Open-Source ReelMimic Turns Any Reference Clip Into Agent-Built 2D Animation
ReelMimic turns any reference clip into an original 2D animation by orchestrating a crew of Claude Code or Codex agents on your own machine.
20 stories
ReelMimic turns any reference clip into an original 2D animation by orchestrating a crew of Claude Code or Codex agents on your own machine.
Ricursive Intelligence trained a 65B dense coding model for $32M that beats open-weight MoE giants up to 15x its size on agentic coding.
A community GGUF port brings Tencent's DSpine draft model to llama.cpp, pushing Qwen3.8-27B decoding to ~105 tok/s on a single RTX 3090.
The all-in-one Python framework unifies vector search, agents, RAG, and multi-model workflows behind a single embeddings database that scales from laptop to cluster.
Nous Research took Alibaba's Qwen3.8-27B, post-trained it on 53M tokens, and cut state-aligned political framing from 22% to under 1%.
Suno's open source Bark model is trending again, offering a fully generative text-to-audio system that handles speech, music, laughter, and sound effects.
LAION releases a 365M-clip audio dataset with 707M captions and a family of open CLAP2 models that set new state of the art on MAEB-30.
AutoTrust Labs released a 27B text decision model that scores option probabilities instead of generating free-form answers, hitting 97 on tool selection.
Voyager's new Rust engine lets coding agents write, render, and iterate on cinematic 3D scenes without any video generation model
A new survey maps how TinyML is reshaping farm sensing on ESP32, STM32, and UAV hardware, exposing serious gaps in energy and memory reporting.
Deletion Tuning learns a binary keep-or-delete mask over gate projections, matching full fine-tuning accuracy while zeroing under 0.05% of weights.
Two Hacker News posts say a team of Qwen3.8-27B models matches or beats Claude Fable 5. Here are the four reasons a second agent helps, and how to test yours.
A 122B mixture-of-experts reasoning model with MIT weights, a 262K context, and 84.5% on SWE-bench Verified, served from a single 80GB GPU.
Prism ML's 2B vision-language model squeezes into roughly half a gigabyte using ternary weights while keeping a 262K context window and image input.
A new 500-question long-video benchmark forces models to not just answer correctly but prove where in the video the evidence lives. Frontier systems collapse.
A solo developer fully fine-tuned Krea 2's 12.8B base model on 196k anime illustrations in about 8 hours, then transferred the result to the fast 8-step Turbo checkpoint.
COEVO treats correctness as a continuous fitness signal and co-evolves it with area, delay, and power to generate Verilog that passes verification and synthesizes efficiently.
A specialized 4B Qwen fine-tune writes complete Three.js pages that preview low-poly assets and export them as GLB files ready for Godot.
OpenCode Browse ships an MCP server plus Chromium extension that lets coding agents drive your actual browser, with your logins, tabs, and extensions intact.
A new benchmark called Muses-Bench stress-tests 26 frontier LLMs on multi-user scenarios, revealing how badly they handle conflicting instructions, privacy, and coordination.