ciflow/trunk/179195: Update
[ghstack-poisoned]
Provide feedback
Saved searches
Use saved searches to filter your results more quickly
Sign up
Appearance settings
Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
update
Running a local LLM on Android with Termux and llama.cpp
What I used Samsung S21 Ultra Termux llama-cpp-cli llama-cpp-server Qwen3.5-0.8B with Q5_K_M quantization from huggingface (I also tried Bonsai-8B-GGUF-1bit from huggingface. Although this is a newer model and required a different setup, which I might write about at a later time, it produced 2-3 TPS and I did not find that to be usable) Installation I downloaded the "Termux" app from the Google Play store and installed the needed tools in Termux: pkg update pkg upgrade -y pkg install llama-cpp -y Downloading a model I downloaded Qwen3.5-0.8B-Q5_K_M.gguf in my phone browser and saved it to my device. Then I opened the download folder shortcut in the browser, selected the GGUF file -> open with: Termux Now the file is accessible in Termux. Running it in the terminal After that, I loaded the
Gemma4:26b's reasoning capabilities are crazy.
Been experimenting with it, first on my buddy's compute he let me borrow, and then with the Gemini SDK so that I don't need to keep stealing his macbook from 600 miles away. Originally my home agent was run through Gemini-3-Flash because no other model I've tried has been able to match it's reasoning ability. The script(s) I have it running through are a re-implementation of a multi-speaker smart home speaker setup, with several rasperry pi zeroes functioning as speaker satellites for a central LLM hub, right now a raspberry pi 5, soon to be an M4 mac mini prepped for full local operation. It also has a dedicated discord bot I use to interact with it from my phone and PC for more complicated tasks, and those requiring information from an image, like connector pinouts I want help with. I've
![viable/strict/1775476886: [vllm hash update] update the pinned vllm hash (#179439)](https://avatars.githubusercontent.com/u/133916390?s=60&v=4)
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Releases
Hong Kong developers test homebuyers with modest price increases after sell-outs
Hong Kong developers are raising prices of new homes this week following sold-out launches in recent days, further testing the appetite of homebuyers amid geopolitical and interest rate uncertainties. Henderson Land Development put another 39 units at its Chester project in Hung Hom on sale on Monday, with 25 homes finding buyers, according to agents. With an average discounted price of HK$22,198 (US$2,831) per square foot, the units were priced 4.57 per cent higher than the 123 units that sold...
Why Microservices Struggle With AI Systems
Adding AI to microservices breaks the assumption that same input produces same output, causing unpredictability, debugging headaches, and unreliable systems. To safely integrate AI, validate outputs, version prompts, use a control layer, and implement rule-based fallbacks. Never let AI decide alone—treat it as advisory, not authoritative. Read All
An Empirical Study of Testing Practices in Open Source AI Agent Frameworks and Agentic Applications
arXiv:2509.19185v3 Announce Type: replace Abstract: Foundation model (FM)-based AI agents are rapidly gaining adoption across diverse domains, but their inherent non-determinism and non-reproducibility pose testing and quality assurance challenges. While recent benchmarks provide task-level evaluations, there is limited understanding of how developers verify the internal correctness of these agents during development. To address this gap, we conduct the first large-scale empirical study of testing practices in the AI agent ecosystem, analyzing 39 open-source agent frameworks and 439 agentic applications. We identify ten distinct testing patterns and find that novel, agent-specific methods like DeepEval are seldom used (around 1%), while traditional patterns like negative and membership tes
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!