A developer built NatureLens, a mobile-first PWA and desktop companion that turns a phone's camera and GPS into an interactive outdoor field guide. The app sends photos to Gemma 4 E4B running locally …
A developer built "Can You Find It?", an open-source game in which a local Gemma 4 E4B vision model running via Ollama on a laptop secretly picks one real object in a photo of a familiar place and iss…
JetBrains released Mellum2.1, a 12B mixture-of-experts coding model with 2.5B active parameters under the Apache 2.0 license, trained primarily with reinforcement learning across millions of sandboxed…
NVIDIA announced that new compact open models, including Nemotron 3.5 Lightning and Qwen3.8-27B, can now run reasoning and agentic AI workloads on Jetson edge devices, eliminating the need for data ce…
Perplexity launched Hybrid Compute on Perplexity Computer, splitting one task across a cloud model for search and reasoning and a local model on Apple silicon Macs for private files, with an orchestra…
Perplexity launched Hybrid Compute on Sep 2, 2026, a feature that splits tasks between cloud models and local AI to protect private data and reduce costs. Available in the Perplexity app for Apple Sil…
On September 1, 2026, Perplexity released Hybrid Compute for its Mac app, which splits tasks between the cloud and the user's Mac based on data sensitivity, with sensitive steps running locally via an…
Researchers introduced the Hybrid Hierarchical Multi-Agent Framework (H²MAF), which fuses EfficientNet-B3 and ConvNeXt-Tiny with multimodal large language models Gemma 4 E4B and Qwen3.5 4B to provide …
Models with fewer than 1 billion parameters, such as LFM 2.5 350M, exhibit unreliable tool calling in agentic systems, prompting developers to implement repair mechanisms and prompt-checking systems i…
Solo indie developer from Japan upgraded the local LLM powering the commercial game Hex Judge from Gemma 2 9B to Gemma 4 E4B, cutting AI conversation startup time on an RTX 2060 from about 45 seconds …
A developer testing MCP server tool calls in Rust found that combining 7 related functions into 1 function with a `command` argument is superior, as small models like LFM and Ling Tiny perform equally…
A developer exploring local AI tasks found that writing custom code to handle natural language queries would be faster than LLM inference but would fail on complex prompts due to the inherent complexi…
Adding relative date support to an MCP server required about four new tool calls with detailed descriptions, increasing prompt tokens to over a thousand and highlighting a tradeoff between agentic AI …
A developer experimenting with local-LLM agentic AI workflows found that combining tool calls—such as having a single `read_blog_post` MCP tool internally perform `find_blog_date`—reduces the number o…
A developer reports greater success and enjoyment using agentic AI with small local models, such as Gemma 4 E4B via llama-cpp and an MCP server, for disliked tasks like website management, compared to…
A developer created a local MCP server in Rust to run commonly-needed commands via an AI agent, arguing that the approach offers error correction, command chaining, and an improved browser-based UI ov…
Alibaba's Qwen3.6 35B A3B scores 32 on Artificial Analysis's Intelligence Index v4.1, outperforming Google's Gemma 4 26B A4B at 26, with Qwen winning 18 of 22 evaluations tested. However, Gemma 4 is 6…
Cue, a voice-activated AI agent, cut latency by 44%, increased dictation usage by 30%, and eliminated marginal inference costs by running Google's Gemma 4 E4B model locally for text polishing, with me…
Researchers propose RepSelect, a method for robust LLM unlearning that isolates forget-set-specific representations by collapsing top principal components of weight gradients, achieving 4-50x larger r…
A developer has completed PuskesmasAI, an offline-first Progressive Web App that uses AI-powered medical triage to assist community health workers in rural Indonesia. The app runs entirely on-device w…