Ornith-1.5, a family of open-weight foundation models built on Qwen3.5 and Gemma4, trains through a self-improvement loop that jointly generates tasks, constructs scaffolds, and runs reinforcement lea…
ZAI released GLM 5.3, a post-trained coding model built on GLM 5.2's architecture, and Zcode, a dedicated agentic development desktop app designed to run GLM models. Zcode features project indexing, M…
Block released Bird, a free, open-source desktop app that unifies multiple AI coding agent harnesses—including Claude Code, Codex, GitHub Copilot, Amp, Cursor agent, and Goose—into a single interface,…
Lovable's no-code AI platform enables non-technical builders to create a text-only, ad-free news aggregator website called Decaf News, which goes from idea to live URL in an afternoon without writing …
Lovable, Replit, Bolt, Codex, and Claude Code each fit different software shapes, and choosing based on the project's structure—not AI trends—is critical, according to a comparison guide. Lovable offe…
DeepSeek Harness, an open-source coding agent framework built entirely from plugins, gained roughly 165,000 GitHub stars within a week of release, signaling strong developer interest in a customizable…
Docker has introduced Docker Sandboxes, isolated micro VMs designed to safely run AI coding agents like Claude Code and Codex, with hardware-enforced isolation, network policies, credential proxying, …
A new framework from AI builders categorizes personal software projects into five shapes—local tools, web apps, native apps, background services, and hardware projects—and advises choosing the shape b…
Zhipu AI's GLM 5.3 large language model scored 28.3 on terminal-bench 3, up from GLM 5.2's 4.6, without changing base weights, according to a hands-on review. In live tests, GLM 5.3 added a real-time …
Zhipu AI's GLM 5.3 is available only via API, with open weights delayed by roughly two weeks due to the model's strong cybersecurity capabilities, according to the company. The model, which retains th…
Lovable is an AI platform that generates fully functional, hosted applications—including front end, database, user authentication, and a live public URL—from plain-English descriptions in a chat inter…
Deep Reinforce's Ornith 1.5 35B-A3B, a mixture-of-experts language model activating about 3 billion parameters per token, outperforms Qwen3.6-35B-A3B on every published coding and agentic benchmark, s…
Empero released Qwen3.8-9B Distill, a full-parameter fine-tune that transfers reasoning from the Qwen3.8 2.4T A95B teacher into the Qwen3.5-9B architecture, trained on roughly 70,000 curated chain-of-…
Zed Delta, a standalone AI coding agent app from Zed, pairs an AI agent with Delta DB, a version control layer that records every code change and links it to the conversation that produced it, offerin…
Deep Recurse released Ornith 1.5 35B-A3B, a mixture-of-experts language model with 35 billion total parameters and 3 billion active per token, built on a Qwen3.5-MoE architecture for coding, reasoning…
Ornith 1.5 9B, a 9-billion-parameter dense language model from ornith-ai, failed real-world agentic and coding tasks in a local test on a single Nvidia A100 GPU, despite benchmark claims that it rival…
Tether released Qwak, a free, open-source local AI platform that bundles text generation, retrieval-augmented generation (RAG), fine-tuning, image and video generation, and speech into a single offlin…
Alibaba's Qwen team released Qwen 3.8 27B, a dense 27 billion parameter multimodal model that scored 51 on the Artificial Analysis agentic index, trailing only Kimi K2, a 2.8 trillion parameter model.…
LM Studio's local LLM performance hinges on whether users run MLX on Apple Silicon Macs or GGUF on Windows GPU laptops, with Apple's unified memory enabling larger models while discrete GPUs like the …
Alibaba's Qwen 3 27B, a dense open-weight multimodal model, can be run locally using DeepSeek Harness, achieving agentic performance close to Claude 4.5 on benchmarks. On a two-node NVIDIA DGX Spark c…