Design
Preset
Background
Text
Font
Size
Width
Account Pricing

The Git Times

AI Models
Claude Opus 5 $25/M GPT-5.6 Luna $1.20/M Gemini 3.1 Pro Preview $12/M Grok 4.6 $6/M DeepSeek V4 Pro 0813 $3.56/M Qwen3.8 2.4T A95B $6/M Kimi K3 $15/M
Full Markets →

SAGE Router Dynamically Chooses Agent Collaboration Mode at Runtime 🔗

Evaluates SELF, COLLABORATE, or HANDOFF options using progress-aware utility under real-world constraints

wang2122/sprix-sage-router · Python · ▲ 249 in 1d 3d old

The wang2122/sprix-sage-router project introduces SAGE—State-Aware Graph Exchange—as a decision layer that determines how AI agents should collaborate during task execution. Built for Agent2Agent (A2A) networks, it moves beyond static discovery by evaluating three runtime routing modes—SELF, COLLABORATE, or HANDOFF—within a single auditable utility function.

Rather than relying on disconnected heuristics, SAGE weighs progress, dependencies, agent complementarity, and contextual trust to decide whether an agent should continue alone, recruit collaborators, or hand off the task.

It assigns remaining requirements to executors, converts dependency edges into an inspectable communication topology, and estimates critical-path latency. A learned outcome model replaces fixed success equations with a regularized online predictor that adapts from execution evidence, while bounded beam search explores multiple team prefixes to avoid greedy commitments. Crucially, SAGE rewards marginal requirement coverage over redundant high-ranked agents, ensuring teams are formed for complementary strength, not prestige.

Designed to sit above the A2A protocol, SAGE consumes Agent Cards, messages, and artifacts but adds the intelligence needed for dynamic, constraint-aware orchestration under permission, budget, and deadline limits. Its progress-aware replanning means active execution—completed nodes, failures, and transferred context—directly influences routing decisions, enabling adaptive recovery and re-planning mid-task.

The catch: As a recent research output with no open issues and only one fork, SAGE remains early-stage; its real-world efficacy in large-scale, heterogeneous agent networks has yet to be validated beyond controlled benchmarks.

Why this leads today Sprix AI’s state-aware routing enables reliable, scalable collaboration in multi-agent systems by letting agents dynamically self-manage, collaborate, or handoff based on context — a necessary primitive for production AI workflows.

Use Cases
  • Dynamic agent team formation for complex task execution
  • Runtime adaptation of collaboration strategies in AI workflows
  • Constraint-aware routing under budget, deadline, and permission limits

Source: wang2122/sprix-sage-router — based on the project README.

More on the Front Page

Vanity-eth Enables Offline, Auditable Vanity Address Generation for Bitcoin and Ethereum 🔗

CPU-based multi-process search supports Legacy, SegWit, Taproot, and EIP-55 with zero network calls

Leutenegger/vanity-eth · Python · 801 stars 0d old

Leutenegger/vanity-eth is a Python tool that generates cryptocurrency vanity addresses entirely offline, ensuring private keys never leave the user’s machine. It searches for private keys whose derived addresses match user-specified patterns—prefix, suffix, or substring—across Bitcoin address types (Legacy, Nested SegWit, Native SegWit, Taproot) and Ethereum (EIP-55).

The tool uses multi-process parallelization to distribute the brute-force search across CPU cores, with live progress reporting on attempts, rate, and elapsed time. Built for auditability, it includes independent verification helpers (verify_eth_address, verify_btc_address) to confirm that a printed private key correctly derives the shown address before any funds are moved. Dependencies are minimal and transparent: coincurve for secp256k1 operations, pycryptodome for Keccak-256, and click/tqdm for CLI and progress. No telemetry, no external key material, and no network requests are made—by design. The README stresses that users must run it on a fully controlled machine, preferably air-gapped when handling real funds, and never expose private keys to untrusted software. While readable and secure, the tool intentionally sacrifices speed for simplicity and inspectability, making it unsuitable for high-throughput use cases compared to GPU- or C++-based alternatives.
The catch: It is significantly slower than specialized GPU or C++ vanity miners, trading performance for transparency and ease of audit.

Use Cases
  • Developers generating secure vanity addresses for personal wallets
  • Auditors verifying key-to-address derivation in isolated environments
  • Educators teaching cryptocurrency address generation and privacy practices

Source: Leutenegger/vanity-eth — based on the project README.

OpenBot Gives AI Agents Isolated Computers With Full Audit Trails 🔗

Each bot runs in its own browser container with policy-governed tool access and component output

CopilotKit/OpenBot · TypeScript · 1.8k stars 4d old

OpenBot lets you deploy AI coworkers that each get a dedicated browser, file system, and toolset—no shared logins or cross-bot data leaks. Every action is pre-approved by a CEL policy and recorded in an immutable trail before execution.

Bots stop and hand control back when they hit restricted tasks, then resume after human intervention. Built on AG-UI, it accepts any agent framework or custom code, runs via Docker Compose with your chosen model and PostgreSQL, and outputs UI components instead of just text. Three example bots ship as config: General Assistant, Knowledge, and Risk Analyst.
The catch: Early alpha (v0.0.1) with 33 open issues means rough edges and active changes—expect bugs and shifting APIs.

Use Cases
  • Developers testing AI agents with secure browser automation
  • Compliance teams deploying governed AI for risk analysis
  • Companies running private AI coworkers on internal infrastructure

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

T3 Code Unifies Agent Control Across Mobile, Web, and Desktop Apps 🔗

Latest release adds drag-to-reorder pinned threads and improved remote reconnect stability

pingdotgg/t3code · TypeScript · ▲ 19.8k in 1d 6mo old

T3 Code provides a centralized control surface for managing AI agent subscriptions like Claude Code, Codex, Cursor, Grok Build, and OpenCode through iOS, Android, web, and Electron apps. The v0.

0.33 release introduces drag-to-reorder pinned threads in the web interface and fixes for mobile reconnects that previously shifted thread lists. Remote access from phones or other machines remains a core feature, with server sync maintained via Node.js 22.16+ backends. Developers can install the desktop app from GitHub Releases or package registries, though the project notes it is still early-stage with expected bugs. The catch: Contributions are largely closed except for small fixes, limiting community-driven evolution despite active development.

Use Cases
  • Developers sync agent settings across phone and laptop
  • Teams standardize agent workflows via centralized control
  • Power users customize agent interfaces with theme-aware sidebars

Source: pingdotgg/t3code — based on the README and release notes.

Wake Unifies Mac Coding-Agent Sessions in One Searchable App 🔗

Built with Rust and GPUI, it indexes local agent histories without network calls.

iAmCorey/Wake · Rust · 355 stars 2d old

Wake gathers every coding-agent session on a Mac into a single native app, letting users browse, search, and resume conversations instantly. It reads data read-only from directories like `~/.

claudeand~/.codex`, using SQLite FTS5 for full-text search and tree-sitter for code highlighting across 30+ languages. Sessions can be resumed with one click in Terminal/iTerm at the original project directory, and all data stays local with zero network requests. The app supports starring, pinning, and exporting to Markdown, while respecting agent-specific limitations like encrypted or cloud-stored histories. The catch: Wake does not support agents that encrypt local data (Cursor, Windsurf, Trae) or store sessions in the cloud (Amp, Factory, Warp), limiting its universal applicability.

Use Cases
  • Developers searching past Claude conversations by keyword
  • Engineers resuming Codex sessions from a unified interface
  • Teams auditing agent interactions across multiple tools locally

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

Next.js Adds Turbopack Fixes for Shared Runtime Chunks in v16.3.1 🔗

Release preserves async-module runtime and improves HMR for dynamic imports in layouts

vercel/next.js · JavaScript · 141.9k stars Est. 2016

Vercel’s Next.js v16.

3.1 patches Turbopack to retain async-module runtime in shared chunks and collapse nested promises in the analyzer. It also fixes image response preservation after optimization and restores live headers() views in SSR. Backported changes address styled-jsx styling in Pages Router and HMR for dynamic imports from layouts. The framework continues to integrate Rust-based tooling for faster builds, supporting full-stack React apps with server rendering and static generation.
The catch: Open issues exceed 4,000, signaling ongoing complexity in balancing cutting-edge features with stability across diverse deployment targets.

Use Cases
  • Enterprises building SEO-optimized React sites with incremental static regeneration
  • Developers creating full-stack apps using App Router and server actions
  • Teams deploying Jamstack sites with hybrid rendering and edge middleware support

Source: vercel/next.js — based on the README and release notes.

Python CLI Enables GetContact Number Lookups Without Android or Browser 🔗

Tool handles authentication, CSV batch processing, and account rotation via WhatsApp verification

xdreizein666/getcontact-cli · Python · 323 stars 1d old

xdreizein666/getcontact-cli is a Python command-line tool that interacts directly with the GetContact API to retrieve phone number information such as display names, tags, and email addresses. It supports batch lookups via CSV files, automatic result saving to a results/ directory, and dynamic account switching when rate limits trigger CAPTCHA or HTTP 403 responses.

Users must provide their own GetContact credentials, obtained through WhatsApp verification, and are responsible for complying with privacy norms and terms of service. The project positions itself as an independent research effort for understanding mobile backend communication and API authentication flows.
The catch: Requires manual credential management and poses privacy compliance risks if misused, with no built-in safeguards against abusive or large-scale data harvesting.

Use Cases
  • Developers testing GetContact API integration logic
  • Researchers studying mobile service authentication mechanisms
  • Individuals performing occasional phone number lookups via CLI

Source: xdreizein666/getcontact-cli — based on the project README.

CyberLeek repo verifies GTA 6 leaks with on-chain proof and scam alerts 🔗

Community researchers map Vice City details while exposing fake Discord and Twitter imposters

zyrexdz/cyberleek-leak-research · Unknown · 427 stars 2d old

The zyrexdz/cyberleek-leak-research project breaks down the August 2026 GTA 6 gameplay leaks using technical video analysis, Arweave hashes, and on-chain evidence. It details map layouts, vehicle mechanics, and audio cues like the Tate McRae song, all tied to verified leak files.

Researchers have confirmed build age, physics behavior, and in-game CarPlay logos through frame-by-frame checks. The repo includes scam warnings about fake @cyberleek X accounts and Discord servers spreading unverified claims like Robux currencies or fake missions. The catch: The project relies on community-submitted evidence and lacks formal peer review, leaving room for misinformation despite its verification efforts.

Use Cases
  • Game analysts verify leaked GTA 6 footage authenticity
  • Modders study Vice City map details for accuracy
  • Researchers track OSINT trails from dark web leak sources

Source: zyrexdz/cyberleek-leak-research — based on the project README.

Open Source AI Agents Gain Tools for Autonomy and Control 🔗

New projects give agents dedicated environments, memory guardrails, and cross-platform agency

Trendai-agents
CopilotKit/OpenBotwang2122/sprix-sage-routerbrowser-use/macos-harnessfufankeji/deepseek-harness-studioiAmCorey/Wakelycorp-jp/sim-usedust-tt/dustyvgude/lean-ctxlangfuse/langfusevolcengine/OpenVikingbojieli/ai-agent-bookifixai-ai/iFixAiinternet-court/internet-court-skillPrimeIntellect-ai/prime-agentcalesthio/OpenMontagett-a1i/archifychaitanyagiri/munder-difflinakitaonrails/ai-memoryayghri/i-have-adhdWenyuChiou/awesome-agentic-ai-zhblader/humanizeryc-software/qmcan1357/oh-my-piwhiteguo233/OpenBiliClawTencentCloud/TencentDB-Agent-Memoryop7418/guizang-ppt-skillDevin-AXIS/iPolloWorkHKUDS/CLI-Anythingmaka-agent/maka-agentK-Dense-AI/scientific-agent-skillsyizhiyanhua-ai/fireworks-tech-graphagegr/pi-webHKUDS/nanobotZSeven-W/openpencilalibaba/open-code-reviewvercel-labs/agent-browsercolbymchenry/codegraphcurl/curlvercel-labs/fxjustrach/codedb

A clear pattern is emerging in open source AI agent development: the shift from experimental prompts to structured, autonomous systems with dedicated runtime environments and safety controls. Repos like browser-use/macos-harness and lycorp-jp/sim-use now provide thin, platform-specific harnesses that grant LLMs direct, unmediated control over macOS, iOS, and Android interfaces — treating the agent not as a chatbot but as an actor with real system access.

Meanwhile, yvgude/lean-ctx introduces a local Rust binary that enforces context boundaries, reducing token usage by 60–90% while logging every read, write, and learn operation — effectively giving agents a provable memory ledger.
Collaboration is also being formalized: wang2122/sprix-sage-router implements state-aware routing for agent-to-agent handoffs, enabling dynamic SELF/COLLABORATE/HANDOFF logic in multi-agent networks. For persistence and sharing, TencentCloud/TencentDB-Agent-Memory offers a team-level memory hub that converts interactions into reusable, governed assets like skill graphs and code maps.
On the usability front, fufankeji/deepseek-harness-studio delivers a zero-code desktop launcher for Windows and macOS with plugin discovery and AI-recommended workflows, lowering the barrier to agent deployment.
The catch: Despite rapid innovation, many of these tools remain early-stage, with overlapping functionality, limited interoperability beyond specific frameworks (like MCP or AG-UI), and untested real-world reliability at scale — suggesting a Cambrian explosion that may yet consolidate around fewer, more robust standards.

Use Cases
  • Developers automate macOS workflows using vision-enabled agents
  • Teams share governed agent memories across coding and design tasks
  • Researchers deploy local-first agents with verifiable context boundaries

Local-First AI Agent Tooling Accelerates Across Open Source 🔗

Projects enable LLMs to control systems, manage context, and automate skills with minimal token overhead

Trendllm-tools
SigmanticAI/apex-inference-chipbrowser-use/macos-harnessNervercc/gpt_nerverbawadou/ai-data-extractormilvus-io/milvusdust-tt/dustyvgude/lean-ctxlangfuse/langfusecathrynlavery/diagram-designTiger3807861189/J-Space-Cognition-Suite-V3.6mukul975/Anthropic-Cybersecurity-Skillsvirgiliojr94/book-to-skillinternet-court/internet-court-skilllidge-jun/opencodexzhaoxuya520/reverse-skilllightningpixel/modlycoreyhaines31/marketingskillsWenyuChiou/awesome-agentic-ai-zhcitrolabs/ego-liteccch1mneyyy/dsh-TUITencentCloud/TencentDB-Agent-MemoryDevin-AXIS/iPolloWorkK-Dense-AI/scientific-agent-skillsZhuLinsen/daily_stock_analysisAlishahryar1/free-claude-codeshy3130/tickflow-stock-panelalibaba/open-code-reviewlnxjedi/gopherbotcolbymchenry/codegraphjustrach/codedb

A clear pattern is emerging in open source: developers are building lightweight, local-first tools that extend LLMs into practical agents by reducing reliance on cloud APIs and maximizing control. Projects like yvgude/lean-ctx (Rust) act as a context intelligence layer, cutting token use by 60–90% through selective memory and access gating, while browser-use/macos-harness gives LLMs direct, thin-layer control over macOS systems.

Skill automation is gaining traction via repos such as Nervercc/gpt_nerver (PyQt6 + JSON-RPC) for ChatGPT Plus workflows and alibaba/open-code-review (Go), which integrates LLMs into deterministic code review pipelines with fine-tuned rules. Vector search infrastructure like milvus-io/milvus (Go) underpins agent memory systems, complemented by TencentCloud/TencentDB-Agent-Memory (TypeScript), which structures agent knowledge into reusable assets. Meanwhile, colbymchenry/codegraph (C) and justrach/codedb (Zig) provide localized code intelligence to minimize tool calls, and dust-tt/dust (TypeScript) offers a customizable agent platform. These projects collectively shift AI agent development from prompt engineering toward observable, controllable, and efficient local execution—prioritizing precision, privacy, and reproducibility over raw scale.
The catch: Much of this tooling remains fragmented across languages and agent frameworks, with limited interoperability; many skills and context layers are tightly coupled to specific LLMs like Claude Code, raising concerns about long-term portability and whether early efficiency gains will hold under complex, real-world agent workflows.

Use Cases
  • Developers automate macOS tasks using LLMs via thin harness
  • Engineers reduce LLM token use with local context filtering
  • Teams deploy self-hosted agent workspaces for code and office work

Open Source Shifts Toward Specialized, Purpose-Built Web Tools 🔗

Developers favor lightweight, domain-specific frameworks over monolithic stacks for targeted web solutions.

Trendweb-frameworks
jeremy-prt/bloubzyrexdz/cyberleek-leak-researchxdreizein666/getcontact-clivega-org/vega-apptrezor/trezor-suitePostHog/posthogscrapy/scrapyvercel/next.jsdnshe/DNSHE-FreeDomainsrmyndharis/OpenWAagegr/pi-webNaiboWang/EasySpiderpf-robotics/kachaka-apicurl/curljstrieb/github-statsbyte4geek/weatherstationTarnadas/smmdbphaserjs/phaserdefold/defold

A clear pattern emerges in recent open source activity: a move away from all-encompassing web frameworks toward narrowly focused tools solving specific problems. Instead of adopting broad solutions, developers are assembling purpose-built components.

For example, vega-org/vega-app delivers a TypeScript-based media streaming client for Android, prioritizing video playback efficiency over general UI capabilities. Similarly, rmyndharis/OpenWA provides a self-hosted WhatsApp API gateway, enabling direct integration without relying on official or opaque SDKs. In the crawling space, NaiboWang/EasySpider offers a visual, no-code interface for web scraping, lowering the barrier for data extraction tasks traditionally requiring custom scripts. Meanwhile, agegr/pi-web supplies a dedicated web UI for interacting with AI coding agents, streamlining prompt engineering and output review. Even established projects like scrapy/scrapy are being used in more surgical ways — embedded as scraping engines within larger automation workflows rather than as standalone frameworks. This reflects a technical shift: teams now prefer modular, interoperable tools that excel at single functions, reducing complexity and vendor lock-in. The trend is evident in how new projects avoid frontend/backend monoliths, instead offering clean APIs or domain-specific UIs that plug into existing stacks.
The catch: While this modularity promises agility, it risks creating integration overhead — developers may spend more time connecting disparate tools than building features, and the lack of standardized interfaces between these niche projects could lead to fragmentation, especially when maintenance or security updates lag across the ecosystem.

Use Cases
  • Developers build custom media streaming interfaces
  • Teams automate WhatsApp messaging via self-hosted gateways
  • Analysts extract web data using visual, no-code scrapers

Deep Cuts

DeepSeek Harness Studio turns AI agents into one-click desktop apps 🔗

It bundles plugin discovery, smart recommendations, and visual enhancements for Windows and macOS

fufankeji/deepseek-harness-studio · TypeScript · ▲ 40 in 1d

DeepSeek Harness Studio lets developers launch AI-powered agents without writing code. Built with TypeScript and Electron, it offers a zero-config desktop environment where users discover, install, and manage plugins through an intuitive interface.

The tool pushes trending plugins, uses AI to suggest relevant add-ons, and enhances workflows with visual tuning. For builders, it removes friction in agent experimentation—no CLI, no setup, just double-click to start. Whether prototyping conversational agents or automating tasks, the studio accelerates iteration by treating AI capabilities as pluggable components.
The catch: It's early-stage and niche, so plugin ecosystems and documentation remain sparse outside Chinese-speaking communities.

Use Cases
  • Developers testing DeepSeek-based agents without coding
  • Teams managing AI plugin workflows across Windows and macOS
  • Beginners exploring agent capabilities through visual plugin discovery

Source: fufankeji/deepseek-harness-studio — based on the project README.

Quick Hits

coldcard-airgap Leutenegger/coldcard-airgap (Python): Provides offline utilities for Coldcard wallet users to inspect PSBTs, manage entropy, split seeds, encode BBQr, and verify firmware — no internet needed. 608
cve-2026-41940-PoC lanicer/cve-2026-41940-PoC (Python): Demonstrates a critical authentication bypass in cPanel/WHM, enabling unauthorized admin access via exploited session handling flaws. 322
ai-data-extractor bawadou/ai-data-extractor (Python): Extracts and structures chat histories from AI coding assistants like Claude Code and Cursor for analysis, fine-tuning, or audit — free and open-source. 532
ai-trader-bot MIgHTy-alIeN/ai-trader-bot (Solidity): Implements an on-chain arbitrage bot controlled by off-chain automation, executing profitable trades across DEXs via smart contract logic. 2.7k
desktop-fly DenisSergeevitch/desktop-fly (Swift): Brings a biologically realistic 3D fruit fly to your macOS desktop, simulating neural activity from the FlyWire connectome in real time. 653
apex-inference-chip SigmanticAI/apex-inference-chip (Python): Delivers an FPGA-based LLM inference chip running Qwen2.5-0.5B with bit-exact RTL transformer layer, achieving 0.56 tok/s — a 140× speedup over prior designs. 649
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.

From the labs & arXiv

OpenClaw v2026.7.1-2 fixes npm plugin updates for official tool compatibility 🔗

Patch resolves singleton-array metadata handling to restore plugin install and update flows

openclaw/openclaw · TypeScript · ▲ 130 in 1d 8mo old · Latest: v2026.7.1-2

The latest OpenClaw release patches a regression in npm plugin updates that blocked official plugin installations. Version 2026.

7.1-2 accepts singleton-array metadata from newer npm clients, correcting a parsing issue that prevented tracked plugins from updating to patch releases. This fix, detailed in issue #108336, restores expected behavior for developers relying on OpenClaw’s extensible tool ecosystem. OpenClaw remains a TypeScript-based personal AI assistant that runs locally, connecting models and tools through a Gateway to messaging channels like WhatsApp, Slack, and Signal. It provisions Node.js runtimes automatically or uses existing installs (Node 22.22.3+, 24.15+, or 25.9+). The assistant treats inbound messages as untrusted input and requires explicit approval for pairing with unknown senders on DM-capable channels. While the core architecture supports hosted and local models, voice, canvas, and device-local actions via companion apps, the project carries significant maintenance overhead: 5,838 open issues indicate ongoing stability and feature work. The fix is narrow but necessary—without it, plugin updates fail silently, undermining the assistant’s promised extensibility.
The catch: The high volume of open issues suggests unresolved reliability concerns that may affect production use despite the assistant’s local-first, privacy-focused design.

Previously in The Times “covered” — Aug 18

Use Cases
  • Developers extending AI assistant with custom tools
  • Teams deploying private assistants across messaging platforms
  • Users avoiding cloud data exposure with local model execution

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

More Stories

Dify adds multi-select tool inputs and node locator in latest release 🔗

Workflow enhancements streamline debugging and configuration for agentic AI builders

langgenius/dify · TypeScript · ▲ 114 in 1d Est. 2023

The langgenius/dify project released version 1.16.

1 with workflow tool nodes now supporting multi-select dropdown inputs, letting users choose multiple values when configuring tool parameters (#39346). A new workflow node locator lets developers click a node_id in run logs or error messages to highlight the corresponding node in the editor canvas (#38187). Additional updates include improved block selector keyboard navigation, agent DSL YAML export from the sidebar, and expanded knowledge tracing for RAG pipeline observability. These changes reduce friction in complex workflow iteration and team collaboration.
The catch: Despite active development, 965 open issues suggest ongoing stability challenges in agent orchestration and self-hosted deployment consistency.

Previously in The Times “covered” — Aug 13

Use Cases
  • AI teams building multi-tool agent workflows
  • Developers debugging RAG pipelines via visual logs
  • Enterprises self-hosting LLM apps with compliance needs

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

Data Engineering Zoomcamp Adds Kestra Workflow Orchestration Module 🔗

Free 9-week course now includes hands-on training with Kestra for pipeline automation

DataTalksClub/data-engineering-zoomcamp · Jupyter Notebook · 44.8k stars Est. 2021

The DataTalksClub/data-engineering-zoomcamp repository updated its curriculum to include a dedicated module on Kestra, an open-source workflow orchestration tool, alongside existing content on Docker, Spark, dbt, and Kafka. The course teaches learners to build end-to-end data pipelines using pre-recorded lectures, scored homework, and a final project, with live cohort features like deadlines and peer review.

Materials are available via Jupyter Notebooks, Slack, and Telegram, supporting both self-paced and cohort-based learning. The catch: The course relies on pre-recorded content, limiting real-time instructor interaction despite its "live cohort" labeling.

Use Cases
  • Developers learn pipeline construction with Docker and Spark
  • Analysts gain dbt modeling and BigQuery optimization skills
  • Data scientists practice end-to-end workflow automation with Kestra

Source: DataTalksClub/data-engineering-zoomcamp — based on the project README.

Hermes Agent v0.20.4 patches skill evaluation and cron media delivery 🔗

Release fixes NVIDIA security checks on skill installs and cron attachment handling

NousResearch/hermes-agent · Python · ▲ 477 in 1d Est. 2025

The Hermes Agent v0.20.

4 patch release rolls up ~74 merged PRs since v0.20.3, including NVIDIA SkillEvaluator Tier 1 advisory scanning that now performs license and security checks during skill installs. Cron media-send hardening adds configurable timeouts, manual-run attachment support, and missed-fire surfacing for unreliable deliveries. SessionDB event-loop-thread contention fixes reduce instability under load, while hermes update now provides parked-branch honesty to prevent silent update failures. The update also resolves tabbed SESSIONS|BOTS sidebar rendering issues and improves Bot Mode group-chat Markdown handling and cross-machine routing. Built on Python, the agent supports multi-model backends and portable TUI across Telegram, Discord, and CLI. The catch: With 34,072 open issues, the project’s rapid feature expansion may outpace stability fixes for production-critical workflows.

Previously in The Times “covered” — Aug 15

Use Cases
  • DevOps teams automate serverless skill deployment via cron scheduler
  • Researchers build self-improving agents using portable VPS or GPU clusters
  • Power users delegate complex tasks to isolated subagents for parallel execution

Source: NousResearch/hermes-agent — based on the README and release notes.

Quick Hits

AutoGPT AutoGPT empowers anyone to build and deploy autonomous AI agents that reason, act, and adapt — no PhD required. 186.7k
openai-cookbook The OpenAI Cookbook offers practical, ready-to-run notebooks showing how to leverage GPT models for real-world applications. 75.4k
langchain LangChain provides the modular framework to chain LLMs with tools, memory, and data for sophisticated AI agent development. 144.7k
ComfyUI ComfyUI delivers a flexible, node-based GUI for designing and executing complex diffusion workflows with full API and backend control. 128.7k
airllm AirLLM enables efficient 70B parameter LLM inference on a single 4GB GPU through optimized quantization and memory management. 32k

RTAB-Map 0.23.8 adds CUDA-enabled OpenCV and RealSense2 support 🔗

Windows binaries now bundle OpenCV 4.13.0 with GPU acceleration and Intel RealSense SDK 2.56.5

introlab/rtabmap · C++ · ▲ 2 in 1d Est. 2014 · Latest: 0.23.8

The RTAB-Map SLAM library updated its Windows dependencies to include CUDA-accelerated OpenCV 4.13.

0 and RealSense2 2.56.5 for T265/D400-series cameras. Built with Qt 6.8.3 and PCL 1.15.0, the release maintains ROS 1 Noetic and ROS 2 Humble-through-Rolling compatibility. Despite active commits, 578 open issues persist, including Kinect-for-Xbox-360 failures on Windows 11.
The catch: Windows 11 users cannot rely on Kinect for Xbox 360 sensors due to unresolved driver incompatibilities.

Use Cases
  • Robotics researchers mapping indoor spaces with RGB-D cameras
  • Mobile developers integrating SLAM into Android/iOS scanning apps
  • ROS 2 Humble users fusing lidar and visual odometry for localization

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

More Stories

mavros 2.15.0 fixes param race crash and updates ROS 2 toolchain 🔗

Refactored mavconn now supports MAVLink v2 signing for secure autopilot links

mavlink/mavros · C++ · 1.2k stars Est. 2013

The mavros project released version 2.15.

0, patching a race condition in its param plugin that previously caused "Promise already satisfied" crashes during concurrent parameter updates. The update refactors the mavconn library to add MAVLink v2 signing support, enabling authenticated communication with modern autopilots over UDP, TCP, or serial links. Dependency bumps include urllib3, Sphinx, and fastcrcr for improved documentation builds and devcontainer consistency. Tested against PX4 and ArduPilot in SITL, the bridge remains ROS 2 Humble-through-Rolling native.
The catch: Open issues remain high at 413, with no indication of reduced maintenance burden despite recent activity.

Previously in The Times “covered” — Aug 14

Use Cases
  • Developers arm/disarm PX4 drones via ROS 2 services
  • Engineers stream IMU and GPS data as standard sensor_msgs
  • Researchers proxy GCS commands over UDP to simulated FCUs in CI pipelines

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

ROS 2 Documentation Streamlines Live Builds with Sphinx-Autobuild 🔗

Nightly Jenkins updates keep ROS.org docs current for 24.04 dev environments

ros2/ros2_documentation · Python · 983 stars Est. 2018

The ros2/ros2_documentation repo now emphasizes local iteration via sphinx-autobuild, which rebuilds docs on file save and auto-refreshes browsers at http://localhost:2022. Contributors use make html for static builds or make multiversion to test rolling branch output, both leveraging Sphinx’s parallel -j auto flag.

Dependencies are locked via pinned Python versions in constraints.txt, ensuring reproducibility on Noble (Ubuntu 24.04). Nightly Jenkins jobs push changes to docs.ros.org.
The catch: 205 open issues suggest ongoing maintenance strain despite active commits.

Previously in The Times “covered” — Aug 13

Use Cases
  • Documentation writers test live changes locally
  • Devcontainer users preview docs via port forwarding
  • Release engineers validate multisite builds pre-deploy

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

Quick Hits

autoware_universe autowarefoundation/autoware_universe: A modular, open-source autonomous driving stack in C++ enabling perception, planning, and control for self-driving vehicles across diverse environments. 1.7k
ardupilot ArduPilot/ardupilot: A battle-tested C++ flight stack powering ArduPlane, ArduCopter, ArduRover, and ArduSub for reliable, multi-domain autonomous navigation of aerial, ground, and underwater robots. 15.7k
easy_handeye2 marcoesposito1988/easy_handeye2: A Python-based ROS2 tool that automates hardware-independent hand-eye calibration, streamlining robot-arm-to-sensor alignment with minimal setup and maximum accuracy. 295
gopherbot lnxjedi/gopherbot: A Go-powered ChatOps engine for Slack that automates DevOps workflows through chat commands, blending Hubot’s flexibility with Jenkins-like pipeline control. 61
ed tue-robotics/ed: A C++ library that generates structured environment descriptors from sensor data, enabling robots to semantically understand and reason about dynamic surroundings for smarter navigation. 56
rosclaw Self-evolving runtime infrastructure for Physical AI and embodied agents. Ground AI agents into robot bodies with e-URDF, sandbox safety, capability routing, praxis capture, physical memory, runtime intervention, and skill evolution. 183

OpenCTI's latest release adds on-demand CI deployments for faster threat intel workflows 🔗

The update fixes critical backend bugs and stream handling and AI summary reliability issues

OpenCTI-Platform/opencti · TypeScript · ▲ 2 in 1d Est. 2018 · Latest: 7.260817.0

OpenCTI-Platform/opencti released version 7.260817.

017.0, introducing on-demand deployment for pre-built feature branch images in CI pipelines (#17620). This lets security teams spin up isolated CTI environments instantly from any branch, accelerating testing of new integrations or rule sets without disrupting production. The release also resolves a critical worker bug where malformed STIX messages caused indefinite queue blocking (#17686), a fix vital for maintaining real-time threat data ingestion. Additional fixes address case template authorization (#17618), Attack Pattern upsert crashes (#17421), and unreliable AI-generated summaries (#12432). Built with TypeScript and STIX2 schemas, OpenCTI remains a hub for correlating observables, TTPs, and attribution across tools like MISP and MITRE ATT&CK.

The catch: Despite active development, 2,136 open issues signal persistent complexity in managing CTI data lifecycles, and the Enterprise Edition’s gated features may limit advanced capabilities for budget-conscious teams.

Previously in The Times “covered” — Aug 18

Use Cases
  • Security analysts testing new threat intel integrations
  • SOC teams correlating observables across MISP and ATT&CK
  • Threat hunters validating AI-assisted summary accuracy

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

More Stories

GitHub repo equips AI agents with 817 structured cybersecurity skills 🔗

Latest release adds MITRE F3 framework and 55 new skills across AI, supply chain, and firmware domains

mukul975/Anthropic-Cybersecurity-Skills · Python · ▲ 3 in 1d 5mo old

The mukul975/Anthropic-Cybersecurity-Skills project provides 817 production-grade cybersecurity skills mapped to six frameworks including MITRE ATT&CK, NIST CSF 2.0, and the newly added MITRE F3.

Skills follow the agentskills.io standard and integrate with Claude Code, GitHub Copilot, Cursor, and 20+ platforms. Release v1.3.0 expanded coverage with 55 new skills in AI security, supply chain, and hardware/firmware domains, plus depth in identity, cloud-native, and DFIR areas. Each skill includes framework-specific guidance for tasks like memory forensics, Kerberoasting detection, and cloud breach scoping.
The catch: Open issues number 47, and offensive techniques require explicit authorization to avoid legal risk.

Use Cases
  • Security teams train AI agents for automated threat hunting
  • Red teams simulate attacks using structured skill libraries
  • Junior analysts accelerate incident response with framework-guided steps

Source: mukul975/Anthropic-Cybersecurity-Skills — based on the README and release notes.

Community scripts add Umbraco while updating for Debian 13 🔗

Latest release includes bug fixes and a breaking distro update for Proxmox VE users

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

The community-scripts/ProxmoxVE project released Umbraco as a new one-click installer and updated existing scripts to align with Debian 13, a breaking change affecting multiple services. Bug fixes addressed build paths in storyteller, OOM risks in scanopy, and iperf3 hangs in networkoptimizer.

Users gain streamlined setup for CMS and media tools via pvesh commands, with default and advanced modes tailoring resource allocation. Scripts now require manual intervention for legacy configurations post-Debian 13 shift.
The catch: Breaking distro updates may force reconfiguration of older containers, increasing maintenance overhead for long-term homelab setups.

Previously in The Times “covered” — Aug 16

Use Cases
  • Deploy Home Assistant in under five minutes via shell
  • Update Jellyfin media server with post-install helper
  • Migrate services to Debian 13 using updated script templates

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

Quick Hits

mitmproxy mitmproxy/mitmproxy: Interactively inspect, modify, and replay HTTP/HTTPS traffic to debug APIs and test security vulnerabilities in real time. 44.8k
bbot blacklanternsecurity/bbot: Automatically map and scan internet-exposed assets recursively to uncover hidden attack surfaces for red team operations. 10.5k
faraday infobyte/faraday: Centralize, correlate, and prioritize vulnerability findings from multiple scanners into a collaborative, actionable remediation workflow. 6.7k
wazuh wazuh/wazuh: Detect threats across endpoints and cloud environments with unified XDR and SIEM capabilities, including log analysis, file integrity monitoring, and active response. 16.6k
infisical Infisical/infisical: Securely manage, rotate, and audit secrets, certificates, and privileged access across teams and infrastructures with end-to-end encryption and RBAC. 28.9k
hosts StevenBlack/hosts: Block ads, trackers, malware, and unwanted content by deploying curated, extensible hosts files system-wide for enhanced privacy and security. 30.9k

Netdata v2.2.11.0 Adds Network Flow Monitoring and AI Alerts for DevOps Teams 🔗

New technical previews expand visibility into traffic patterns and cloud-native troubleshooting without central data collection.

netdata/netdata · Go · ▲ 11 in 1d Est. 2013 · Latest: v2.11.0

Netdata’s latest release introduces Network Monitor Dashboard and support for NetFlow, IPFIX, and sFlow in technical preview, enabling real-time traffic analysis across hybrid infrastructures. The update also enhances SNMP device coverage and adds a trap listener for legacy network gear.

On the observability front, Netdata Cloud now includes AI-powered troubleshooting via MCP (Model Context Protocol) and unified log ingestion from OpenTelemetry, correlating metrics, traces, and logs in a single interface. Endpoint monitoring extends to macOS, Windows, and FreeBSD alongside Linux, broadening its reach in heterogeneous environments. Built in Go, Netdata maintains its zero-configuration ethos—deploying in seconds with per-second granularity and minimal overhead, validated by a University of Amsterdam study showing superior energy efficiency in Dockerized workloads. The Prometheus collector overhaul improves metric labeling and scrape reliability for teams using it as a remote write target.
The catch: While powerful for real-time debugging, Netdata’s AI features rely on cloud connectivity for full functionality, which may conflict with air-gapped or strict data-locality requirements despite its distributed architecture.

Previously in The Times “covered” — Aug 16

Use Cases
  • DevOps teams monitoring Kubernetes traffic flow and application dependencies
  • SREs using AI to correlate logs and metrics during incident response
  • Infrastructure engineers tracking SNMP-enabled devices alongside cloud services

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

More Stories

Vercel Labs releases fx, a minimalist coding agent in Zig 🔗

The 7.8 MiB CLI tool enables model-agnostic AI-assisted coding with Unix shell ergonomics

vercel-labs/fx · Zig · 1.8k stars 1w old

Vercel Labs open-sourced fx, a coding agent harness and CLI written in Zig, designed for lightweight, embeddable AI-assisted development. At just 7.

8 MiB, it prioritizes minimalism and performance, offering a Unix shell-like interface rather than a heavy terminal IDE. Users interact via slash commands like /provider to switch between Vercel AI Gateway and OpenAI Codex, /model to list available models, and /fast to access priority inference tiers. Sessions persist locally, with tab titles reflecting workspace context and active model. Commands such as fx session resume restore prior work, while --prompt-permissions enables clean, headless approval flows. Transcript scrollback survives resize and reflow, and terminal attention bells alert multiplexers to waiting input.
The catch: As a v0.0.4 release with 64 open issues, fx remains early-stage and may lack stability for production-critical workflows.

Use Cases
  • Developers seeking lightweight AI-assisted coding in local terminals
  • Engineers embedding AI agents into larger toolchains or CI pipelines
  • Researchers testing model-agnostic LLM interactions with minimal overhead

Source: vercel-labs/fx — based on the README and release notes.

Rust-Powered CodeGraph Syncs Changes in Under a Second 🔗

Version 1.5.0 replaces parser with native Rust kernel for instant graph updates

colbymchenry/codegraph · C · 67.5k stars 7mo old

The Rust engine release rebuilds CodeGraph's parsing core, enabling live graph sync in under a second even on 27,000-file repositories. It adapts worker pools and caches to actual system resources, indexing the Linux kernel on a 2-core VPS in under 12 minutes — down from 26.

The graph remains byte-for-byte identical to the prior engine. The catch: 432 open issues suggest ongoing stability challenges despite performance gains.

Use Cases
  • Agents get real-time context for large codebases
  • Local indexing avoids token limits and tool calls
  • Cross-language projects maintain synchronized knowledge graphs

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

CC Switch v3.20.0 adds Pi management and fixes WSL path issues 🔗

Release includes database migration from v16 to v17 with automatic backup on upgrade

farion1231/cc-switch · Rust · ▲ 183 in 1d Est. 2025

The latest release of CC Switch integrates Pi as its ninth managed application, offering structured editing for providers, prompts, Skills, session browsing, and usage tracking without altering Pi’s native login or defaults. It enables Codex to bind multiple ChatGPT accounts securely, preventing cross-account billing, and restores Claude Code’s built-in WebSearch under GPT routing.

A critical WSL path bug affecting configuration updates is resolved. The release includes a database migration from v16 to v17, with automatic backups before upgrade and rollback requiring manual restoration. The catch: Open issues remain high at 2,325, indicating ongoing stability and feature maturity challenges despite rapid development.

Previously in The Times “covered” — Aug 14

Use Cases
  • Developers switch between Claude Code and Codex agents
  • Teams manage multiple AI provider accounts securely
  • Users reduce API costs via sponsored relay services

Source: farion1231/cc-switch — based on the README and release notes.

Quick Hits

bitcoin Bitcoin Core provides the reference implementation for secure, decentralized peer-to-peer transactions and blockchain validation. 90k
ragflow RAGFlow combines advanced retrieval-augmented generation with AI agent capabilities to deliver a powerful, context-aware foundation for LLM applications. 89k
rust Rust enables developers to write memory-safe, high-performance systems without sacrificing control or productivity. 115.8k
vaultwarden Vaultwarden offers a lightweight, Rust-based, Bitwarden-compatible password manager server for self-hosted secure credential storage. 65.7k
hugo Hugo is an ultra-fast static site generator that lets builders create responsive websites with minimal configuration and maximum speed. 89.5k

Tarnadas/smmdb enables cross-platform Super Mario Maker level sharing via Protocol Buffers 🔗

Last updated August 2026 with self-hosting docs and sanitized backup scripts for privacy

Tarnadas/smmdb · TypeScript · 76 stars Est. 2017

Tarnadas/smmdb solves the fragmentation of Super Mario Maker courses across Wii U, 3DS, and emulators like Cemu and Citra by converting levels into a shared Protocol Buffer format. Built with TypeScript, React, and Node, it lets users upload courses via a public API, then download them on any supported platform using a save file editor—bypassing Nintendo’s restricted 3DS download features.

The project relies on user uploads since it cannot access Nintendo’s servers, meaning only community-shared levels are available. Recent work focuses on self-hosting: developers can run their own instance using Docker, with environment variables for Google OAuth, MongoDB, and cookie secrets. A sanitization script (node scripts/sanitize-backup.js) removes personal data like emails and API keys before sharing backups.

The catch: SMMDB depends entirely on user-generated uploads, so coverage is sparse for niche or older levels, and emulator users still need legal game copies to play online features beyond level access.

Use Cases
  • Homebrew 3DS players download custom Wii U Maker levels
  • Cemu emulator users access community courses without Wii U hardware
  • Developers integrate SMMDB API to sync levels across tools and platforms

Source: Tarnadas/smmdb — based on the project README.

More Stories

ESP8266 Weather Station Delivers Lab-Grade Sensing with Self-Healing Air Quality Sensor 🔗

v1.0.4 adds ENS160 diagnostics, pure English DST, and full mechanical assembly guides for reliable DIY deployment

byte4geek/weatherstation · C++ · 122 stars 4w old

The byte4geek/weatherstation project transforms an ESP8266 into a high-precision environmental monitor, tracking temperature, humidity, pressure, air quality (TVOC/eCO2), luminosity, wind, and rain via a responsive web dashboard. It features dynamic sensor detection, customizable moving-average filters, Home Assistant MQTT Auto-Discovery, and hardware factory reset logic, all persisted in NVS flash.

The latest release introduces ENS160 air quality sensor diagnostics—including hotplate resistance checks and delta-threshold compensation writes—to reduce I2C traffic and verify sensor health without rebooting. Builders get single-click JSON backup/restore, Metric/Imperial toggling, and a mobile-optimized Compact UI that fits all metrics on one screen. The catch: The ENS160 sensor requires a 3-minute warm-up period, and the project’s reliance on specific hardware (like the ENS160 and compatible pressure sensors) may limit flexibility for builders using alternative sensor suites.

Use Cases
  • Homeowners monitoring backyard microclimate data for gardening
  • Educators teaching IoT environmental sensing in STEM classrooms
  • Makers integrating weather data into Home Assistant automations

Source: byte4geek/weatherstation — based on the README and release notes.

ESPectretre v2.8.0 refines Wi-Fi motion detection with hardened detection and ML improvements 🔗

Release switches traffic generator to ping and enables Hampel filtering by default for better reliability

francescopace/espectre · Python · ▲ 7 in 1d 9mo old

The ESPectre project released version 2.8.

0, updating its Wi-Fi spectre analysis motion detection system for ESP32 devices. Key changes include switching the default traffic generator from DNS to ICMP ping to avoid router rate limits and enabling Hampel outlier filtering by default. The ML detector saw reliability gains across ESP32-C6, S3, and C3 chips with retrained models and temperature-scaled scoring. Detection logic now uses a unified 12-subcarrier window and edge-driven binary publishing aligned between firmware and Micro-ESPectre. The catch: The ML detector remains experimental and requires manual asset selection, limiting plug-and-play use for advanced features.

Previously in The Times “covered” — Aug 16

Use Cases
  • Home automation enthusiasts tracking room occupancy without cameras
  • Renters installing motion sensing in apartments with no wiring changes
  • Developers testing privacy-preserving sensing for smart office environments

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

Ibex RISC-V Core Adds Bit-Manip Extension Support in Latest Update 🔗

Core now supports RISC-V B extension for embedded control and security applications

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

The lowRISC Ibex 32-bit RISC-V CPU core, written in SystemVerilog, has added support for the Bit Manipulation (B) extension in its latest commit. This enhances its suitability for embedded control tasks requiring efficient bit-level operations.

Ibex remains heavily parametrizable, supports RV32I/E/M/C extensions, and targets production use with extensive verification and multiple tape-outs. It is actively maintained, with the last push occurring today.
The catch: Verification status varies by configuration, with some setups still marked experimental or minimal, requiring users to assess maturity for their specific use case.

Use Cases
  • Embedded controllers implementing RISC-V with custom extensions
  • Security-focused IoT devices needing bit-manipulation acceleration
  • Academic projects studying parametrizable CPU core design and verification

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

Quick Hits

insect-detect Automates insect monitoring using camera trap data with AI-powered Python software for real-time detection and data collection. 66
ghw A Go library enabling robust hardware discovery and inspection for system-level device enumeration and diagnostics. 1.9k
automotive-skills-suite Provides 100+ installable Claude skills for automotive engineering standards, each paired with a confirmation reviewer for validated compliance workflows. 2.4k
alphabets Open-source mechanical split-flap display (Solari-style) enabling retro-modern information displays with physical actuation. 133
awesome-open-hardware Curated collection of tools, guides, and resources to support and accelerate open source hardware project development. 1k

PlayCanvas Engine Adds VSM Shadow Fix in Latest Patch Release 🔗

Patch v2.21.4 resolves out-of-range depth artifacts in virtual shadow maps for WebGL2 and WebGPU renderers

playcanvas/engine · JavaScript · ▲ 3 in 1d Est. 2014 · Latest: v2.21.4

The PlayCanvas engine team released v2.21.

4, addressing a visual artifact in variance shadow mapping (VSM) where out-of-range depth values caused incorrect shadow casting in complex scenes. The fix, contributed by @mvaligursky in pull request #9124, discards invalid depth samples during shadow map generation, improving stability for developers using high-precision lighting in WebGL2 and WebGPU-backed 3D applications.

This patch arrives amid ongoing work to unify the engine’s rendering pipeline across WebGL2 and WebGPU, with recent commits focusing on resource synchronization and shader compatibility. The engine continues to support glTF 2.0 asset streaming, 3D Gaussian Splatting, and WebXR for AR/VR, all accessible via TypeScript or JavaScript scripts. Local development requires Node.js 18+, with project scaffolding available through create-playcanvas.

Despite its maturity—over 12 years of active development—the engine maintains a high issue count, with 537 open issues signaling ongoing triage demands. While the core remains stable for web-based games and interactive visualizations, builders targeting cutting-edge GPU features may encounter gaps in experimental WebGPU extensions compared to native engines.

The catch: The engine’s abstraction layer, while simplifying cross-platform deployment, can limit low-level GPU tuning access for developers needing fine-grained control over compute-heavy workloads or custom rasterization techniques.

Previously in The Times “covered” — Aug 20

Use Cases
  • Build browser-based 3D games with physics and animation
  • Create AR/VR product visualizations using WebXR
  • Deploy interactive architectural walkthroughs with glTF streaming

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

More Stories

Phaser 4.2.1 fixes stencil rendering and ESM build issues 🔗

Patch resolves framebuffer clearing and ScaleManager resize bugs in HTML5 game framework

phaserjs/phaser · JavaScript · 40.2k stars Est. 2013

The latest Phaser release v4.2.

1 addresses critical rendering flaws, including incorrect stencil inversion and framebuffer clearing that broke visual effects. It also fixes ESM incompatibilities in CombineColorMatrix, ImageLight, and Texture modules, plus a ScaleManager resize failure when parent containers change. Built on 13 years of active development, Phaser supports WebGL and Canvas for 2D games deployable to web, mobile, and platforms like YouTube Playables and Discord Activities via third-party tools. The framework’s large 8 MB source file is mostly documentation; the minified build is 345 KB gzipped.
The catch: Despite frequent commits, 130 open issues suggest ongoing stability challenges in edge-case rendering and platform integrations.

Use Cases
  • Build HTML5 games for desktop and mobile browsers
  • Create Discord Activities and YouTube Playables
  • Deploy games to iOS, Android, or Steam via wrappers

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

Quick Hits

Pumpkin Pumpkin-MC/Pumpkin (Rust): Run blazing-fast, lightweight Minecraft servers with minimal setup — ideal for developers seeking performance and simplicity. 10.8k
WickedEngine turanszkij/WickedEngine (C++): A modern, feature-rich 3D engine delivering cutting-edge graphics and flexibility for high-fidelity game development. 7.2k
godot godotengine/godot (C++): Build and deploy 2D and 3D games across all platforms with a free, open-source engine designed for creator freedom. 115.9k
phantom-camera ramokz/phantom-camera (GDScript): Add cinematic, smooth camera controls to Godot 4 projects — inspired by Cinemachine, with zero setup hassle. 3.5k
renpy renpy/renpy (Ren'Py): Create immersive visual novels with ease — a powerful, scripting-driven engine trusted by indie and professional storytellers. 6.7k
godot-ai hi-godot/godot-ai (GDScript): Integrate production-grade AI tools and MCP servers into Godot with a snap-install, free, and fun-to-use toolkit. 1.8k
MonoGame MonoGame/MonoGame (C#): Develop cross-platform games using a mature, flexible framework that lets you write once and deploy everywhere. 14.3k
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