatted output via LLMs. On the integration side, apache/camel continues to evolve as a connective tissue, offering 350+ connectors to route data across systems using Java, YAML, or XML. For scalable backend needs, ytsaurus/ytsaurus provides a fault-tolerant, C++-built big data platform optimized for throughput and consistency. Together, these repos signal a shift: infrastructure is no longer just about storage or compute, but about purpose-built pipelines that ingest, transform, visualize, and act on data in context-specific ways — often AI-augmented and language-agnostic.
The catch: Much of this remains early-stage, with fragmented tooling, inconsistent APIs, and limited interoperability between niches — raising concerns about long-term maintainability and vendor lock-in at the module level, even as individual components show promise.

Use Cases
  • Engineers build real-time robotic dashboards from sensor logs
  • Analysts convert market reports into animated presentations
  • Developers connect SaaS APIs using low-code integration routes

Deep Cuts

GDScript Town Simulates AI Agents in a Living Mini-World 🔗

Builders prototype emergent behaviors without complex ML infrastructure

mewamew/my_ai_town · GDScript · ▲ 59 in 1d

mewamew/my_ai_town lets developers spawn AI agents in a Godot-powered pixel town using GDScript. Agents navigate, interact, and adapt through simple rule-based systems, revealing surprising social dynamics.

It’s a sandbox for testing AI sociology, dialogue loops, or pathfinding quirks without training models. Watch barters form, conflicts arise, or traditions emerge from basic triggers. Ideal for rapid iteration where Unity or Python stacks would overcomplicate early exploration. The catch: Its niche Godot focus and minimal documentation leave most builders unaware of its prototyping power.

Use Cases
  • Game designers testing NPC social systems
  • Researchers observing emergent agent behavior
  • Educators teaching AI concepts visually

Source: mewamew/my_ai_town — based on the project README.

Quick Hits

airship 0xnyn/airship: A TypeScript-powered visual editor inspired by Figma, designed to streamline UI creation within Claude Code, Codex, and OpenCode environments. 259
chatbot-template shadcn-ui/chatbot-template: A lean, production-ready chatbot starter using Next.js, Vercel AI Gateway, and shadcn/ui for fast, customizable AI interactions. 490
llvm-project llvm/llvm-project: A comprehensive, modular compiler infrastructure enabling optimized code generation, analysis, and toolchain development across languages and platforms. 39.8k
Who shipped it

The Business Desks

Fresh on Hugging Face

Model Drops

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

The AI Wire

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

OpenClaw 2026.7.1-2 fixes npm plugin updates for official tool compatibility 🔗

Patch resolves singleton-array metadata issue blocking plugin installs from newer npm clients

openclaw/openclaw · TypeScript · ▲ 156 in 1d 8mo old · Latest: v2026.7.1-2

OpenClaw’s latest release v2026.7.

1-2 addresses a critical workflow blocker: npm plugin updates now accept singleton-array metadata from newer npm clients, ensuring official plugins install and update to correction releases. The fix, pulled in #108336, unblocks a core feature—extending the assistant via tools, skills, and plugins—after users reported failures when trying to add official extensions. OpenClaw remains a TypeScript-based personal AI gateway running locally on macOS, Linux, and Windows, connecting models to WhatsApp, Telegram, Slack, and other channels through a unified Control UI. It provisions Node.js runtimes (22.22.3+, 24.15+, or 25.9+) when needed and operates as a pnpm workspace, rejecting plain npm installs at the repo root. With 81,127 forks and 5,609 open issues, the project shows active adoption but ongoing maintenance pressure. The assistant treats inbound messages as untrusted input and requires manual approval for pairing unknown senders on DM-capable channels—a deliberate security stance.
The catch: Despite its cross-channel reach, OpenClaw demands significant devops overhead for secure remote exposure, with sandboxing and Gateway hardening left to the operator via separate guides, making casual multi-user deployment risky without deep expertise.

Previously in The Times “covered” — Aug 10

Use Cases
  • Developers extending AI via custom tools on local devices
  • Teams testing private assistants in Slack or Discord workflows
  • Individuals self-hosting voice-controlled automation with companion apps

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

More Stories

Dify adds workflow node locator and tool multi-select for faster AI app debugging 🔗

Export agent configurations as YAML and improved block selector streamline team collaboration on LLM workflows

langgenius/dify · TypeScript · ▲ 119 in 1d Est. 2023

The latest Dify release introduces a workflow node locator that lets developers click a node ID in logs to instantly highlight it in the editor, cutting debug time. Workflow tool nodes now support multi-select dropdown inputs, enabling flexible parameter configuration without custom code.

Agent apps can be exported as DSL YAML directly from the sidebar for version control or backup. Block selector improvements add keyboard navigation and category previews for faster node discovery. These updates target friction in collaborative AI workflow design, especially for teams iterating on RAG pipelines and agent logic.
The catch: Despite active development, over 1,000 open issues suggest ongoing stability challenges in complex multi-agent orchestration at scale.

Previously in The Times “covered” — Aug 5

Use Cases
  • AI engineers debug LLM workflows using node-level log tracing
  • Teams version-control agent configurations via exported YAML DSL
  • Low-code builders configure multi-tool agents without writing glue code

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

NVIDIA Releases Cosmos 3 for Physical AI World Modeling 🔗

Omnimodal foundation model unifies language, video, audio, and action inputs for robotics and simulation

NVIDIA/cosmos · Jupyter Notebook · 11.5k stars Est. 2024

NVIDIA's Cosmos 3 release delivers an open omnimodal world model family designed for Physical AI applications. Built on a Mixture-of-Transformers architecture, it processes and generates text, images, video, audio, and action sequences within a unified framework.

The release includes Cosmos3-Nano (16B) and supports flexible input-output configurations for world understanding and generation tasks. Developers can use it for robot action prediction, video-to-world simulation, and multimodal reasoning with integrations like vLLM-Omni, TensorRT-LLM, and NIM. Benchmarks show state-of-the-art open performance across understanding, generation, and robotics tasks. The catch: High computational demands and dependency on NVIDIA-specific tooling may limit accessibility for teams without GPU infrastructure or those seeking vendor-neutral stacks.

Use Cases
  • Robotics teams train policy models using action-conditioned rollouts
  • Autonomous vehicle developers simulate edge-case driving scenarios
  • Smart infrastructure planners generate sensor-fused environmental forecasts

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

Microsoft’s ML-For-Beginners adds GitHub Copilot tips for data science learners 🔗

New Discord series offers hands-on guidance for using AI pair programming in Scikit-learn workflows

microsoft/ML-For-Beginners · Jupyter Notebook · 89.3k stars Est. 2021

The microsoft/ML-For-Beginners project now promotes a Learn with AI Series running September 18–30, 2025, where participants receive tips on using GitHub Copilot for data science tasks within the curriculum’s 26 lessons. The course teaches classic machine learning using Scikit-learn and Python, avoiding deep learning, and includes quizzes, assignments, and project-based exercises.

It remains available in Jupyter Notebook format with no recent code changes beyond community updates.
The catch: The curriculum avoids deep learning and modern ML frameworks, limiting its relevance for roles requiring TensorFlow, PyTorch, or large model experience.

Previously in The Times “covered” — Aug 6

Use Cases
  • Students learning Scikit-learn basics through guided notebooks
  • Educators adopting structured ML coursework for classrooms
  • Beginners pairing ML study with GitHub Copilot experimentation

Source: microsoft/ML-For-Beginners — based on the project README.

Quick Hits

AutoGPT AutoGPT empowers anyone to build and deploy autonomous AI agents without deep expertise, focusing on real-world task automation through accessible, modular tools. 186.5k
hermes-agent Hermes-Agent evolves alongside your workflow, learning from interactions to become a personalized, adaptive AI partner that grows smarter with use. 229.2k
ColossalAI ColossalAI slashes the cost and complexity of training massive AI models by optimizing parallelism and memory efficiency, making cutting-edge AI feasible for more teams. 41.4k
system-prompts-and-models-of-ai-tools This repo aggregates open-source system prompts and internal AI tool configurations from leading AI coding assistants, offering a treasure trove for prompt engineers and tool builders. 142.8k
spec-kit Spec-Kit jumpstarts Spec-Driven Development with ready-to-use templates and tooling, helping teams define clear, testable requirements before writing a single line of code. 126.3k

Rerun 0.36.0 Adds Experimental 3D Gaussian Splatting for Robotics Data 🔗

New release enables web-based viewing of large multimodal datasets beyond RAM limits

rerun-io/rerun · Rust · ▲ 5 in 1d Est. 2022 · Latest: 0.36.0

Rerun’s latest release introduces experimental support for 3D Gaussian splatting, a rendering technique that enhances visualization of complex point cloud and sensor data in robotics workflows. The update also expands the Viewer catalog to handle larger-than-RAM files directly in the browser, enabling remote teams to inspect multimodal logs — including images, video, joint states, and transforms — without local downloads.

Multi-sink support for GrpcServerSink improves scalability for distributed logging pipelines. SDKs in Python, Rust, and C++ remain central, with the Python SDK bundling the Viewer for quick starts via pip install rerun-sdk. Breaking changes include stricter application ID naming (max 180 characters, ASCII alphanumerics and select symbols) and relocated ParquetReader options in Python. The project continues to ingest data from MCAP, rrd, and LeRobot formats, syncing multimodal streams for real-time inspection and direct streaming into training jobs.
The catch: Experimental features like Gaussian splatting and web catalog lack stability guarantees, and the nasm-dependent video decode optimization in Rust requires manual setup, potentially limiting adoption in performance-sensitive deployments.

Previously in The Times “covered” — Aug 11

Use Cases
  • Robotics teams sync multimodal sensor logs for debugging
  • ML engineers stream live data directly into training pipelines
  • Remote collaborators inspect large datasets via browser-based viewer

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

More Stories

RT Corporation updates CRANE-X7 ROS 2 packages for Jazzy with new MoveIt examples 🔗

Release 5.0.0 adds RealSense namespace support and mock components for simulation

rt-net/crane_x7_ros · Python · 127 stars Est. 2018

The rt-net/crane_x7_ros project released version 5.0.

0, its first ROS 2 Jazzy Jalisco update, adding RealSense camera_namespace support and mock components for testing. It includes new moveit_py example packages alongside existing C++ and Python samples for Gazebo simulation and MoveIt 2 configuration. Maintained since 2018, the repository supports Ubuntu 24.04 and provides USB setup guidance for the CRANE-X7 arm. Despite recent activity, the project notes development is not open, with feature requests subject to internal guidelines.
The catch: While actively maintained, the project lacks free community support and prioritizes internal roadmap over external feature requests.

Use Cases
  • Control CRANE-X7 robotic arm via ROS 2 on Ubuntu
  • Simulate crane_x7 movements in Gazebo with MoveIt 2
  • Learn robotics using C++ or Python example codebases

Source: rt-net/crane_x7_ros — based on the README and release notes.

Google DeepMind Enhances MuJoCo with Surface Velocity Modeling for Conveyor Systems 🔗

Release 3.11.0 adds geom/surfacevel to simulate treadmills and turntables without dynamic joints

google-deepmind/mujoco · C++ · ▲ 16 in 1d Est. 2021

The latest MuJoCo release introduces geom/surfacevel, enabling accurate simulation of moving surfaces like conveyor belts and turntables using static geometries. This update allows friction-driven motion of contacting bodies without adding degrees of freedom, improving efficiency in industrial robotics and automation workflows.

Surface velocity fields compose naturally with body motion and are visualized via contact-point arrows in the native GUI. The feature supports complex contact dynamics in manufacturing and logistics simulations where belt-driven part handling is common.
The catch: Despite performance gains, the C++-centric API and limited native language bindings beyond Python may steepen the learning curve for teams relying on non-Python stacks or rapid prototyping in other environments.

Previously in The Times “covered” — Aug 6

Use Cases
  • Robotics engineers simulating assembly line part transport
  • Biomechanics researchers modeling joint movement on treadmills
  • Game developers creating physics-based interactive environments with moving platforms

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

Quick Hits

glim GLIM provides a flexible, point cloud-driven framework for robust 3D localization and mapping, enabling precise spatial awareness in dynamic environments. 1.7k
openarm OpenArm delivers a fully open-source humanoid arm designed for physical AI research, excelling in contact-rich tasks requiring dexterity and real-time adaptation. 2.8k
ardupilot ArduPilot offers a unified, battle-tested open-source autopilot suite for planes, copters, rovers, and subs, supporting diverse autonomous vehicle development. 15.7k
panda-py Panda-py enables real-time, low-latency Python control of Franka Emika robots, bridging high-level AI with precise robotic actuation. 177
rtabmap RTAB-Map delivers a scalable, loop-closure-aware SLAM system for real-time 3D mapping and navigation from RGB-D, stereo, or LiDAR sensors. 3.9k
navigation2 Navigation2 provides a modular, production-ready ROS 2 framework for robot navigation, featuring plug-and-play planners, controllers, and recovery behaviors. 4.6k

HackTricks Adds Japanese Search Index to Boost Global Pentest Access 🔗

Automated ja release improves non-English usability for security researchers worldwide

HackTricks-wiki/hacktricks · CSS · ▲ 14 in 1d Est. 2020 · Latest: searchindex-ja

The HackTricks wiki released searchindex-ja, an automated Japanese language index built to improve accessibility of its pentesting techniques for Japanese-speaking security researchers. This update parses the project’s extensive CTF, real-world app, and research-derived tricks into a localized search structure, reducing language barriers in time-sensitive engagements.

Built with CSS and maintained by STM Cyber, the project now serves localized content via http://localhost:3337 when deployed locally, supporting teams using its methodology alongside tools from Intigriti, Modern Security, and 8kSec Academy. The release follows a pattern of incremental i18n efforts, though no corresponding searchindex-en rebuild was noted in recent commits.
The catch: Despite frequent pushes, the project’s CSS-only stack limits dynamic interactivity, and open issues remain unaddressed for over 300 days on average, raising questions about active maintenance depth beyond documentation updates.

Previously in The Times “covered” — Aug 7

Use Cases
  • Pentesters adapting CTF write-ups for Japanese teams
  • Security trainers localizing offensive security curricula
  • Bug bounty hunters cross-referencing techniques in ja environments

Source: HackTricks-wiki/hacktricks — based on the README and release notes.

More Stories

TruffleHog Adds Vault Token Detection and Tightens Jira Verification 🔗

Latest release improves secret classification and validation across cloud and DevOps platforms

trufflesecurity/trufflehog · Go · ▲ 33 in 1d Est. 2016

TruffleHog v3.96.

0 adds detection for HashiCorp Vault tokens and refines Jira token verification to reduce false positives. The tool now logs analyze errors for Postgres and Anthropic analyzers, improving debuggability. It also introduces TargetNotFoundError for missing scan targets and updates Cloudflare detectors for 2026+ credential formats. With over 800 secret types classified and live validation via login attempts, TruffleHog helps teams catch exposed API keys, passwords, and keys in Git, Slack, wikis, and logs. The catch: Heavy reliance on active verification can trigger rate limits or alerts in monitored systems during scans.

Use Cases
  • DevOps engineer scanning GitHub repos for leaked AWS keys
  • Security team monitoring Slack for exposed database passwords
  • Compliance officer validating Stripe secret permissions in wikis

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

Wazuh Agent Fixes Address Clock Rollback and AWS SQS Detection Gaps 🔗

Manager updates improve cluster stability and API auth while removing deprecated database daemon

wazuh/wazuh · C++ · ▲ 10 in 1d Est. 2015

Wazuh v4.14.

7 patches critical agent issues: AWS SQS subscriber now resolves correct accounts for cross-account IAM roles, and keepalive scheduling survives system clock rollbacks to prevent false disconnections. eBPF FIM fixes restore file move/rename visibility on older kernels like Amazon Linux 2. Manager-side, cluster payload buffering and archive decompression are stabilized, API auth gains bounded thread pools and regex timeouts, and the deprecated wazuh-dbd daemon is dropped. The catch: Open issues near 3,000 suggest ongoing complexity in scaling distributed deployments despite recent stability gains.

Use Cases
  • Detect malware and rootkits on cloud workloads
  • Monitor file integrity in containerized environments
  • Analyze logs for PCI-DSS compliance audits

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

Quick Hits

web-check An all-in-one OSINT tool that analyzes any website for security insights, intelligence gathering, and vulnerability discovery. 34.5k
bettercap A versatile network reconnaissance and MITM framework supporting Wi-Fi, Bluetooth, CAN-bus, and more for advanced penetration testing. 19.8k
maigret A username-based OSINT tool that aggregates personal data from over 3,000 websites to build detailed digital dossiers. 36.6k
ProxmoxVE A collection of community-driven shell scripts that simplify Proxmox VE deployment, configuration, and maintenance tasks. 29.2k
sniffnet A Rust-based network traffic monitor that provides real-time, user-friendly visibility into internet activity and connections. 40.5k
hosts A curated hosts file aggregator that blocks ads, malware, and unwanted content with optional extensions for porn, social media, and more. 30.8k

Hugo v0.164.0 Boosts Site Build Speed With Pandoc Citations And Dark/Light Syntax Highlighting 🔗

Release fixes performance regression in large sites and adds HexEncode/HexDecode template functions for developers

gohugoio/hugo · Go · ▲ 15 in 1d Est. 2013 · Latest: v0.164.0

Hugo’s latest release, v0.164.

0, sharpens its edge as a Go-based static site generator by introducing Pandoc citation support and dark/light syntax highlighting pairs via Chroma. Developers gain new template functions—encoding.HexEncode, encoding.HexDecode, and crypto.Hash—for direct data manipulation in layouts. The update also resolves a performance regression from v0.128.0 that slowed builds on larger sites, addressing a pain point noted in community discussions. Layout rendering now supports subpaths in .Render calls, and RST markup gains a syntaxHighlight option. These changes reinforce Hugo’s focus on speed and flexibility without overhauling its core pipeline.

The catch: Hugo’s rapid release cadence and extensive feature set can overwhelm newcomers, and its Go-only plugin model limits extensibility for teams invested in JavaScript or Python toolchains.

Previously in The Times “covered” — Aug 8

Use Cases
  • Build documentation sites with Pandoc citations and syntax highlighting
  • Generate fast blogs using dark/light code themes and asset pipelines
  • Deploy enterprise sites needing subpath layouts and crypto template functions

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

More Stories

RTK v0.45.0 Adds Transparent Hooks for Mistral Vibe CLI Agents 🔗

New rewrite feature compresses multiline command outputs to reduce LLM token load

rtk-ai/rtk · Rust · ▲ 235 in 1d 6mo old

The latest RTK release v0.45.

0 introduces transparent hook support for Mistral Vibe CLI, allowing agents to automatically route commands through the proxy without manual configuration. A new rewrite feature now processes multiline blocks—like log files or diff outputs—applying smart filtering and deduplication before they reach the LLM context. Built as a single Rust binary with zero dependencies, RTK installs to ~/.local/bin and supports over 100 common dev commands including git, ls, and cargo. The tool claims 60-90% token reduction by stripping boilerplate, grouping similar outputs, and collapsing redundant lines.
The catch: Hook-based automation only works in Bash environments; Claude Code’s native tools like Read and Grep bypass the proxy unless invoked manually via rtk read or rtk grep.

Previously in The Times “covered” — Aug 6

Use Cases
  • Developers reducing Claude Code token costs during debugging sessions
  • AI agents optimizing context usage in autonomous coding workflows
  • Rust teams minimizing LLM expenses in CI/CD pipelines with frequent command output

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

Loongson and Phytium Support Added to LEDE Fork for Chinese Edge Gateways 🔗

Project maintains 19,225 forks while addressing 866 open issues in active development

coolsnowwolf/lede · C · 31.6k stars Est. 2017

The coolsnowwolf/lede repository provides a Lean-based LEDE source fork targeting Loongarch64 and Phytium D2000 architectures for OpenWrt-based firmware. Recent commits added kernel bumps to 5.

4.246/6.1.34, MultiPath TCP support, and fixes for crypto/sha1-arm.ko missing errors. It enables multi-system deployment (OpenWrt/Debian/Armbian) on RuiSha network computing devices with 2.5G Ethernet options. Builders use it to compile custom gateways for IoT and small business networks, following strict WSL case-sensitivity workarounds and avoiding root user compilation. The catch: 866 unresolved issues suggest ongoing stability challenges despite frequent kernel updates and hardware enablement patches.

Use Cases
  • Develop OpenWrt firmware for Loongson-based edge gateways
  • Build multi-system images for RuiSha Phytium D2000 hardware
  • Compile customized firmware with MPTCP and 2.5G NIC support

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

Rust Compiler Patch Fixes LLVM-Induced Miscompilation in 1.97.1 🔗

Rust 1.97.1 addresses critical optimization bug affecting release builds

rust-lang/rust · Rust · ▲ 49 in 1d Est. 2010

The rust-lang/rust project released version 1.97.

1, patching a miscompilation triggered by LLVM optimizations that could corrupt release builds. The fix reverts a rustc change and bumps the LLVM submodule to include the upstream correction. Rust’s compiler, standard library, and toolchain — including Cargo, rustfmt, Clippy, and rust-analyzer — remain central to systems and application development. Despite its maturity, the project carries 12,726 open issues, reflecting ongoing complexity in language evolution and tooling maintenance.
The catch: Rapid release cadence and deep LLVM integration increase the risk of subtle, hard-to-diagnose codegen regressions in performance-critical code.

Use Cases
  • Building operating systems and kernels with memory safety
  • Developing high-performance web services and APIs
  • Creating embedded firmware for resource-constrained devices

Source: rust-lang/rust — based on the README and release notes.

Quick Hits

bitcoin Bitcoin Core provides the secure, decentralized foundation for Bitcoin transactions, enabling trustless peer-to-peer value transfer with robust cryptographic validation. 89.9k
traefik Traefik dynamically routes and secures microservices traffic in cloud-native environments with automatic service discovery and zero-configuration load balancing. 64.4k
PowerToys PowerToys enhances Windows productivity through customizable utilities like FancyZones and PowerToys Run, letting builders tailor their workflow for efficiency. 137.7k
git Git’s source code mirror enables developers to study and contribute to the world’s most widely used version control system via email-based patch submission. 62.5k
grpc gRPC delivers high-performance, language-agnostic RPC communication with strong typing and efficient serialization, ideal for scalable distributed systems. 45.3k

Tulip CC and AMYboard bring portable Python synthesis to ESP32-S3 hardware 🔗

New web and desktop versions expand access beyond $29.90 AMYboard and $59 Tulip CC kits

shorepine/tulipcc · C · ▲ 1 in 1d Est. 2022 · Latest: tulip

The Tulip Creative Computer and AMYboard now run as web apps and native desktop clients for Mac, Linux, and Windows Subsystem for Linux, letting users code synthesizers, games, or art without buying hardware. Built on MicroPython, the AMY synthesis engine, and LVGL graphics, the system targets the ESP32-S3 chip for low-latency audio and touchscreen interaction.

The latest rolling release, TULIP4_R11, over-the-air updates Tulip CC devices via tulip.upgrade() on every push to main, ensuring real-time firmware stays current.
AMYboard adds professional audio I/O: 10Vpp modular levels, CV jacks, S/PDIF, and audio input, positioning it as a Eurorack-compatible sound processor at under $30. Tulip CC combines a touchscreen display, MIDI hardware, and sensor access in a self-contained unit that boots instantly to a Python prompt. Both projects are fully open source, with hardware schematics and code available for DIY assembly or purchase through Makerfabs.
Developers can prototype musical instruments, interactive installations, or lo-fi sequencers using familiar Python syntax, leveraging built-in wavetable synthesis, FFT analysis, and real-time graphics. The web version mirrors most features, enabling collaboration and sharing via browser-based IDEs.
The catch: The ESP32-S3’s dual-core architecture limits polyphony and complex DSP, making heavy reverb or multi-voice FM synthesis prone to glitches without careful optimization.

Previously in The Times “covered” — Aug 10

Use Cases
  • Build portable synths with Tulip CC for live electronic music
  • Design Eurorack modules using AMYboard’s CV and audio jacks
  • Teach DSP and real-time systems via browser-based Tulip tutorials

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

More Stories

ESPectre 2.8.0 refines Wi-Fi motion detection with hardened calibration and ML detector upgrades 🔗

Release replaces DNS traffic with ping, enables Hampel filtering by default, and improves neural network reliability across ESP32 chips

francescopace/espectre · Python · ▲ 3 in 1d 9mo old

The latest ESPectre release sharpens its Wi-Fi sensing core, adopting ICMP ping as the default traffic generator to bypass DNS-restrictive routers. Hampel outlier filtering now runs enabled by default, reducing false positives in noisy RF environments.

Machine learning detection, introduced as an experimental feature, sees improved reliability through chip-grouped training and a revised 9-feature neural network model that outputs a gradual temperature-scaled Movement Score for smoother Home Assistant integration. Builders still need an ESP32-C6 or S3 with CSI support and a 2.4GHz Wi-Fi network to deploy the sensor, which operates without cameras or microphones. The catch: The ML detector remains labeled experimental, with no guarantees on performance across all housing layouts or motion types, and feedback is actively sought via GitHub discussions.

Previously in The Times “covered” — Aug 7

Use Cases
  • Home automation enthusiasts adding presence sensing
  • Renters deploying camera-free motion monitoring
  • Developers testing Wi-Fi sensing for occupancy analytics

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

Ibex RISC-V Core Gains Bit-Manipulation Support for Embedded Security 🔗

LowRISC updates Ibex with B-extension verification and tape-out readiness

lowRISC/ibex · SystemVerilog · 2k stars Est. 2017

Ibex, the 32-bit RISC-V CPU core from lowRISC, now includes verified support for the Bit Manipulation (B) extension in SystemVerilog, enhancing its suitability for security-focused embedded control. Recent commits show active development, with the last push just one day ago and ongoing verification of RV32IMBC configurations.

The core remains parametrizable, offering trade-offs between multiplier architecture, performance, and area, with Yosys synthesis showing ~23kGE for baseline builds. Ibex has seen multiple tape-outs and is used in PULP-based systems, though formal verification for advanced extensions like B is still marked as amber in some configs, indicating ongoing work. The catch: Full verification of security-critical bit-manipulation instructions remains incomplete, requiring careful validation before deployment in hardened systems.

Use Cases
  • Embedded controllers needing RISC-V with compressed and bit-manip instructions
  • Low-power IoT devices requiring configurable CPU core with M and C extensions
  • Academic tape-outs verifying RISC-V core security extensions in silicon

Source: lowRISC/ibex — based on the project README.

Stream Deck SDK simplifies plugin building with Node.js CLI 🔗

Enables hot-reload debugging and action-based event handling for custom hardware integrations

elgatosf/streamdeck · TypeScript · 255 stars Est. 2023

The Stream Deck SDK provides TypeScript tools for building plugins that interact with Elgato Stream Deck hardware via Node.js 24+.

Developers use the streamdeck create CLI to scaffold projects, then run npm run watch for hot-reloading during development. Actions inherit from SingletonAction to handle key presses, dial turns, and other inputs. Debugging works with VS Code or Chrome using manifest-configurable inspect modes. Recent commits show active maintenance with updates to plugin bundling and manifest validation.
The catch: The SDK requires Node.js 24+, excluding older LTS versions and complicating deployment on constrained or legacy systems.

Use Cases
  • Builders create custom macro controllers for video editing
  • Developers design interactive dashboards for streaming software
  • Makers integrate hardware knobs with DAW transport controls

Source: elgatosf/streamdeck — based on the project README.

Quick Hits

maker.js Maker.js lets builders design and model 2D vector shapes for CNC and laser cutting with precise, scriptable TypeScript tools. 2k
makerpnp MakerPnP streamlines PCB assembly by combining part planning, pick-and-place automation, and Gerber viewing in a Rust-powered workflow. 48
glasgow Glasgow is a versatile open-source hardware tool that acts as a multi-protocol "Swiss Army Knife" for debugging and interfacing with electronic circuits. 2.2k
childrens-clock Children's Clock is an IoT-enabled, child-friendly timepiece that teaches time-telling through engaging visuals and simple interactions. 54
FlightTracker FlightTracker turns a Raspberry Pi into a real-time aviation monitor, displaying live aircraft, satellite, weather, and time data on a dot matrix display. 187

Rust-based Minecraft server Pumpkin gains Bedrock support after two years of development 🔗

Project nears 1.0.0 with dual-edition compatibility and improved chunk handling

Pumpkin-MC/Pumpkin · Rust · ▲ 9 in 1d Est. 2024

Pumpkin, a Minecraft server written in Rust, has added preliminary Bedrock Edition support after two years focused on Java Edition. The update enables cross-protocol hosting, allowing administrators to run both Java and Bedrock clients on a single instance—a rare feature in open-source server software.

Built for performance, Pumpkin uses multi-threading and asynchronous I/O to reduce tick latency. Its TOML-based configuration replaces complex plugin.yml files, simplifying setup for custom rules, view distances, and feature toggles. Recent commits show progress in chunk saving algorithms (Linear and Pump variants) and entity AI for villagers and bosses, addressing long-standing gaps in vanilla-compatible behavior.

The project now tracks 724 forks and averages less than a day between commits, indicating active maintenance. Documentation highlights RCON, query protocol, and proxy support for Bungeecord and Velocity, positioning Pumpkin as a potential alternative to Paper or Velocity in proxy chains.

The catch: Bedrock Edition remains marked as work-in-progress, with entity spawning, liquid physics, and advancements incomplete—builders needing full Bedrock compatibility should test carefully before production use.

Previously in The Times “covered” — Aug 9

Use Cases
  • Host Java and Bedrock players on one server
  • Run low-latency servers with Rust performance
  • Customize server behavior via TOML config files

Source: Pumpkin-MC/Pumpkin — based on the project README.

More Stories

Magpie Upscales Legacy Windows Apps with HLSL-Powered Clarity 🔗

Lightweight tool applies Anime4K and FSR filters to boost resolution without fullscreen mode

Blinue/Magpie · HLSL · 14.3k stars Est. 2021

Blinue/Magpie uses HLSL shaders to upscale individual Windows 10/11 windows, letting users run older games or apps at higher clarity via Anime4K, FSR, or CRT effects. Its WinUI interface supports light/dark themes and multi-monitor setups, while auto-hide cursor and improved scaling prevent interruptions from pop-ups.

The tool requires DirectX 11 and works best when target apps use native resolution scaling.
The catch: Upscaling quality varies by source content, and some legacy apps may still exhibit artifacts or instability despite compatibility tweaks.

Use Cases
  • Gamers upscaling 16-bit era titles on modern high-DPI displays
  • Developers testing UI scaling behavior across resolutions
  • Archivists enhancing clarity of abandoned Windows software without source access

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

Open 3D Engine 2605.0 adds Wwise audio integration for AAA game workflows 🔗

Release includes CMake 3.24+ requirement and Visual Studio 2019 setup for multi-platform builds

o3de/o3de · C++ · ▲ 3 in 1d Est. 2021

O3DE’s latest release integrates the Wwise Audio Engine Gem, enabling spatial audio and middleware control in simulations and games. Developers must install Visual Studio 2019 16.

9.2+, CMake 3.24.0, and configure Git LFS for large assets. The engine supports Windows and Linux for high-fidelity 3D worlds without licensing fees. Despite active commits, 3,483 open issues highlight ongoing stability challenges in animation and physics systems.
The catch: Complex setup and dependency management may deter small teams seeking lightweight, out-of-the-box tools.

Use Cases
  • Game studios building cross-platform AAA titles
  • Simulation creators rendering cinematic 3D environments
  • Developers implementing custom audio pipelines with Wwise integration

Source: o3de/o3de — based on the README and release notes.

Quick Hits

libgdx LibGDX lets Java developers build and deploy high-performance games across desktop, mobile, and web with a single codebase. 25.3k
godot Godot Engine empowers creators to design 2D and 3D games for any platform using an intuitive, open-source workflow. 115.6k
MonoGame MonoGame enables C# developers to craft cross-platform games with familiar XNA-like APIs and broad hardware support. 14.3k
raylib Raylib offers a lightweight, beginner-friendly C library for rapid prototyping and learning game programming fundamentals. 34.3k
SpacetimeDB SpacetimeDB provides Rust developers with a real-time, distributed database optimized for low-latency game state synchronization. 25k
pyxel Pyxel delivers a minimalist, retro-inspired game engine in Python for creating 8-bit style games with minimal setup. 17.6k
The Git Times AI Desk
Ask about today's stories — or hit “Ask about this” on any article to focus on one.

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

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