A self-hosted persistent memory system for AI coding assistants, Friday is a FastAPI-based "cognitive memory" layer that plugs into editors and agent CLIs via the Model Context Protocol (MCP). It solves AI "amnesia" by persisting cross-session semantic memories, vector indexes, and a Neo4j knowledge graph so agents like Cursor, Claude, Antigravity, or custom Codex setups can share one unified brain. Core MCP tools include add_memory, add_fact, memory_search, and get_context; an auto-graph engine extracts entities from memories into Neo4j nodes and edges automatically. A live Neural Studio provides an Obsidian-style knowledge graph browser with filtering, CRUD, and an inspector, while layers combine Mem0, ChromaDB-style vector search, and Neo4j for contextual retrieval and structured facts.
Friday emphasizes practical self-hosting and low friction: Docker Compose boots the full stack, .env configuration wires in Mem0 and DeepSeek APIs, and a simple curl POST can store a persistent memory for immediate recall. Claimed benefits include dramatic token cost reduction (~8,000 → ~800 tokens per task) and much higher context-recall accuracy (~30% → ~94%), plus a versioned facts ledger and audit trail. The repo targets developers who want private, centralized memory for multi-agent workflows and supplies step-by-step MCP client configs and example integrations to make a shared brain across local and remote toolchains.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
Researchers developed BITCOS, a new storage layout that reduces the bits per weight in ternary large language models below the traditional 1.58-bit limit. This method improves model compression and inference speed on multiple hardware platforms, achieving up to 1.28 times faster processing. (arxiv.org)
Barndoor AI has acquired Diaphora, the creator of the open-source workflow engine Frags, to enhance governed AI automation for enterprise workflows. The combined platform enables organizations to build repeatable, governed AI workflows that can be securely scaled across teams. (barndoor.ai)
UBTECH has opened a factory in China's Guangxi region capable of producing over 10,000 humanoid robots annually, with one robot made every 10 minutes. The plant uses robots and driverless vehicles for automation, mainly producing the Walker S and Cruzr series. (en.sedaily.com)
A 4-billion-parameter model called Qwen was trained using reinforcement learning to generate query plans that are 81% faster than those by Postgres. The model improves query optimization by learning from performance measurements and adjusting its strategies accordingly. (rohanbansal.com)
A reverse-engineered Jev-like model is available on GitHub, with recent updates and commits. The project includes code, documentation, examples, tests, and licensing information. (github.com)
A fleet of AI agents requires a new type of operating system that manages authority and ownership across multiple agents, rather than a simple harness for individual models. The proposed agent OS must coordinate shared authority over context, tools, execution, and governance to ensure reliable and secure multi-agent operations. (pentad.ai)
Today's best Hacker News stories, summarized and screenshotted, one email a day.