Grok Build 업데이트·에이전트 메모리 보안·오픈소스 에이전트 도구 동향

Elon Musk

@elonmusk

Try Grok Build

Andrew Milich

@milichab

Big updates next week :)

// Bad Memory in Agents // (bookmark it) Persistent memory is what makes an agent feel useful across sessions. It is also a place an attacker can leave something behind. New work tests prompt injection against memory in Claude Code and OpenAI Codex, across different models. Tricking an agent into overwriting its own memory from untrusted external content is genuinely hard. Payloads already sitting in those files attack current and future sessions just fine. Strength helps on some goals and not others. Opus 4.7 and GPT-5.5 never allowed credential exfiltration (0% attack success). Unauthorized tool use landed with high success across almost every model tested. One planted rule quietly pins a pip install to PyYAML 5.3.1, a version with known vulnerabilities, during routine environment setup. Why does it matter? An injection no longer has to win the session it arrives in. It has to get written down once and wait. Defenses need to protect memory writes without killing the adaptation that makes memory worth having. Paper: https:// arxiv.org/abs/2607.14611 Learn to build effective AI agents in our academy: https:// academy.dair.ai

LangChain

@LangChain

We've built a full software engineering agent factory in house, and open-sourced every piece of it. @BraceSproul 's blog dives into each of those components, what they're used for, and why them being OSS matters

Sharing my new skill! It keeps track of high-signal X accounts for top AI news, papers, projects, etc. Total gamechanger for me. Built with X MCP tools. Give your agent the skill and tell it to generate the artifact with top stories. Works for Codex, Claude, Hermes, OpenClaw, or whatever you use. 3 steps: 1. Set up X MCP - X API: https:// docs.x.com/tools/mcp#x-mc p-%E2%80%94-x-api … 2. Install skill here: https:// github.com/dair-ai/dair-a cademy-plugins/tree/main/plugins/x-agent-intelligence … 3. Run prompt: "Use the x-agent-intelligence skill to build a self-contained local feed from my X MCP connection; ask for my source handles if needed, save feed.html, and validate it." It should generate a nice, beautiful HTML artifact like the one shown in the clip. You can tune it however you want. You can then set a schedule/automation to do this daily or whatever cadence you prefer. I have it every 4 hours. You will need to curate the X accounts yourself, but I have shared a few good ones under the assets. You can ask your agent to tune it to however you like. I have also shared my personal feed with our community here: https:// academy.dair.ai/dashboard/reso urces/x-agent-intelligence … I understand if it gets tricky to set up. Please reach out to me in the community forum. I plan to do a little tutorial or live session soon to help others reproduce the process. You can also store the feed as a wiki, as I have in my own implementation, but that's optional. If you encounter any issues or have ideas on how to improve it, please open a PR.

"The debate between LLM-Wiki and Proxy-Pointer is not really about retrieval. It is about when semantic understanding should occur." Partha Sarkar breaks down key differences between two approaches to enterprise RAG.

"Each upgrade earned its place against a concrete failure of the baseline: the typo that missed the page, the TOC the keyword search ignored, the list flattened into prose." Angela Shi walks us through concrete ways to improve the performance of a RAG pipeline.

Almost half capacity You guys show up every time we ask That's why Local & Opensource AI is going to win See you next Saturday