agent-memory is a local-first, agent-agnostic long-term memory runtime for AI agents. It stores memories as plain Markdown files, with a rebuildable SQLite index used as a cache rather than the source of truth. Conversation-boundary hooks trigger writes, while an independent sleep-time Manage pass consolidates, ages, and forgets memories by value; unattended deletion is limited to proposals that require confirmation, and superseded or archived material remains available. The runtime provides three recall paths: deterministic MEMORY.md injection at session start, BM25 search with progressive disclosure, and direct access through the store's directory tree using tools such as ls and grep. Memory records use frontmatter for names, abstracts, status, timestamps, links, weights, and provenance, with free-form Markdown bodies. Its CLI, MCP server, and host-agent hooks share the same validation, hash-diff, and reindexing path, and it can be used by agents that run shell commands, including Claude Code and Codex CLI. The repository documents a Python 3.12-or-later setup using uv, commands for initialization, recording, recall, rebuilding, sleep-time consolidation, and proposal approval, plus host setup for Claude Code and Codex. It does not include an LLM client; reasoning is delegated to the host agent's CLI, so the library requires no separate model keys or billing surface.
ai-evaluation-framework is an open-source framework by Dreamers Inc. for benchmarking model-based solutions by accuracy, p95 latency, and estimated cost against user-defined ground-truth cases. Each task specifies fields and a scoring rule per field, including exact identifiers, normalized dates, numeric tolerances, containment, token overlap, and expected-missing values; cases can also provide alternate accepted renderings. It runs models from OpenAI, Anthropic, or an OpenAI-compatible gateway, compares results in a table, emits detailed JSON reports, and caches responses by model, task, and case so scoring rules can be changed without another model call. The repository uses document-field extraction as its worked example but accepts arbitrary text inputs such as OCR output, transcripts, and PDF text. A separate feedback server collects thumbs-up or thumbs-down responses with required comments, stores append-only JSONL data, and exposes statistics, exports, and candidate cases for human review; it does not train models. The repository includes tests using scripted mock models and is licensed under Apache-2.0.
Anti Slop is an open-source set of rules and agent skills for AI coding agents that filters generic AI-generated user interfaces, copy, and code without prescribing colors, fonts, layouts, or other visual styles. Its core contains 38 mandatory rules across Hard Gate, Purpose-Gate, and Quality Locks tiers; a Liveliness Toolkit with ENERGY, RHYTHM, and MOTION controls; a Design Read; and a mandatory four-part PASS/FAIL Delivery Gate. Task-specific skills are organized as separate SKILL.md folders and loaded additively, while a project’s DESIGN.md supplies the visual direction. The repository provides installation through the npx antislop-ai picker, the skills.sh-compatible npx skills add command, or a Claude Code marketplace plugin, with support for agents including Claude Code, Codex, Antigravity, OpenCode, Cursor, Gemini CLI, and Hermes.
Blender is a free and open-source 3D creation suite developed by the Blender Foundation. It provides tools for modeling, rigging, animation, simulation, rendering, compositing, motion tracking, video editing and game creation. Blender is distributed under the GNU General Public License.
Blip is a Linux iMessage client for the Omarchy bar and desktop, developed by Fred Nix and contributors. It uses an existing Mac signed into Messages as a gateway: Linux tools communicate with Mac-side bridge tools over SSH, which read the Mac's chat database and use Messages.app automation to send messages. The Linux client renders conversations, group chats, unread indicators, notifications, search, link previews, photos and files, while a Mac-side watcher provides near-real-time updates. Blip requires an awake, reachable Mac running macOS 13 or newer, with Messages in iCloud, SSH key authentication, and the required macOS permissions. It does not use a hosted server, telemetry, or a Linux-side message cache; message text and attachment data travel through the SSH connection, although opened attachments and link previews may be cached locally. It is distributed as an Omarchy/Quickshell plugin with TypeScript, QML, and Mac bridge tools under the MIT license.
BoardUI is a React design system for agentic interfaces and dashboards, combining AI-product components such as chat, thinking indicators, streaming agent logs, task lists, web-search trails, composers, and an agent runtime with dashboard components such as tables, charts, forms, cards, navigation, authentication, and design tokens. Its CLI copies individual components and their dependencies, or the complete free catalog, into a Next.js project as source files rather than installing a runtime dependency, so developers can modify the code directly. The system uses React Aria Components for accessible controls, Tailwind CSS v4, semantic design tokens, light and dark theme classes without runtime CSS, and Figma-based typography and styling rules. The repository includes a working AI chat application with a streaming chat endpoint and an agent runtime that accepts provider keys for supported services or an OpenAI-compatible server, reads keys server-side, and streams replies to the chat UI. It also provides an MCP server for browsing and installing components, along with agent rules specifying design tokens, type scale, and conventions.
Camera to Blender is an open-source workflow that turns a phone or webcam photo of a real object into a 3D model and imports it into Blender. Its pipeline sends the image through optional Gemini background removal, submits it to the Tripo3D API for model generation, and relays the result to Blender through a WebSocket add-on. The project includes a phone-oriented web camera app, a Python relay server, and a Blender add-on. The server runs on the computer with Blender; phone access uses an HTTPS ngrok URL, while laptop webcam use can run locally. It requires Python 3.10+, Blender 3.0+, and a Tripo3D API key; a Gemini API key is optional. The project is distributed under the MIT license.
chippytea is a native macOS menu-bar application for finding and removing storage clutter, including app and developer caches, logs, Xcode data, Python bytecode, build folders, downloads, and large personal files. It runs locally without an account, telemetry, or automatic deletion, and uses a Rust engine with a SwiftUI interface to scan selected locations, maintain a SQLite index, estimate sizes, perform safety and ownership checks, and show the consequences before cleanup. Users can move review-only files to Trash or permanently remove eligible developer caches, build files, and dependencies; its bounded duplicate check rechecks selected files before moving one copy to Trash. The app supports macOS 14 and later, requires an Apple Silicon Mac for local builds, and is MIT licensed.
Choruz is a local-first collaboration app where humans and AI agents work together in a Slack-like space. Each agent runs a real CLI in its own workspace, preserving that CLI's models, tools, and session capabilities while allowing work to be handed to people or other agents through direct messages, groups, mentions, threads, tasks, and files. It provides isolated company and agent workspaces, dedicated directories or Git worktrees, an integrated terminal, file browser and editor, SSH runtime hosts, and browser-based remote control. It supports Claude Code, Codex, Pi, Grok, OpenCode, and webhook-driven external agents, with REST APIs, WebSocket synchronization, webhook agents, Slack and Telegram bridges, and optional plugins. Choruz is in pre-release development and requires Rust, Node.js, pnpm, PostgreSQL, and at least one supported agent CLI when run from source. Its source code and software documentation are licensed under the MIT License; visual assets have separate licensing records.
A utility for combining conditional class names. The video contrasts it with `cn`, which combines class-name handling with Tailwind Merge.
cn is a JavaScript engine for conditionally joining Tailwind CSS class names and resolving conflicting utilities. It is a drop-in replacement for clsx and tailwind-merge, preserving their APIs and output semantics while learning repeated variadic call sequences so stable calls can skip redundant work. The package has no dependencies and is framework-agnostic, running in browsers, Node, Bun, Deno, edge runtimes, and server templates for frameworks including React, Vue, Svelte, Solid, and Astro. It supports configurable Tailwind class groups, prefixes, themes, and validators through cn/config, and includes a CLI for build-time processing and a lighter strings-only variant. It is maintained by aidenybai and shadcn.
Codenotch is a macOS app that pins a small notch to a screen edge and displays usage limits, reset windows, and activity states for coding assistants including Claude Code, Cursor, Codex, Antigravity, and GLM. It can show whether a provider is working, finished, or waiting for user input, with separate rings for separate Claude Code profiles and live session details on hover. Each provider adapter reads credentials, sessions, logs, databases, language-server RPCs, or provider endpoints already used by the corresponding local tool; Codenotch does not sign in to services itself. A UsageStore polls the adapters, preserves the last good reading across launches, labels each reading by fidelity, and exposes failures as statuses such as stale, needsAuth, or error rather than inventing a percentage. The notch can be placed on any screen edge, expands when approached, and can be configured to show or hide its Dock and menu-bar icons. The app is written for macOS, supports a demo mode with fixed sample data, and uses Sparkle for signed background updates. Its README notes that the provider interfaces may change because some readings come from internal endpoints or local application state. The project is licensed under the MIT License.
Commerce Agents is Anthropic's reference implementation for building shopping agents for customers and merchant agents for back-office staff with Claude. It defines each agent through prompts, skills, tool contracts, grounding, memory, approval gates, and backend interfaces, and provides runnable examples for retail, travel, telecom, and entertainment. The shopping agent searches and compares products, plans purchases, fills carts, answers order and policy questions, and maintains customer memory. The merchant agent analyzes performance, manages listings and inventory alerts, handles pricing and promotions, and drafts campaigns; its writes are staged until a host approves them. The agents can run through the Anthropic Messages API, Claude Agent SDK, or Managed Agents, and the repository provides a Claude Code plugin for scaffolding or reviewing commerce agents.
Fable51 Worlds is an open-source, AI-assisted world-generation project that turns a text brief, photograph, or video into a walkable browser scene implemented as a pure Three.js application. Claude Fable 5.1 agent swarms research the location, collecting map geometry, elevation, transit, street information, and storefront data; generation scripts produce procedural façades, street furniture, vehicles, vegetation, fixtures, and other assets; and the runtime assembles terrain, streets, buildings, props, crowds, and traffic from JSON specifications without a game engine, proprietary 3D tiles, or downloaded meshes. Verification uses Playwright to drive the application, capture fixed viewpoints, and compare them with photographs from corresponding locations. Independent reviewer agents covering architecture, geography, technical art, and interaction produce reports for subsequent fix cycles, and each world includes a quality-assurance report.
Fable orchestrator is a small, local-first routing skill for Codex that separates planning and adjudication from code implementation. Fable 5.1 plans tasks and makes final decisions without writing code or owning the workspace; Codex validates the bounded graph it returns, runs ready implementation nodes in parallel when useful, collects evidence, verifies the result, and requests further adjudication when needed. Implementation work is delegated through callable OpenCode Go agents, with GPT-5.6 Luna assigned to normal implementation and DeepSeek V4 Flash to loops and repeated high-throughput execution. The repository contains an installable skill, agent configuration, an invocation script, an installer, and shell-based tests. It does not provide a proxy, dashboard, model catalog, credential store, or API-key management. The installer supports dry runs, copying into a Codex skills direct
A research artifact containing a complete machine-checked Lean 4 proof of Fermat's Last Theorem, following the Frey–Serre–Ribet–Wiles and Taylor–Wiles argument and built on Mathlib. The theorem states that for natural numbers n, a, b, and c with n ≥ 3 and a, b, and c positive, a^n + b^n ≠ c^n. Lean's kernel checks the proof, which uses only Lean's three standard axioms—propext, Classical.choice, and Quot.sound—and the repository excludes sorry statements, added axioms, native_decide, unsafe code, and related escape mechanisms. A comparator independently replays the environment against a Mathlib formulation, while Nanoda, an independent Lean kernel implemented in Rust, checks an exported version of the environment. The repository also includes PROOF-PATH.md and an offline HTML presentation of the proof's theorems, definitions, dependencies, and generated documentation. It is released under the Apache License 2.0 and is described as not maintained and not accepting contributions.
ffmpeg-skill is an agent skill that gives Claude Code, Cursor, Codex, and other agents a local video- and audio-editing workflow built on FFmpeg and ffprobe. It uses a probe → edit, preferring stream copying where possible → check → verify process, with typed Python scripts rather than shell strings. Its 21 tools cover cutting, joining, silence removal, duration and aspect-ratio fitting, captions, overlays, graphics, HDR-to-SDR conversion, LUTs, audio cleanup, loudness, synchronization with drift correction, multicamera editing, delivery checks, project rendering, and batch processing. Each tool supports structured results and dry runs, and the machine-readable contract generates an MCP interface whose tool definitions are derived from the scripts. The skill runs locally without cloud services, API keys, or Python dependencies beyond the standard library; it requires FFmpeg 5.0 or later. It is a tool for local, agent-controlled workflows that probe, edit, verify, and render video.
Flea is a keyboard-first graphical file manager for Omarchy on Linux, built as a Quickshell front end over a Rust backend. It keeps the full directory in the backend while sending the interface only the rows that fit on screen; thumbnails are requested only for visible rows after the listing stops moving and are sandboxed with bubblewrap and prlimit. It provides list, Miller-columns, and thumbnail-grid views; previews for text, images, video, audio, PDFs, and archive contents; directory tabs; path navigation; filtering and subtree search; and an undo journal for operations such as copying, moving to trash, renaming, compression, extraction, and conversion. The application supports local and removable volumes, SMB, SFTP, FTPS, WebDAV, and NFS mounts through GIO, plus Taildrop sharing and Dropbox as a destination. It can register itself as Omarchy's default file manager and as the desktop file chooser through the XDG file-chooser portal. The repository documents installation through Omarchy's package manager or an AUR package, optional archive, image-conversion, and Taildrop dependencies, a newline-delimited JSON protocol between the UI and backend, and an MIT license.
FrontierHarness Eval is an open evaluation benchmark and repository for measuring how coding-agent harness configurations affect software-engineering task results when the underlying model and runtime are held constant. It publishes benchmark definitions, task instructions and metadata, harness-version metadata, normalized aggregate and task-level results, and an agent-neutral workflow for evaluating additional harnesses. The benchmark runs the same model through multiple harness configurations on 30 tasks, using a frozen golden checkpoint and a fresh restore for each task. Its scripts provision the environment, run trials, record verifier-based pass/fail outcomes, measure cost, cache behavior, and speed, then normalize results and generate charts and reports. The repository also provides a CLI and skill-based workflow for evaluating a third-party harness, with infrastructure failures marked separately from task failures. The project focuses on terminal-based software-engineering tasks and states that its results may not generalize to other knowledge-work domains. The repository is published under the frontier-harness-eval organization, with isolated runtimes and checkpoint restores provided by Runta.
Gemini is Google's AI model and assistant accessible through gemini.google.com. It processes mixed media including video, audio, images, and text, and can convert recordings into SOPs, summarize meetings, extract decisions, and draft follow-up emails. In research and development contexts, it serves as an alternative LLM backend for tasks such as adversarial audits of requirements and resume evaluation.
gpuix-svelte is an experimental Svelte custom renderer for GPUI, Zed's GPU-accelerated UI framework, distributed through @gpuix/native. It renders ordinary Svelte components into native desktop windows without a web view, while retaining Svelte component syntax, reactive state, scoped styles, and hot reload. The project includes demos for a counter, tic-tac-toe, Hacker News, and a liquid-glass control center, plus a browser target that renders the same components into a WebGPU canvas through GPUI's WebAssembly build. Its styling layer parses supported inline styles and class rules into GPUI's layout and paint properties rather than using a browser CSS engine. Supported behavior includes flex and grid layout, logical-pixel sizing, colors, typography, scrolling, hover and active states, CSS variables, portals, anchored overlays, keyboard and focus helpers, and custom components such as Scroller. The package also provides headless testing utilities that exercise GPUI's rendering, hit testing, input, and event pipeline. The repository is work in progress and requires an unreleased Svelte custom-renderer build bundled with the package. It supports Node.js and Bun, includes prebuilt native binaries, can compile examples into standalone executables for macOS, Linux, and Windows, and is licensed under MIT.
Human Atlas is an open-source, browser-based interactive 3D anatomy explorer. It uses selectable BodyParts3D meshes to let users toggle anatomical system layers, search structures, and view exploded arrangements that show how body parts fit together. The video describes it as running locally as a static site without accounts or API keys.
J-Space Cognition Suite is a model-agnostic, inference-time control system packaged as a cross-platform AI-agent Skill for deep reasoning, long-horizon work, tool use, verification, and recovery. It leaves model weights and training unchanged, organizing an agent's working representations through one entry point and nine selectively loaded modules supported by four references. Its fast, full, and loop operating modes use selective workspace loading, a shared broadcast hub for constraints and values, dense internal reasoning traces, explicit intermediate steps before conclusions, metacognitive routing of confidence and failure signals, and bounded empirical verification. An optional standard-library controller records durable task state, including goals, next actions, checkpoints, open questions, seams, and recovery state. The suite is intended for low-friction integration with AI hosts that support Skills or equivalent system/developer-instruction mechanisms.
Kitter is a local-first skill manager for agent skills across project folders. It keeps one canonical copy of each skill in a shared library, links skills into individual projects according to their needs, reports which skills are loaded, and estimates their context cost.
kugiri is a text-preparation library for animation that splits text into lines, words, and characters exactly where the browser has already wrapped it. It measures browser wrapping to preserve layouts such as balanced headings and vertical text while exposing individual text elements to animation tools.
Lean 4 is a programming language and theorem prover used to formalize mathematical proofs, including proofs generated by AI systems. Its project repository provides the language implementation along with theorem-proving and functional-programming tutorials, reference documentation, examples, installation guidance, and instructions for building from source.
M3E Canvas is a browser-based Material 3 Expressive design tool that turns interface sketches into prompts for AI coding tools. It provides drag-and-drop components, multiple phone and desktop screens, connected tap and swipe navigation, layers and groups, theme controls, previews, and PNG export. Designs can be converted into prompts in English, Japanese, Chinese, or Korean for Android or web targets, including behavior notes for individual components. An optional bring-your-own-key AI helper can draft behavior notes and screen descriptions by sending requests directly from the browser to OpenAI, Claude, Gemini, or DeepSeek. The app is a static Next.js export, saves work in browser localStorage, and is free and MIT-licensed.
Model Context Protocol (MCP) is an open, standardized protocol layer for connecting large language models and AI agents with external data sources, hosted infrastructure tools, and other contextual data. It defines formats, metadata, protocol schemas, and APIs for sharing information and tools, attaching, referencing, and validating context such as documents, embeddings, and provenance, while supporting scoped authentication and permissions. MCP translates JSON requests from an agent into calls to service APIs, including CRM, container-management, and GKE capabilities, and can provide design-system context and related tools for generating consistent applications. The official project publishes its specification, documentation, and protocol schema; the schema is defined first in TypeScript and also provided as JSON Schema for broader compatibility. The protocol was created by David Soria Parra and Justin Spahr-Summers, is hosted at modelcontextprotocol.io, and is licensed under the MIT License.
Nanoda is a Rust-based tool that independently checks formal proof declarations in a Rust implementation.
NoGraphicsAPI is a minimal graphics API built on modern Vulkan extensions, developed by sebbbi. It lets shaders follow GPU pointers directly while the application manages texture and sampler descriptors in GPU memory, reducing traditional binding and synchronization machinery. The repository presents it as an implementation closely related to the author's "No Graphics API" blog post and SIGGRAPH talk.
OrcaReplay is an open-source replay and debugging tool for AI-agent executions, built by the OrcaRouter.ai team. It records an agent run as a local trace, including model requests and responses, tool calls and results, shell exit codes and timing, filesystem snapshots, MCP traffic, and selected network activity. It runs as a local proxy and capture layer around an unmodified agent. A proxy records model traffic, while PATH shims, JSON-RPC interception, filesystem snapshots, fetch hooks, and optional TLS interception capture effects that model transcripts cannot show. The resulting timeline can be viewed, exported, queried through JSON or MCP, and analyzed with recorded versus inferred causal edges. The same recorded stream supports offline replay with the network blocked, forking from a derived checkpoint onto another model, and comparison of several models with an optional verification command. Replay uses the recorded conversation and workspace state; interactive terminal sessions can be approximate because prompts and interactive-only tools may not exist on the wire. The CLI is installed with npm, requires Node 20 or newer, stores runs under `.orca/runs/`, and the code, CLI, viewer, adapters, and trace format implementation are released under Apache-2.0; the trace specification is CC BY 4.0. Traces are local and redacted on write, but the project describes them as sensitive and treats redaction as best-effort.
Pictaria Server is a self-hosted photo-intelligence and curation server for Immich libraries, developed by Pictaria. It provides a private web dashboard for collection insights, optional AI enrichment, human photo review, and Smart Albums that synchronize saved Immich searches on a schedule. Its curation workflow groups photos from the same moment into similarity stacks and suggests which photos to keep using frame-worthiness scores; an optional AI referee can compare grouped photos, while human decisions take precedence. Enrichment classifies selected images with operator-hosted Ollama, LM Studio, llama.cpp, or other OpenAI-compatible endpoints, as well as supported cloud providers. Smart Albums can apply saved searches and a Best of mode that ranks results using enrichment data, curation decisions, and photo scores. The server can run as a Docker container or directly on Node.
React is a JavaScript library for building web and native user interfaces. It uses declarative views that update the appropriate components when application data changes, and lets developers compose encapsulated, state-managing components into larger interfaces. React code can use JSX, an HTML-like syntax for defining interface elements, although JSX is optional. The library is designed for gradual adoption, can render on the server with Node, and supports mobile applications through React Native. Its core repository is developed openly on GitHub under the MIT license.
Reef is open-source continual-learning infrastructure for self-improving AI agents, developed by Human-Agent-Society. It connects agent inference, interaction records, feedback, learning jobs, evaluation, and versioned delivery, supporting updates to model weights as well as agent harnesses such as prompts, rules, and skills. Each learning cycle has four stages: Reef serves requests and records interactions; matches later scores or structured feedback to those records; produces a candidate update from eligible records; and evaluates the candidate against a configured selection policy before publishing it. Rejected candidates leave the previous release serving, while accepted updates are committed and delivered without restarting the service. The project provides an HTTP API with OpenAI- and Anthropic-compatible inference endpoints, feedback reporting, scenario-based releases, and integrations for training or inference components including Slime and SGLang. It can be installed from PyPI as `reef-infra` or from source; artifact and checkpoint management requires Git LFS.
Reverify is an AI-agent verification toolkit for reverse engineering and related code analysis. It places deterministic tools between an agent and its claims: the model proposes hypotheses about a binary or candidate implementation, while parsers, disassemblers, pattern scanners, emulators, equivalence checks, and optional analysis engines compare them with ground truth and return evidence-backed VERIFIED, REFUTED, INCONCLUSIVE, OBSERVED, or INVALIDATED results. Its pure-Python core handles PE, ELF, and Mach-O parsing, x86/x64 and ARM/ARM64 disassembly, byte and pattern inspection, CPU micro-emulation, Protobuf/TLV dissection, and Frida hook generation; optional installations add Capstone, Unicorn, LIEF, Z3, or angr for deeper analysis. The project exposes the verification loop through a command-line interface and an MCP server that agents such as Claude Code and Cursor can call. It also records verified, observed, proved, and refuted results in a content-keyed local ledger, allowing grounded facts and negative findings to survive context resets, compaction, and new sessions. A rollover and orchestration system hands sessions off through files rather than model-written summaries, keeping verified facts separate from unverified notes. Reverify includes claim types for bytes, typed reads, instructions, patterns, emulation, behavioral or formal equivalence, imports, exports, sections, and—when angr is installed—functions, calls, references, and reachability. It is distributed from PyPI and licensed under the MIT License. The repository states that it is intended for authorized reverse engineering, malware analysis, CTF work, interoperability research, and software the user owns or is permitted to analyze.
Rust is a systems programming language and toolchain developed by the Rust project. Its compiler, standard library, documentation, package manager and build tool Cargo, formatter rustfmt, linter Clippy, and editor support are maintained in the main repository. The language uses a rich type system and ownership model to enforce memory and thread safety at compile time, while targeting fast, memory-efficient software for critical services, embedded devices, and integration with other languages. Rust is primarily distributed under the MIT and Apache 2.0 licenses, with portions under BSD-like licenses.
SlopMonster is a Python linter for detecting formulaic or suspicious AI-generated writing in HTML pages, Markdown, plain text, landing pages, READMEs, emails, and scripts. Its scorer checks AI-associated vocabulary, sentence constructions, punctuation cadence, rule-of-three rhythms, and unsupported sales claims, assigns a score out of 5, and can fail a build when copy scores below 5/5. Its workflow scores the draft, performs a three-pass rewrite that removes suspect vocabulary and sentence shapes and replaces them with specific copy, sends the draft to a different model family for cleansing, and scores the result again. The cleansing script can use a Codex or Claude CLI, refuses to route a draft back to its own model family, and prints a prompt when no supported rival CLI is available. A GitHub Actions workflow is included as a reusable build gate; the scorer itself uses only Python's standard library, while the cleansing step requires an external AI CLI or manual prompt handling. The repository also includes regression tests, a catalogue of detection rules, rewrite principles, and worked before-and-after examples. It is distributed under the MIT license and can also be installed as an agent skill for Claude Code or used by other agents through its plain-Markdown instructions.
SQLite is a self-contained, serverless, zero-configuration, transactional SQL database engine implemented as a C library. It was created and is primarily maintained by D. Richard Hipp and the SQLite development team and is widely used as an embedded database in applications, devices, and browsers.
stop-stutter is a macOS utility for reducing game-streaming stutter associated with Apple's peer-to-peer Wi-Fi interface, known as AWDL. It disables the interface while selected streaming applications run and restores it afterward without disconnecting regular Wi-Fi. The repository lists support for Moonlight, GeForce NOW, Punktfunk, Parsec, and Steam Link, and describes a native SwiftUI interface with Liquid Glass. It is distributed under the MIT license.
superlocal is a desktop-focused, provider-agnostic email client and provider gateway for combining multiple mailboxes into one keyboard-friendly inbox. It keeps provider credentials and capabilities separate, supports Gmail and inbound connections, and uses local encrypted storage.
tailwind-merge is a utility for combining Tailwind CSS classes while resolving conflicts between them. The video describes it in the context of using `cn` instead of combining it with `clsx`.
Three.js is an open-source JavaScript 3D library developed in the mrdoob/three.js project for assembling and rendering procedural scenes in browser applications, including Fable 51 Worlds, Human Atlas, browser FPS worlds, CAD workspaces, and physics games. It creates scenes containing cameras, geometry, materials, meshes, and animations, with WebGL and WebGPU renderers and SVG and CSS3D renderers available as addons. Its official site and repository provide documentation, examples, tools, an editor, community resources, and a JavaScript module distribution.
A curated map of modern time-series forecasting architectures, covering classical baselines, linear models, transformers, diffusion models, and pretrained forecasters. It explains the mechanisms and trade-offs of representative approaches and provides compact, runnable PyTorch implementations, with a short README for each architecture.
Tripo AI is an online AI 3D-model generator that converts text prompts, photographs, images, and sketches into 3D models. The video presents it in a camera-to-Blender workflow for generating a model from a photograph. Its generation modes include detailed geometry and production-oriented quad meshes. The service also provides automatic part segmentation for 3D printing, exports STL, 3MF, and OBJ files, automatic character and creature rigging, and PBR texture generation.
Try Omarchy for Windows is a Windows launcher maintained by Omacom that runs the full Omarchy Linux desktop in a window without dual booting, VMware, or VirtualBox. It uses QEMU on the Windows Hypervisor Platform (WHPX), boots a prebuilt x86_64 Arch image with Omarchy, and renders the desktop through virtio-gpu, virgl, and Venus Vulkan via WINQ-EMU, with automatic CPU-rendering fallback. The console-less launcher enables Windows virtualization, downloads and SHA256-verifies the GPU runtime and guest image, creates and supervises the virtual machine, and stores its data in a chosen local folder. It provides GPU or CPU rendering, clipboard sharing, an optional shared Windows folder, port forwarding, backups, diagnostics, updates, and optional Start-menu and Desktop shortcuts. The project also includes reproducible guest-image build tooling, QMP controls, PowerShell scripts, and an experimental offline portable mode. The application is built in Go for Windows; the repository's scripts and documentation are MIT-licensed, while Omarchy and the guest image contents have separate licenses.
TypeScript is a programming language developed by Microsoft as a superset of JavaScript. It adds optional static types for application-scale JavaScript and compiles to readable, standards-based JavaScript for browsers, hosts, and operating systems. The compiler is distributed through npm, with stable and nightly builds, and the project provides a browser-based playground, documentation, roadmap, and community resources.
UNREEL is a personal AI video streaming service that generates shows while they are watched instead of serving prerecorded episodes. A showrunner LLM writes shots incrementally, and MiniMax H3 Max Turbo renders them through fal; each story shot uses the previous shot’s last frame as its image-to-video input, preserving visual continuity as the stream advances. The player maintains a render buffer and swaps to the next clip when the current one ends, while captions display dialogue written verbatim by the showrunner. The application is built with Next.js, React, and TypeScript, and uses Nano Banana 2 for key art and Gemini 2.5 Flash through fal’s LLM router for the showrunner. It includes chained story films and hard-cut “chaos” channels, and requires a fal API key for generation.
Place deterministic verification tools between an AI agent and claims it makes about a binary. The system checks proposed offsets, instruction patterns, or behaviors against actual bytes, returns verified, refuted, or inconclusive with evidence, and stores grounded facts and reputations in a local ledger.
Build cloud-powered assistants that help customers shop and help staff operate a retail back office. Use shared prompts, skills, tool contracts, and approval gates across multiple runtimes, while keeping merchant changes staged until a person approves them.
Provide a local video-editing workflow that lets coding agents probe footage, perform edits, verify the result, and rerender changes without sending footage to the cloud.
Searchable transcript of GitHub Trending Today #48: reverify, keslr_connect, anti-slop, boardui, ffmpeg-skill, chippytea — Github Awesome (14:48). Search for a phrase, then click its timestamp to jump straight to that moment in the video.
Captions sourced from the original video on YouTube, published by Github Awesome. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.
00:00 Welcome back to GitHub awesome. This is GitHub trending today number 48. 35 trending open- source projects on GitHub right now. Let's get into it. Reverify puts deterministic tools between an AI agent and any claim it makes about a binary. The model proposes an offset instruction pattern or behavior. The verifier checks the actual bites and returns verified, refuted, or inconclusive with evidence.
00:26 It also stores grounded facts and reputations in a local ledger that survives context resets. It is a dualuse reverse engineering toolkit, so use it only on binaries you own or have permission to analyze. Kesler Connect lets your app recognize verified humans without building another signup maze. Its TypeScript packages handle open ID connect, express guards, and network identity lookups.
00:50 So services on Kesler can identify the member behind a device or require an explicit login when consent matters. The security defaults fail closed, including fixed RS 256 and mandatory PKCE. Anti-slop is a rules document that stops AI coding agents from shipping the same generic UI, gradients, glass morphism, marketing buzzwords. It's not a style guide.
01:16 It doesn't tell the agent what colors or fonts to use. It filters out slop through 36 rules and a checklist with one verification question per rule. A pointer line in your agents.mmd sends the agent to those rules when UI work is relevant across clawed code or codecs. Board UI gives React developers a design system built for AI products with chat, thinking states, agent logs, and the dashboard pieces around them.
01:42 Its CLI copies components directly into your project, so you own the source instead of depending on a blackbox package. The free repository includes 62 items, React area accessibility, and a working bring your own key chat app. This repo is a complete machine checked proof of Firmat's last theorem in lean 4 with zero shortcuts, no sorry statements, no extra axioms, just Lean's three standard ones.
02:08 It's independently verified twice, once by Lean's own kernel and once by Nanoda, a Rust implementation checking 152,000 declarations across 60,475 modules with zero errors. It's Anthropic's own artifact unmaintained. FFmpeg Skill gives coding agents a practical video editor without sending footage to the cloud. It teaches a probe, edit, verify workflow and ships local scripts for cuts, captions, audio cleanup, color conversion, multicam sync, and delivery checks.
02:40 You can describe a complete edit in project.json, rerender after changes, or expose the tools through MCP. Tailwind class conflicts usually mean combining CLSX with Tailwind merge. CN replaces both through the same familiar APIs, so existing projects can migrate without rewriting every component. It has zero dependencies, works across frameworks and runtimes, and supports custom Tailwind configurations.
03:07 The maintainers report a 30 time speed up for common component calls backed by reproducible benchmarks. Anthropics Commerce Agents is a reference blueprint for building two kinds of cloudpowered assistants. one that helps customers shop and another that helps staff run the back office. The same prompts, skills, tool contracts, and approval gates work across three run times with runnable examples for retail, travel, telecom, and entertainment.
03:37 Merchant changes stay staged until a person approves them. Chippy finds the storage clutter developers tend to forget old build folders, package caches, Xcode data, downloads, and large files. The native Mac OS menu bar app is built with Swift UI and Rust, runs entirely locally, and sends no telemetry. It explains each cleanup candidate before you choose trash or permanent deletion, then tracks credited space with a playful chip counter.
04:07 JSpace Cognition Suite is an agent skill that manages how models handle long toolheavy tasks without changing their weights. It selects lightweight or persistent workflows, records durable task state, and adds checkpoints for verification and recovery. The optional Python controller keeps that state inside the project while modules load only when needed.
04:30 Try Omari for Windows runs the full Omari Linux desktop in a window without dual booting or setting up Virtual Box. A small Go launcher configures queuing on Windows hypervisor platform, downloads a pre-built Arch image, and uses your GPU through Virgel and Venus Vulcan. When supported, your virtual disc runtime and settings stay inside one removable folder.
04:53 M3 canvas turns interface ideas into something an AI coding tool can actually follow. You drag Material 3 expressive components onto phone or desktop screens, connect interactions, preview taps and swipes, then export the design as a structured prompt for Android or the web. Themes, layouts, and behavior notes travel with it, and everything saves in your browser.
05:17 Blip puts iMessage inside the Omari bar on Linux, while a Mac you already own handles Apple's side of the conversation. The Linux client reads and sends messages over a dedicated SSH connection with group chats, attachments, search notifications, and roughly two-c message updates. There's no hosted server, account, or Linux message cache. Reef is infrastructure for agents that improve from real feedback while they keep serving requests.
05:44 It records each interaction, matches later scores or comments to that run, produces a candidate update, evaluates it, and publishes only the version that passes your policy. The update can change model weights or the agents prompts, rules, and skills. Slot Monster treats awkward AI copy like a code quality problem. Its Python llinter scans pages, markdown or plain text for familiar vocabulary, formulaic sentence shapes, punctuation ticks, repetitive rhythms, and suspicious proof claims.
06:17 A rewrite workflow then cleans the draft with a different model family and scores it again, while a GitHub actions check can block weak copy from shipping. Fable 51 Worlds turns a text, image, or video brief into a walkable place built entirely with code. Agent swarms research the location, generate procedural assets, assemble everything in 3JS, then compare fixed camera views against real photographs.
06:42 Because the result is source code rather than a baked video, you can inspect or edit individual pieces without regenerating the whole scene. Flea is a keyboard first Linux file manager built specifically for Omari. A Rust backend keeps the full directory while the quick shell interface renders only the rows and thumbnails you can see which makes huge folders more manageable.
07:06 It includes grid and Miller column views, media previews, tabs, remote mounts, and an undo journal for file operations. GPUselt lets you build GPU rendered desktop interfaces with ordinary spelt components, but without a web view. It replaces Spelt's DOM target with GPUi, the native UI framework behind Zed, while keeping familiar component syntax, reactive state, scoped styles, and hot reload.
07:34 Pre-built native binaries mean you can try it without installing Rust and demos run across Mac OS, Windows, and Linux. Time series atlas organizes modern forecasting architectures into one map from classical baselines and linear models to transformers, diffusion, and pre-trained forecasters. Each directory explains the core mechanism, trade-offs, and original references without burying you in paper jargon.
08:00 Four representative models include compact PyTorch implementations that train on synthetic data in under a minute on a laptop. Superlo is a desktop focused email client with a provider agnostic inbox SDK underneath. It combines multiple mailboxes into one keyboard friendly view while keeping credentials, identities, and provider capabilities separate.
08:22 Gmail and inboundnew connections work today with sanitized HTML rendering, encrypted connection credentials, and local SQLite storage. Pictaria server turns your image library into a self-hosted photo curation system. It builds collection insights, groups similar shots into stacks with suggested keepers, and keeps smart albums synchronized from saved searches.
08:46 Optional AI enrichment works with local Alama or OpenAI compatible models, while human review always wins. Unreal is a streaming service where nothing is pre-recorded. A language model writes the next shot while a video model renders it live, so you're watching the show get written as it happens. Each clip starts from the last frame of the one before, so characters and lighting hold together instead of resetting.
09:12 14 original films, plus live channels doing absurdest rapid cut news. Ker fixes the mess of copying the same agent skill into every project folder and forgetting which copy is actually current. It keeps one canonical version of each skill and links it into projects instead of duplicating it. So updating one copy updates everywhere it's installed. It also flags every skill an agent has loaded, managed or not, and estimates the context cost of each.
09:45 Frontier harness eval tests how much the coding agent wrapper matters when the underlying model stays fixed. The maintainers ran Kimmy K3 through 12 harness configurations on 30 software engineering tasks, publishing prompts, task metadata, and normalized results. Their headline finding is similar pass rates with costs differing by 17.5x. Human Atlas lets you take apart a 3D human body in your browser.
10:10 You can toggle anatomical systems, search for structures, and spread visible pieces into an exploded view to see how they fit together. Built with React and 3JS, it runs locally without accounts or API keys and deploys as a static site. Its anatomy comes from an adult male reference, so it doesn't represent every human variation. Codinach puts your coding assistance usage limits on the edge of your Mac screen.
10:39 Hover to see reset times or check whether Claude is working or waiting for you. The panel folds away and can sit on any edge, including around the MacBook notch. Codeex readings come from local logs. Claude uses an unpublished usage endpoint so that integration can break when the provider changes it. Fable is a routing layer for codecs that won't let one model both plan and write the code.
11:04 It splits the job into three tiers. One model plans and adjudicates. Another handles standard implementation. A third just grinds through the loops. Fable never touches your codebase itself. It hands codeex a bounded graph to validate and run. No proxies, no dashboards, nothing extra holding your credentials. AI evaluation framework compares models on accuracy, response time, and cost using answers you define as correct.
11:33 Each field gets its own scoring rule, so an invoice number can require an exact match, while a total allows a small tolerance. It caches responses, so you can change scoring without paying for another run. The included cases and prices are placeholders, so replace them before trusting the comparison. Cororus is basically Slack except your coding agents get their own seat at the table.
11:57 Clawed code, codecs, whatever you're running gets a real CLI workspace instead of a sandbox chat window. So it keeps its actual tools intact. You get channels, threads, taskboards, and file sharing between humans and agents in one place, plus SSH and browserbased remote control. Camera to Blender turns a photo of an object into a generated 3D model you can send straight into your scene.
12:20 Open the camera interface on your phone, take a shot, and Tripo 3D builds the model. Optional Gemini Processing removes the background, while a Blender add-on handles the import. The relay runs on your computer, but Generation uses external APIs, so you'll need a tripo key. No graphics API explores how small a graphics API can get using modern Vulcan extensions.
12:45 Shaders follow GPU pointers directly while your application manages texture and sampler descriptors in GPU memory. That removes much of the usual binding machinery and synchronization focuses on memory hazards. It needs specific extensions currently presents windows only on Windows. Cougar prepares text for animation by measuring where the browser actually wrapped it.
13:10 then splitting it into lines, words, or characters. That preserves layouts such as balanced headings and vertical text while giving CSS, Gap, or other animation tools individual elements to move. It has no dependencies, and you can restore the original markup. When fonts or container widths change, your code must split the text again. Stop stutter targets a specific cause of choppy game streaming on Macs.
13:35 Apple's other L peer-to-peer Wi-Fi interface. It disables that interface while selected apps like Moonlight or GeForce Now are running, then restores it when they quit. Your regular Wi-Fi stays connected. It can help when AWDL causes latency spikes, but Airdrop and some continuity features may be unavailable during play and unrelated stutter needs a different fix.
14:01 Agent memory gives coding agents a shared memory store made of plain markdown files. Clawed code, codecs, and other agents can write and retrieve the same notes, while a rebuildable SQLite index handles search. Hooks capture memories at conversation boundaries, and a separate consolidation pass organizes them over time. Your agent breaks something at 2 a.m.
14:24 and the run is gone by morning. Orca replay records the whole execution bite for bite so you can replay it offline as many times as you want without burning a token. It captures things regular logs miss shell exit codes file rights. You can also fork a run at any checkpoint onto a different model keeping everything else the same.