Turn text into videos, images, and speech. Your face on an AI avatar? Done. Photos speaking? You got it! Ever wanted to clone yourself? Now, it's possible!
Joined January 2023
- Tweets 4,486
- Following 14
- Followers 1,629
- Likes 1,721
147 Photos and videos
JUST IN: Advanced tutorial explores coding FineWeb dataset. Stream, filter, deduplicate, tokenize, and analyze large-scale web corpus efficiently. Includes quality filters, near-duplicate detection, token count verification, domain analysis, and summary.
4
JUST IN: US government order leads to Anthropic disabling Claude Fable 5 and Mythos 5 for all customers. Directive cites national security concerns, Anthropic disputes over narrow jailbreak claim. Other models not affected. #AI #Anthropic #USGov
29
JUST IN: Moonshot AI releases Kimi K2.7-Code, a coding model surpassing K2.6 on Kimi Code Bench v2 with a 21.8% improvement. K2.7-Code features a Mixture-of-Experts model with impressive parameters and enhanced performance. #AI #Coding #MoonshotAI
37
JUST IN: Tutorial on implementing spatial graph neural networks for urban function inference using city2graph, OSMnx, and PyTorch Geometric. Learn about building spatial features, proximity graphs, training a GraphSAGE model, and running heterogeneous GNN experiments.
10
JUST IN: Google launches Gemini-SQL2, a breakthrough text-to-SQL system with 80.04% accuracy on BIRD Single-Model Leaderboard. Powered by Gemini 3.1 Pro, it elevates natural language skills for data services. No confirmation on integration plans yet.
25
JUST IN: Moonshot AI introduces Kimi Work, a local desktop agent running on Kimi K2.6 with a 300-sub-agent swarm. Kimi Work reads local files, drives your browser, and performs scheduled tasks. It targets knowledge workers seeking file access and live sessions.
18
JUST IN: Tutorial on MONAI for 3D Spleen Segmentation. Implementing UNet on Medical CT Volumes with transforms, training, validation, and visualization. Includes mixed precision training and DiceCE loss. #MedicalImaging #UNet #MONAI #CTSegmentation
10
JUST IN: Nous Research releases Profile Builder for Hermes Agent. Build agent identity, model, skills, and MCP servers in one guided flow within the project's web dashboard. Easily set up distinct agents with new browser form, enhancing usability and customization options.
1
1
51
JUST IN: Google AI team introduces DiffusionGemma, a 26B MoE open model for text generation. It uses text diffusion instead of autoregressive decoding, offering up to 4x faster generation on dedicated GPUs. Model supports multiple languages and multimodal inputs. #GoogleAI
1
30
JUST IN: AI coding landscape in 2026 introduces top agents like Atoms for end-to-end product building, Devin AI for well-defined tasks, GitHub Copilot for real-time code suggestions, and more tools for specialized needs. Expect further integrations and reliability enhancements.
30
JUST IN: Google introduces Gemini 3.5 Live Translate, a real-time speech-to-speech audio model for 70 languages across various platforms like Meet and Translate. The model offers continuous streaming translation, maintaining speaker nuances.
27
JUST IN: Xiaomi MiMo and TileRT collaborated to achieve 1000 tokens per second on a 1T model using three key techniques. The speed boost comes from FP4 quantization, DFlash speculative decoding, and TileRT runtime on commodity GPUs.
69
JUST IN: Microsoft AI releases MAI-Transcribe-1.5 with 2.4% WER, top FLEURS accuracy, and up to 5x faster long-audio transcription. AI model supports 43 languages, improved keyword biasing, and integrates with various Microsoft products.
17
JUST IN: Tutorial on using GEPA for reflective prompt optimization. From weak seed prompts to structured feedback, explore evolving prompts for improved language model performance on arithmetic word problems. Held-out validation shows optimized prompt benefits.
1
39
JUST IN: Harness-1, a 20B retrieval subagent trained with reinforcement learning on gpt-oss-20b, separates search decisions from bookkeeping. It achieves high recall and keeps code open for public use. Exciting developments in AI research!
36