PulseAugur
EN
LIVE 01:40:36

User builds fully offline coding agent with Qwen3.6-35B-A3B on laptop

A user has developed a fully offline, local coding agent utilizing the Qwen3.6-35B-A3B model and llama.cpp, running entirely on their personal computer without any cloud dependencies. This agent features persistent memory through a combination of a vector database, knowledge graph, and a diary, creating a dynamic map of the codebase. It also includes modular skills such as push alerts, browser control, and local text-to-speech, with the agent retaining memory of decisions made over the past six months, all managed on a laptop GPU. AI

IMPACT Demonstrates the capability of running sophisticated AI agents locally on consumer hardware, reducing reliance on cloud infrastructure.

RANK_REASON User-developed tool utilizing open-source components.

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

User builds fully offline coding agent with Qwen3.6-35B-A3B on laptop

COVERAGE [1]

  1. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    I bloated my coding agent until it became a local ecosystem. Qwen3.6-35B-A3B on llama.cpp, fully offline, zero cloud, nothing leaves my PC. 🧠 Persistent memory:

    I bloated my coding agent until it became a local ecosystem. Qwen3.6-35B-A3B on llama.cpp, fully offline, zero cloud, nothing leaves my PC. 🧠 Persistent memory: vector DB + knowledge graph + diary 🗺️ A living map of my codebase I grep before opening files blindly 🔔 Push alerts · …