Jul 15, 2026
The Filesystem Your Agent Already Speaks
An argument for building agent harnesses around Unix filesystem conventions the model already understands, and where …
#ai
#agents
#tooling
Newsworthy
Short writeups of interesting found articles and projects related to Framework Labs’ endeavors in Ai, software, and tooling.
Read →
Jul 13, 2026
ECC: 228K Stars for an Agent Harness
Everything Claude Code (ECC) turned one developer's ten months of configuration work into a 228,000-star harness.
#ai
#agents
#tooling
#claude-code
Jul 3, 2026
GraphRAG and the Pyramid I'd Only Half-Built
Microsoft's GraphRAG turns a knowledge graph and hierarchical community summaries into something close to a curated, …
#ai
#rag
#graphrag
#llm
Jun 30, 2026
Private MCP Servers Available to ChatGPT (not Codex or Claude Code)
OpenAI's approach to connecting private MCP servers to hosted AI products without ever exposing them publicly.
#ai
#mcp
#security
#agents
Jun 24, 2026
SproutRAG: The Pyramid Idea, Grown Without an LLM
SproutRAG builds multi-granularity retrieval from attention signals alone, no LLM summarization calls required to get …
#ai
#rag
#research
#llm
Jun 19, 2026
HippoRAG: Retrieval That Works Like Memory, Not a Database
A short look at HippoRAG's associative-memory approach to retrieval, and why skipping the vector database matters more …
#ai
#rag
#llm
May 26, 2026
Agent Vault: The Lock, Not the Key
Infisical's Agent Vault brokers real credentials to AI agents at the network edge, so the agent completes the request …
#ai
#agents
#security
#self-hosted
May 11, 2026
OpenClaw alternative: Hermes Agent
Nous Research's Hermes Agent pauses after complex tasks to write itself new skill files — a look at the Reflective Phase …
#ai
#agents
#llm
#self-hosted
Jan 24, 2026
MemRL: Let the Model Stay Frozen, Let the Memory Do the Learning
MemRL swaps fine-tuning for a self-evolving memory scored by reinforcement learning that runs on CPU.
#ai
#agents
#rag
#llm
Nov 30, 2025
General Agentic Memory: Indexed, Uncompressed Total Recall
A paper that stops compressing agent history into summaries and instead keeps everything, searching it properly at query …
#ai
#agents
#memory
#research
Sep 5, 2025
RAG's Embedding Ceiling: A Proof, Not a Bug
DeepMind proves dense embeddings hit a hard, dimension-bound retrieval ceiling, and their LIMIT benchmark shows it …
#ai
#rag
#llm
#research
Aug 27, 2025
HRM: The Loop Mattered, Not the Hierarchy
A 27-million-parameter model beat trillion-parameter LLMs on ARC-AGI — until ablations showed the 'brain-inspired' …
#ai
#llm
#benchmarks
#research
Aug 23, 2025
Fast Agent: MCP-native Pythonic workflows
A look at Fast Agent, an open-source Python framework built around MCP as the native substrate for agent workflows …
#ai
#mcp
#agents
#tooling
Jul 29, 2025
OxyGent: GUI for model/agent workflows
A look at OxyGent, JD's open-source Python framework that collapses agents, tools, and models into one interchangeable …
#ai
#agents
#multi-agent
#mcp