Discover Amazing Content, Share Life Moments
Connect Our Wonderful World
100M Parameters, Runs on CPU, 6x Real-Time Speed: Kyutai Labs Releases a Lightweight TTS That Eliminates the Need for a GPU
Kyutai Labs has quietly launched Pocket TTS, a lightweight text-to-speech model with 100 million parameters that runs on CPU, featuring 200ms latency and requiring only 5 seconds of audio for voice cloning. It is open-sourced under the MIT license, allowing free commercial use. This is arguably the most practical implementation of local AI speech synthesis to date.
Tencent Hy3 Released: 295B MoE, Progress from Hy2 to Hy3 in Six Months, Performance Matches Trillion-Parameter Flagship Models
Tencent Hunyuan has released the Hy3 model, a 295B-parameter MoE with 21B activated parameters. It matches the performance of trillion-parameter models across multiple benchmarks, is open-sourced under Apache 2.0, and offers a two-week free API.
Ran Out of Claude Quota Again? This Open-Source Tool Automatically Switches You to Free Models
An open-source tool called 9Router lets you automatically switch between more than 40 AI providers. When your Claude quota runs out, it automatically falls back to cheaper models, and switches to free models if those are also unavailable. It supports Claude Code, Cursor, Copilot and more. Setup takes just two steps, and it's completely free.
DeepTutor v1.5 Released: An Open-Source AI Tutor That Remembers All Your Learning Trajectories
The DeepTutor project open-sourced by a HKU team has been upgraded to v1.5. The core update integrates tutoring, practice, memory, sub-agents, and IM companions into a single agent loop. It is not just another AI tutor with a hodgepodge of features, but a truly inspectable and extensible learning operating system.
35B Model Hits 79 t/s, But 256K Context Is the Real Litmus Test for DGX Spark
llama.cpp benchmark results for Qwen3.6-35B-A3B running on DGX Spark are out: it delivers 79 t/s generation speed at 2K context, but drops to 31 t/s at 256K. The Mia team also released a one-click startup script optimized specifically for 96-128GB VRAM devices. Is long context a cover for poor performance, or a real practical pain point?
AI Agent Automation on SaaS Workflows: Top Scoring Model Fails to Reach 50%, Financial Tasks Prove Most Challenging
Zapier and Artificial Analysis have jointly launched the independent AutomationBench-AA leaderboard, which tests the automation capabilities of AI agents in real-world SaaS workflows. The results show that Claude Fable 5 leads with 48.6% accuracy, but all models violate business rules, and the completion rate for financial tasks is only half that of other domains.
Reddit Uses LLMs to Kill the Garbage Created by LLMs – Makes Sense
Reddit's LLM-powered tool blocks 23 million spam views daily, most of which are also generated by LLMs. Fighting fire with fire, and it's working reasonably well.
2026 Tech Layoff List: Is AI an Excuse or the Truth?
In 2026, tech companies are hitting record revenues while conducting massive layoffs, with AI being the most frequently cited reason. Giants like Microsoft, Google, Meta, and Oracle are cutting headcount, but is AI really replacing humans behind these layoffs? This list documents all major layoff events this year where AI was cited as a reason—dense with information and worth a close look.
Claude Code Can Now Process Videos, With a Clever New Approach
An open-source project enables Claude Code to directly process YouTube videos, Zoom recordings, and local video files. Instead of random frame sampling, it uses FFmpeg for scene change detection and only captures frames where the visual content actually changes. Subtitles are sourced from official tracks or generated via Whisper, and the analysis can be saved to a knowledge base afterward.
Is Your AI Agent Still Repeating What You Said Last Week? This Is “Ghost Memory”
Long-running AI agents stubbornly repeat facts that users mentioned weeks ago, even after those facts have become outdated. A new study from the National University of Singapore has named this failure mode Ghost Memory, and proposed the A-TMA solution that enables memory systems to learn to distinguish between “present”, “past”, and “changes”.
