Hasty Briefsbeta

Bilingual

OSS Alternative to Open WebUI – ChatGPT-Like UI, API and CLI
5 months ago
- Lightweight CLI, API, and ChatGPT-like alternative to Open WebUI for accessing multiple LLMs offline.
- All data kept private in browser storage with support for local and API provider models.
- Configuration via llms.json for providers, models, and default settings.
- Automatic routing to available providers supporting requested models with failover.
- Supports multiple providers including OpenRouter, Ollama, Anthropic, Google, OpenAI, Grok, Groq, Qwen, Z.ai, Mistral.
- OpenAI-compatible API for seamless integration with existing clients.
- Built-in analytics UI for monitoring costs, requests, and token usage.
- Easy provider management with enable/disable functionality and configuration.
- CLI interface for quick interactions and server mode for HTTP API.
- Multi-modal support: text, images (auto-resize and conversion), audio, and documents (PDF).
- Custom chat templates for different modalities and auto-discovery of Ollama models.
- Unified model naming across providers and support for over 160+ LLMs.
- GitHub OAuth for securing web UI and API endpoints with optional user restrictions.
- Docker and Docker Compose support for easy deployment.
- Detailed CLI options for model selection, system prompts, and raw JSON responses.
- Verbose logging and custom log prefixes for debugging.
- Health checks and multi-architecture support (linux/amd64, linux/arm64).
- Extensive documentation for setup, configuration, and troubleshooting.
Cycle-accurate 6502 emulator as coroutine in Rust
5 months ago
- A simple CPU emulator typically uses a loop with a switch statement to interpret opcodes, abstracting away timing details.
- Timing abstraction can cause synchronization issues in systems with timing-sensitive components, like the NES CPU and PPU.
- The solution is to implement the CPU as a state machine that can be stepped one cycle at a time, ensuring accurate synchronization.
- This project demonstrates using Rust coroutines to define CPU emulation semantics, compiling it into a state machine for lockstep execution with the PPU.
- The CPU state includes registers and an I/O data bus, with coroutines yielding I/O events for reads and writes.
- Macros like `fetch!` and `write!` handle I/O events, simplifying the implementation of instruction semantics.
- Example code shows running the CPU and PPU together in lockstep, with the CPU stepping every third cycle to match the PPU's clock rate.
- Performance is sufficient, with the emulator running at ~470 MHz on modern hardware, far exceeding the NES's 1.79 MHz requirement.
- The project includes extensive testing with Harte and Klaus test suites to ensure correctness.
- Built with Rust nightly, the project may be unstable due to compiler changes, but it works with version 1.93.0-nightly.
Show HN: Why write code if the LLM can just do the thing? (web app experiment)
5 months ago
- A web server with no application logic, just an LLM with three tools: database, webResponse, and updateMemory.
- The AI handles HTTP requests by generating responses (HTML, JSON) and executing SQL queries without any pre-written application code.
- Despite being slow (30-60s per request) and expensive ($0.01-0.05 per request), the AI successfully built a functional CRUD contact manager.
- The AI demonstrated emergent behaviors like designing database schemas, writing safe SQL, creating REST-ish APIs, and implementing user feedback.
- Performance issues include high latency, cost, inconsistency in UI, and hallucinations leading to errors.
- The experiment suggests AI's capability to handle application logic is real but currently impractical due to performance limitations.
- Future improvements in inference speed, cost reduction, context memory, and error rates could make this approach viable.
- The vision is a future where computers execute intent directly without layers of code or infrastructure.
- The project is MIT licensed and customizable via a prompt file to build different applications.
CharlotteOS – An Experimental Modern Operating System
5 months ago
- catten is a monolithic kernel developed for the CharlotteOS project, designed for flexibility and use in various environments.
- It features low-level system call interfaces inspired by exokernels, Plan 9, and Fuchsia, with a typesafe system namespace using URIs as paths.
- The kernel supports network access to another host's namespace without local mounting, secured by granular capabilities and mandatory access control.
- Written in Rust and ISA-specific assembly (x86_64 with Intel syntax), with limited C dependencies allowed after vetting.
- Supports x86_64 architecture, UEFI firmware, ACPI, and requires at least 128 MiB memory and 4 GiB storage.
- Compatible with NVMe, USB mass storage, UEFI framebuffers, PS/2 and USB HID input, and USB CDC networking.
- Early development stage; contributions are welcome via GitHub issues, Discord, or Matrix.
- Licensed under GPLv3 or later; contributors must agree to license their work under the same terms.
Solving the NY Times "Pips" game with F#
6 months ago
- Pips is a puzzle game from the New York Times where players cover a shape with dominoes under specific constraints.
- The game uses a backtracking algorithm to solve puzzles efficiently, considering geometric tilings and pruning dead ends.
- Constraints include equal regions (same pip counts), unequal regions (different pip counts), and sum regions (specific pip totals).
- The algorithm checks constraints aggressively during the search to reduce computation time.
- F# is used for implementation due to its functional programming strengths and efficiency with .NET.
- Performance data shows solving times for various puzzles, with the hardest taking over a second and some having thousands of solutions.
- Key data structures include Domino, Cell, Edge, Board, Region, and Puzzle types to model the game state.
- The solver uses tiling trees to guide domino placement and checks validity at each step to ensure constraints are met.
Memlz: Fast compression library for C/C++ on x64/x86
5 months ago
- memlz is a high-speed compression library designed for scenarios requiring speeds close to memcpy().
- Unlike other fast libraries (Snappy, FastLZ, LZAV, LZO), memlz achieves speeds comparable to LZ4 with specific acceleration settings.
- Benchmarks on Intel i7 show decompression speeds vary by data type, with non-cached memcpy() at 14000 MB/s.
- Header-only library; simple integration via `memlz_compress()` and `memlz_stream_compress()` for streaming mode.
- Streaming mode supports compressing data larger than memory or improving ratio for small packets.
- Includes functions to check compressed and decompressed sizes (`memlz_compressed_len()`, `memlz_decompressed_len()`).
- Guarantees safe decompression: no out-of-bounds memory access and regular completion time.
- Eliminates the need for an internal payload queue in streaming mode, unlike LZ4.
- Integration into eXdupe file archiver shows slight speed improvement in streaming mode without significant ratio loss.
- Early beta version; tested only on Intel machines; compatibility may be unstable.
Libcpu: A library to emulate several CPU architectures using LLVM
5 months ago
- libcpu is an open-source library emulating multiple CPU architectures for emulator projects.
- It features frontends for different CPU types and uses LLVM for the backend.
- Supports user mode and system emulation, dynamic and static recompilation.
- Requires CMake version 2.8 or higher.
- Installation instructions provided for Ubuntu and Fedora.
- Build libcpu using 'make' command.
- Run x86 front-end tests with './test/scripts/8086.sh'.
- Developed between 2009-2010, distributed under the 2-clause BSD license.
Value-pool based caching for Java applications
6 months ago
- Mnemosyne is a customizable Java cache library using in-memory databases for efficient updates.
- Supports domain-specific caching algorithms with default FIFO and LRU implementations.
- Works with Spring, with plans for more integrations and distributed system support.
- Uses a Value Pool for objects mapped by IDs to enable simultaneous cache updates.
- Methods are cached via annotations, with support for conditional updates and custom IDs.
- Allows custom cache algorithms by extending AbstractMnemosyneCache.
- Planned features include LFU and S3-FIFO support, better documentation, and distributed caching.
S.a.r.c.a.s.m: Slightly Annoying Rubik's Cube Automatic Solving Machine
5 months ago
- S.A.R.C.A.S.M is a 3D-printed, Teensy-powered robot that scans, solves, and comments on a Rubik’s Cube.
- Includes a Teensy 4.1 main controller and ESP32-CAM for image capture.
- Features an ILI9341 display with custom 2D&3D graphics, animations, and lip-sync.
- Uses stepper motors and servos for cube handling with position sensors to detect faults.
- RGBW lighting synchronized to audio and on-device TTS (espeak-ng) with sarcastic lines.
- Requires a slight modification of Teensy's core to fit the whole code in RAM.
- Project is a work in progress with a messy and incomplete repository state.
Git CLI tool for intelligently creating branch names
5 months ago
- gibr is a CLI tool for creating Git branches connected to issue trackers.
- It supports multiple issue trackers like GitHub, GitLab, Jira, and Linear.
- Automatically generates consistent branch names based on issue details.
- Includes commands like `gibr issues` to list open issues and `gibr 123` to create a branch for issue #123.
- Configurable via `.gibrconfig` file, which can be set up interactively with `gibr init`.
- Offers Git aliases for convenience, allowing commands like `git create 123`.
- Supports placeholders in branch names like `{issuetype}`, `{issue}`, and `{title}`.
- Includes a verbose flag for debug-level logging.
- Encourages contributions and feedback via GitHub issues and discussions.
Show HN: A fast, dependency-free traceroute implementation in pure C
5 months ago
- Fastrace is a high-performance, dependency-free traceroute implementation in pure C.
- Features a fully non-blocking architecture driven by poll() for faster ICMP draining.
- Provides sub-millisecond RTT accuracy with monotonic timing pipeline.
- Offers runtime tuning via CLI flags for hops, probes, concurrency, and timeouts.
- Includes a reverse DNS cache with optional suppression (-n) to optimize lookups.
- Uses expanded socket buffers and smarter probe scheduling for lower latency traces.
- Zero external dependencies, relying solely on standard C libraries and system calls.
- Implements dual socket types: UDP for sending probes and raw ICMP for receiving responses.
- Configurable multi-TTL probing system with tunable concurrency window.
- Precise handling of network protocols including UDP probes, TTL management, and ICMP response processing.
- Structured visual representation of network paths with tree-like format.
- Significant performance improvements over standard traceroute in trace time, memory usage, and CPU utilization.
- Supports Linux, macOS, and other Unix-like systems with raw socket support.
- Includes a Makefile for easy compilation and installation.
- Licensed under the BSD-2 License.
No Code
5 months ago
- The best way to write code is to write no code at all.
- The package allows writing nothing while still having functional code.
- Installation requires Python 3.10+ and a virtual environment is recommended.
- Commands 'no_code' and 'yes_code' transform code to no code and vice versa.
- Includes functions 'no_code.nothing()' and 'no_code.something()' for transformations.
- Inspired by Kelsey Hightower and based on INVISIBLE.js by Martin Kleppe.
Chromium Browser DoS Attack via Document.title Exploitation
5 months ago
- Brash is a critical vulnerability in Chromium's Blink rendering engine, causing browsers to crash in 15-60 seconds.
- The exploit works by abusing the lack of rate limiting on the `document.title` API, allowing millions of DOM mutations per second.
- This saturation of the main thread disrupts the event loop, leading to browser collapse and high CPU usage.
- Affects all Chromium-based browsers (Chrome, Edge, Vivaldi, etc.), but Firefox and Safari are immune.
- The attack involves preloading 100 unique strings to maximize speed and avoid detection.
- Three phases: string generation, burst execution, and continuous updates leading to UI freeze.
- Impact includes system performance degradation, process halting, and potential for coordinated attacks.
- Can be weaponized with delayed or scheduled execution, making it dangerous in critical scenarios.
- Potential real-world impacts: financial market disruption, medical emergencies, and fraud exploitation.
- Intended for educational and security research purposes only, with strict ethical guidelines.
Show HN: Quibbler – A critic for your coding agent that learns what you want
5 months ago
- Quibbler is a background critic for coding agents, automatically correcting failures and enforcing learned rules.
- Prevents common issues like fabricating results, skipping tests, not following coding styles, and hallucinating functionality.
- Supports two integration modes: event-driven monitoring via Claude Code's hook system and universal Model Context Protocol (MCP).
- Setup involves configuring MCP servers or running a hook server, with project-specific instructions in AGENTS.md.
- Quibbler reviews code changes against user intent, validates claims, checks for hallucinations, and ensures testing steps.
- Maintains context and learns project patterns over time, saving rules to .quibbler/rules.md.
- Customizable system prompt and supports different Claude models like Haiku and Sonnet.
- Community support available via Discord and contributions welcome through GitHub issues and PRs.
Kimi Linear: An Expressive, Efficient Attention Architecture
5 months ago
- Kimi Linear achieves 51.0 performance on MMLU-Pro with similar speed as full attention.
- On RULER (128k context length), it shows Pareto-optimal performance (84.3) and a 3.98x speedup.
- Kimi Linear is 6.3x faster TPOT compared to MLA, especially for long sequences (1M tokens).
- Kimi Delta Attention (KDA) is a refined version of Gated DeltaNet with an efficient gating mechanism.
- Reduces KV caches by up to 75% and boosts decoding throughput significantly.
- Open-sourced KDA kernel in FLA and released two model checkpoints trained with 5.7T tokens.
- Hybrid architecture with a 3:1 KDA-to-global MLA ratio reduces memory usage while maintaining quality.
- Superior performance in long-context and RL-style benchmarks on 1.4T token training runs.
- Achieves up to 6x faster decoding and reduces time per output token (TPOT).
- Example code provided for using the Kimi Linear model with Python and deployment via vllm.
Show HN: Front End Fuzzy and Substring and Prefix Search
5 months ago
- @m31coding/fuzzy-search is a frontend library for fast and accurate fuzzy searching of entities by their names and features.
- Key features include speed (<10 ms queries), accuracy (suffix array and n-grams), multilingual support, flexibility (CRUD operations), reliability (no dependencies), and universal compatibility (frontend and Node.js).
- Install via npm: `npm install @m31coding/fuzzy-search` with various module formats available (ESM, CommonJS, UMD, etc.).
- Basic usage involves creating a searcher with `SearcherFactory`, indexing entities, and querying with `getMatches`.
- Supports dynamic updates with `upsertEntities` and `removeEntities`, and handles large datasets efficiently.
- Customizable normalization pipeline for handling different languages and scripts, with default settings optimized for Latin scripts.
- Advanced search techniques include n-grams with character sorting for transposition errors and quality penalties for inexact matches.
- Substring and prefix searches use a suffix array with quality offsets to prioritize matches.
- Configuration options allow tuning for specific needs, such as handling non-Latin scripts or large datasets.
- Open-source and free, contributions and feedback are welcome via GitHub issues.
TailPass
5 months ago
- Tailpass is a Tailscale-powered TCP port forwarding tool for secure network bridging.
- It simplifies connecting web servers, SSH, databases, and other TCP services without complex configurations.
- Components include a Go backend, Vue.js frontend, and PostgreSQL database, all containerized with Docker Compose.
- Quick setup involves cloning the repo and running `docker compose up --build`.
- Services include Frontend (http://localhost:8000), Backend (http://localhost:8082), and Database (localhost:5432).
- Supports binding to localhost, Tailscale IP, or all interfaces for flexible access.
- Future improvements: GUI refinements, health checks, JWT authentication, ORM integration, and deployment optimizations.
- Planned features: service management controls, real-time container metrics, and configurable alerts.
Show HN: Run a GitHub Actions step in a gVisor sandbox
6 months ago
- GitHub Actions with read-only permissions still receive a cache write token, allowing cache poisoning.
- No isolation between steps in a workflow as they run on the same VM with root access.
- Running untrusted code in a workflow with read-only permissions can lead to a false sense of security.
- This GitHub Action runs commands in an isolated gVisor sandbox to mitigate risks.
- The sandbox features include a root filesystem similar to ubuntu-24.04, overlayed access to GITHUB_WORKSPACE, and read-only access to tools installed by setup-* actions.
- Changes to the workspace inside the sandbox can persist on the host if 'persist-workspace-changes' is set to 'true', but this is not safe.
- The action detects and fails if actions/checkout has persisted authentication tokens in GITHUB_WORKSPACE.
- To use this action, disable credential persistence in the checkout step or set 'allow-checkout-credentials' to 'true' (not recommended).
- All tags use GitHub's Immutable Releases for security.
- Example workflow includes running Go tests with different versions and dependencies, staticcheck, and govulncheck in a sandboxed environment.
GooglePlay reports latest F-Droid version of Aves Libre as potential malware
5 months ago
- Bug report: Google Play scanner warning when updating Aves Libre to version 1.13.9 via Droid-ify.
- Expected behavior: Installation should proceed without any warnings.
- Steps to reproduce: Enable security scanner in Google Play, attempt to install Aves Libre 1.13.9, and observe the warning.
- System information includes: Package name 'deckers.thibault.aves.libre', installer 'com.looker.droidify', Aves version '1.13.9-libre', Flutter stable version '3.27.4', Android version '16, API 36', device 'Google Pixel 7a'.
- Additional details: Display specifications, mobile services status, connectivity, system locales, storage volumes, and grants are provided.
Minecraft HDL, an HDL for Redstone
5 months ago
- MinecraftHDL is a digital synthesis flow for Minecraft redstone circuits using industry-standard design tools.
- It can generate digital circuits from Verilog code, though complex designs may not be testable in-game due to block limits.
- The tool currently cannot synthesize sequential circuits (e.g., memory, counters).
- MinecraftHDL was developed as an undergraduate project by three McGill University students.
- It serves as an educational tool for teaching digital design concepts.
- A sample circuit, a 4-to-1 multiplexer, is provided in Verilog and synthesized in Minecraft.
- The project includes resources like screenshots, installation guides, and background theory.
- A 2-bit adder demo shows the quick synthesis and testing process in Minecraft.

first prev36next

About|Login

Bilingual

OSS Alternative to Open WebUI – ChatGPT-Like UI, API and CLI
5 months ago
- Lightweight CLI, API, and ChatGPT-like alternative to Open WebUI for accessing multiple LLMs offline.
- All data kept private in browser storage with support for local and API provider models.
- Configuration via llms.json for providers, models, and default settings.
- Automatic routing to available providers supporting requested models with failover.
- Supports multiple providers including OpenRouter, Ollama, Anthropic, Google, OpenAI, Grok, Groq, Qwen, Z.ai, Mistral.
- OpenAI-compatible API for seamless integration with existing clients.
- Built-in analytics UI for monitoring costs, requests, and token usage.
- Easy provider management with enable/disable functionality and configuration.
- CLI interface for quick interactions and server mode for HTTP API.
- Multi-modal support: text, images (auto-resize and conversion), audio, and documents (PDF).
- Custom chat templates for different modalities and auto-discovery of Ollama models.
- Unified model naming across providers and support for over 160+ LLMs.
- GitHub OAuth for securing web UI and API endpoints with optional user restrictions.
- Docker and Docker Compose support for easy deployment.
- Detailed CLI options for model selection, system prompts, and raw JSON responses.
- Verbose logging and custom log prefixes for debugging.
- Health checks and multi-architecture support (linux/amd64, linux/arm64).
- Extensive documentation for setup, configuration, and troubleshooting.
Cycle-accurate 6502 emulator as coroutine in Rust
5 months ago
- A simple CPU emulator typically uses a loop with a switch statement to interpret opcodes, abstracting away timing details.
- Timing abstraction can cause synchronization issues in systems with timing-sensitive components, like the NES CPU and PPU.
- The solution is to implement the CPU as a state machine that can be stepped one cycle at a time, ensuring accurate synchronization.
- This project demonstrates using Rust coroutines to define CPU emulation semantics, compiling it into a state machine for lockstep execution with the PPU.
- The CPU state includes registers and an I/O data bus, with coroutines yielding I/O events for reads and writes.
- Macros like `fetch!` and `write!` handle I/O events, simplifying the implementation of instruction semantics.
- Example code shows running the CPU and PPU together in lockstep, with the CPU stepping every third cycle to match the PPU's clock rate.
- Performance is sufficient, with the emulator running at ~470 MHz on modern hardware, far exceeding the NES's 1.79 MHz requirement.
- The project includes extensive testing with Harte and Klaus test suites to ensure correctness.
- Built with Rust nightly, the project may be unstable due to compiler changes, but it works with version 1.93.0-nightly.
Show HN: Why write code if the LLM can just do the thing? (web app experiment)
5 months ago
- A web server with no application logic, just an LLM with three tools: database, webResponse, and updateMemory.
- The AI handles HTTP requests by generating responses (HTML, JSON) and executing SQL queries without any pre-written application code.
- Despite being slow (30-60s per request) and expensive ($0.01-0.05 per request), the AI successfully built a functional CRUD contact manager.
- The AI demonstrated emergent behaviors like designing database schemas, writing safe SQL, creating REST-ish APIs, and implementing user feedback.
- Performance issues include high latency, cost, inconsistency in UI, and hallucinations leading to errors.
- The experiment suggests AI's capability to handle application logic is real but currently impractical due to performance limitations.
- Future improvements in inference speed, cost reduction, context memory, and error rates could make this approach viable.
- The vision is a future where computers execute intent directly without layers of code or infrastructure.
- The project is MIT licensed and customizable via a prompt file to build different applications.
CharlotteOS – An Experimental Modern Operating System
5 months ago
- catten is a monolithic kernel developed for the CharlotteOS project, designed for flexibility and use in various environments.
- It features low-level system call interfaces inspired by exokernels, Plan 9, and Fuchsia, with a typesafe system namespace using URIs as paths.
- The kernel supports network access to another host's namespace without local mounting, secured by granular capabilities and mandatory access control.
- Written in Rust and ISA-specific assembly (x86_64 with Intel syntax), with limited C dependencies allowed after vetting.
- Supports x86_64 architecture, UEFI firmware, ACPI, and requires at least 128 MiB memory and 4 GiB storage.
- Compatible with NVMe, USB mass storage, UEFI framebuffers, PS/2 and USB HID input, and USB CDC networking.
- Early development stage; contributions are welcome via GitHub issues, Discord, or Matrix.
- Licensed under GPLv3 or later; contributors must agree to license their work under the same terms.
Solving the NY Times "Pips" game with F#
6 months ago
- Pips is a puzzle game from the New York Times where players cover a shape with dominoes under specific constraints.
- The game uses a backtracking algorithm to solve puzzles efficiently, considering geometric tilings and pruning dead ends.
- Constraints include equal regions (same pip counts), unequal regions (different pip counts), and sum regions (specific pip totals).
- The algorithm checks constraints aggressively during the search to reduce computation time.
- F# is used for implementation due to its functional programming strengths and efficiency with .NET.
- Performance data shows solving times for various puzzles, with the hardest taking over a second and some having thousands of solutions.
- Key data structures include Domino, Cell, Edge, Board, Region, and Puzzle types to model the game state.
- The solver uses tiling trees to guide domino placement and checks validity at each step to ensure constraints are met.
Memlz: Fast compression library for C/C++ on x64/x86
5 months ago
- memlz is a high-speed compression library designed for scenarios requiring speeds close to memcpy().
- Unlike other fast libraries (Snappy, FastLZ, LZAV, LZO), memlz achieves speeds comparable to LZ4 with specific acceleration settings.
- Benchmarks on Intel i7 show decompression speeds vary by data type, with non-cached memcpy() at 14000 MB/s.
- Header-only library; simple integration via `memlz_compress()` and `memlz_stream_compress()` for streaming mode.
- Streaming mode supports compressing data larger than memory or improving ratio for small packets.
- Includes functions to check compressed and decompressed sizes (`memlz_compressed_len()`, `memlz_decompressed_len()`).
- Guarantees safe decompression: no out-of-bounds memory access and regular completion time.
- Eliminates the need for an internal payload queue in streaming mode, unlike LZ4.
- Integration into eXdupe file archiver shows slight speed improvement in streaming mode without significant ratio loss.
- Early beta version; tested only on Intel machines; compatibility may be unstable.
Libcpu: A library to emulate several CPU architectures using LLVM
5 months ago
- libcpu is an open-source library emulating multiple CPU architectures for emulator projects.
- It features frontends for different CPU types and uses LLVM for the backend.
- Supports user mode and system emulation, dynamic and static recompilation.
- Requires CMake version 2.8 or higher.
- Installation instructions provided for Ubuntu and Fedora.
- Build libcpu using 'make' command.
- Run x86 front-end tests with './test/scripts/8086.sh'.
- Developed between 2009-2010, distributed under the 2-clause BSD license.
Value-pool based caching for Java applications
6 months ago
- Mnemosyne is a customizable Java cache library using in-memory databases for efficient updates.
- Supports domain-specific caching algorithms with default FIFO and LRU implementations.
- Works with Spring, with plans for more integrations and distributed system support.
- Uses a Value Pool for objects mapped by IDs to enable simultaneous cache updates.
- Methods are cached via annotations, with support for conditional updates and custom IDs.
- Allows custom cache algorithms by extending AbstractMnemosyneCache.
- Planned features include LFU and S3-FIFO support, better documentation, and distributed caching.
S.a.r.c.a.s.m: Slightly Annoying Rubik's Cube Automatic Solving Machine
5 months ago
- S.A.R.C.A.S.M is a 3D-printed, Teensy-powered robot that scans, solves, and comments on a Rubik’s Cube.
- Includes a Teensy 4.1 main controller and ESP32-CAM for image capture.
- Features an ILI9341 display with custom 2D&3D graphics, animations, and lip-sync.
- Uses stepper motors and servos for cube handling with position sensors to detect faults.
- RGBW lighting synchronized to audio and on-device TTS (espeak-ng) with sarcastic lines.
- Requires a slight modification of Teensy's core to fit the whole code in RAM.
- Project is a work in progress with a messy and incomplete repository state.
Git CLI tool for intelligently creating branch names
5 months ago
- gibr is a CLI tool for creating Git branches connected to issue trackers.
- It supports multiple issue trackers like GitHub, GitLab, Jira, and Linear.
- Automatically generates consistent branch names based on issue details.
- Includes commands like `gibr issues` to list open issues and `gibr 123` to create a branch for issue #123.
- Configurable via `.gibrconfig` file, which can be set up interactively with `gibr init`.
- Offers Git aliases for convenience, allowing commands like `git create 123`.
- Supports placeholders in branch names like `{issuetype}`, `{issue}`, and `{title}`.
- Includes a verbose flag for debug-level logging.
- Encourages contributions and feedback via GitHub issues and discussions.
Show HN: A fast, dependency-free traceroute implementation in pure C
5 months ago
- Fastrace is a high-performance, dependency-free traceroute implementation in pure C.
- Features a fully non-blocking architecture driven by poll() for faster ICMP draining.
- Provides sub-millisecond RTT accuracy with monotonic timing pipeline.
- Offers runtime tuning via CLI flags for hops, probes, concurrency, and timeouts.
- Includes a reverse DNS cache with optional suppression (-n) to optimize lookups.
- Uses expanded socket buffers and smarter probe scheduling for lower latency traces.
- Zero external dependencies, relying solely on standard C libraries and system calls.
- Implements dual socket types: UDP for sending probes and raw ICMP for receiving responses.
- Configurable multi-TTL probing system with tunable concurrency window.
- Precise handling of network protocols including UDP probes, TTL management, and ICMP response processing.
- Structured visual representation of network paths with tree-like format.
- Significant performance improvements over standard traceroute in trace time, memory usage, and CPU utilization.
- Supports Linux, macOS, and other Unix-like systems with raw socket support.
- Includes a Makefile for easy compilation and installation.
- Licensed under the BSD-2 License.
No Code
5 months ago
- The best way to write code is to write no code at all.
- The package allows writing nothing while still having functional code.
- Installation requires Python 3.10+ and a virtual environment is recommended.
- Commands 'no_code' and 'yes_code' transform code to no code and vice versa.
- Includes functions 'no_code.nothing()' and 'no_code.something()' for transformations.
- Inspired by Kelsey Hightower and based on INVISIBLE.js by Martin Kleppe.
Chromium Browser DoS Attack via Document.title Exploitation
5 months ago
- Brash is a critical vulnerability in Chromium's Blink rendering engine, causing browsers to crash in 15-60 seconds.
- The exploit works by abusing the lack of rate limiting on the `document.title` API, allowing millions of DOM mutations per second.
- This saturation of the main thread disrupts the event loop, leading to browser collapse and high CPU usage.
- Affects all Chromium-based browsers (Chrome, Edge, Vivaldi, etc.), but Firefox and Safari are immune.
- The attack involves preloading 100 unique strings to maximize speed and avoid detection.
- Three phases: string generation, burst execution, and continuous updates leading to UI freeze.
- Impact includes system performance degradation, process halting, and potential for coordinated attacks.
- Can be weaponized with delayed or scheduled execution, making it dangerous in critical scenarios.
- Potential real-world impacts: financial market disruption, medical emergencies, and fraud exploitation.
- Intended for educational and security research purposes only, with strict ethical guidelines.
Show HN: Quibbler – A critic for your coding agent that learns what you want
5 months ago
- Quibbler is a background critic for coding agents, automatically correcting failures and enforcing learned rules.
- Prevents common issues like fabricating results, skipping tests, not following coding styles, and hallucinating functionality.
- Supports two integration modes: event-driven monitoring via Claude Code's hook system and universal Model Context Protocol (MCP).
- Setup involves configuring MCP servers or running a hook server, with project-specific instructions in AGENTS.md.
- Quibbler reviews code changes against user intent, validates claims, checks for hallucinations, and ensures testing steps.
- Maintains context and learns project patterns over time, saving rules to .quibbler/rules.md.
- Customizable system prompt and supports different Claude models like Haiku and Sonnet.
- Community support available via Discord and contributions welcome through GitHub issues and PRs.
Kimi Linear: An Expressive, Efficient Attention Architecture
5 months ago
- Kimi Linear achieves 51.0 performance on MMLU-Pro with similar speed as full attention.
- On RULER (128k context length), it shows Pareto-optimal performance (84.3) and a 3.98x speedup.
- Kimi Linear is 6.3x faster TPOT compared to MLA, especially for long sequences (1M tokens).
- Kimi Delta Attention (KDA) is a refined version of Gated DeltaNet with an efficient gating mechanism.
- Reduces KV caches by up to 75% and boosts decoding throughput significantly.
- Open-sourced KDA kernel in FLA and released two model checkpoints trained with 5.7T tokens.
- Hybrid architecture with a 3:1 KDA-to-global MLA ratio reduces memory usage while maintaining quality.
- Superior performance in long-context and RL-style benchmarks on 1.4T token training runs.
- Achieves up to 6x faster decoding and reduces time per output token (TPOT).
- Example code provided for using the Kimi Linear model with Python and deployment via vllm.
Show HN: Front End Fuzzy and Substring and Prefix Search
5 months ago
- @m31coding/fuzzy-search is a frontend library for fast and accurate fuzzy searching of entities by their names and features.
- Key features include speed (<10 ms queries), accuracy (suffix array and n-grams), multilingual support, flexibility (CRUD operations), reliability (no dependencies), and universal compatibility (frontend and Node.js).
- Install via npm: `npm install @m31coding/fuzzy-search` with various module formats available (ESM, CommonJS, UMD, etc.).
- Basic usage involves creating a searcher with `SearcherFactory`, indexing entities, and querying with `getMatches`.
- Supports dynamic updates with `upsertEntities` and `removeEntities`, and handles large datasets efficiently.
- Customizable normalization pipeline for handling different languages and scripts, with default settings optimized for Latin scripts.
- Advanced search techniques include n-grams with character sorting for transposition errors and quality penalties for inexact matches.
- Substring and prefix searches use a suffix array with quality offsets to prioritize matches.
- Configuration options allow tuning for specific needs, such as handling non-Latin scripts or large datasets.
- Open-source and free, contributions and feedback are welcome via GitHub issues.
TailPass
5 months ago
- Tailpass is a Tailscale-powered TCP port forwarding tool for secure network bridging.
- It simplifies connecting web servers, SSH, databases, and other TCP services without complex configurations.
- Components include a Go backend, Vue.js frontend, and PostgreSQL database, all containerized with Docker Compose.
- Quick setup involves cloning the repo and running `docker compose up --build`.
- Services include Frontend (http://localhost:8000), Backend (http://localhost:8082), and Database (localhost:5432).
- Supports binding to localhost, Tailscale IP, or all interfaces for flexible access.
- Future improvements: GUI refinements, health checks, JWT authentication, ORM integration, and deployment optimizations.
- Planned features: service management controls, real-time container metrics, and configurable alerts.
Show HN: Run a GitHub Actions step in a gVisor sandbox
6 months ago
- GitHub Actions with read-only permissions still receive a cache write token, allowing cache poisoning.
- No isolation between steps in a workflow as they run on the same VM with root access.
- Running untrusted code in a workflow with read-only permissions can lead to a false sense of security.
- This GitHub Action runs commands in an isolated gVisor sandbox to mitigate risks.
- The sandbox features include a root filesystem similar to ubuntu-24.04, overlayed access to GITHUB_WORKSPACE, and read-only access to tools installed by setup-* actions.
- Changes to the workspace inside the sandbox can persist on the host if 'persist-workspace-changes' is set to 'true', but this is not safe.
- The action detects and fails if actions/checkout has persisted authentication tokens in GITHUB_WORKSPACE.
- To use this action, disable credential persistence in the checkout step or set 'allow-checkout-credentials' to 'true' (not recommended).
- All tags use GitHub's Immutable Releases for security.
- Example workflow includes running Go tests with different versions and dependencies, staticcheck, and govulncheck in a sandboxed environment.
GooglePlay reports latest F-Droid version of Aves Libre as potential malware
5 months ago
- Bug report: Google Play scanner warning when updating Aves Libre to version 1.13.9 via Droid-ify.
- Expected behavior: Installation should proceed without any warnings.
- Steps to reproduce: Enable security scanner in Google Play, attempt to install Aves Libre 1.13.9, and observe the warning.
- System information includes: Package name 'deckers.thibault.aves.libre', installer 'com.looker.droidify', Aves version '1.13.9-libre', Flutter stable version '3.27.4', Android version '16, API 36', device 'Google Pixel 7a'.
- Additional details: Display specifications, mobile services status, connectivity, system locales, storage volumes, and grants are provided.
Minecraft HDL, an HDL for Redstone
5 months ago
- MinecraftHDL is a digital synthesis flow for Minecraft redstone circuits using industry-standard design tools.
- It can generate digital circuits from Verilog code, though complex designs may not be testable in-game due to block limits.
- The tool currently cannot synthesize sequential circuits (e.g., memory, counters).
- MinecraftHDL was developed as an undergraduate project by three McGill University students.
- It serves as an educational tool for teaching digital design concepts.
- A sample circuit, a 4-to-1 multiplexer, is provided in Verilog and synthesized in Minecraft.
- The project includes resources like screenshots, installation guides, and background theory.
- A 2-bit adder demo shows the quick synthesis and testing process in Minecraft.

Hasty Briefsbeta

OSS Alternative to Open WebUI – ChatGPT-Like UI, API and CLI

Cycle-accurate 6502 emulator as coroutine in Rust

Show HN: Why write code if the LLM can just do the thing? (web app experiment)

CharlotteOS – An Experimental Modern Operating System

Solving the NY Times "Pips" game with F#

Memlz: Fast compression library for C/C++ on x64/x86

Libcpu: A library to emulate several CPU architectures using LLVM

Value-pool based caching for Java applications

S.a.r.c.a.s.m: Slightly Annoying Rubik's Cube Automatic Solving Machine

Git CLI tool for intelligently creating branch names

Show HN: A fast, dependency-free traceroute implementation in pure C

No Code

Chromium Browser DoS Attack via Document.title Exploitation

Show HN: Quibbler – A critic for your coding agent that learns what you want

Kimi Linear: An Expressive, Efficient Attention Architecture

Show HN: Front End Fuzzy and Substring and Prefix Search

TailPass

Show HN: Run a GitHub Actions step in a gVisor sandbox

GooglePlay reports latest F-Droid version of Aves Libre as potential malware

Minecraft HDL, an HDL for Redstone

OSS Alternative to Open WebUI – ChatGPT-Like UI, API and CLI

Cycle-accurate 6502 emulator as coroutine in Rust

Show HN: Why write code if the LLM can just do the thing? (web app experiment)

CharlotteOS – An Experimental Modern Operating System

Solving the NY Times "Pips" game with F#

Memlz: Fast compression library for C/C++ on x64/x86

Libcpu: A library to emulate several CPU architectures using LLVM

Value-pool based caching for Java applications

S.a.r.c.a.s.m: Slightly Annoying Rubik's Cube Automatic Solving Machine

Git CLI tool for intelligently creating branch names

Show HN: A fast, dependency-free traceroute implementation in pure C

No Code

Chromium Browser DoS Attack via Document.title Exploitation

Show HN: Quibbler – A critic for your coding agent that learns what you want

Kimi Linear: An Expressive, Efficient Attention Architecture

Show HN: Front End Fuzzy and Substring and Prefix Search

TailPass

Show HN: Run a GitHub Actions step in a gVisor sandbox

GooglePlay reports latest F-Droid version of Aves Libre as potential malware

Minecraft HDL, an HDL for Redstone