Some personal news: I'm joining @Box as a Senior Developer Relations Engineer! I’ll be helping developers build AI agents that can search, reason over, and act on company content securely. So, why Box? (1/9)

Releasing a dev video showcasing how to give @NousResearch Hermes agents a company brain with @Box soon! If you thought gbrain for Hermes was cool, now imagine Hermes having the same company context as your teammates and get work done.

I witnessed this dude go in the recording room and come out within an hour with this banger video. If you like how Fable codes, watch how it excels on knowledge work as well.

As DevRel @Box, I can't wait to enlighten developers about bringing AI to their workplace. "96% of organizations agree agents need access to company content — only 36% have connected agents to it at scale." This is the most important gap that needs be filled for AI. Everyone knows AI agents can 10x productivity. But most success AI stories come from individual AI enthusiasts or solo-entrepreneurs. Why? Because giving your AI agent company data introduces a whole different group of problems to solve. What files should these agents have access to? Who's permission level are they following? How can we audit if AI goes rogue? With these questions unanswered, that gap isn't gonna close anytime soon. This is why Box is going all in on leveraging its enterprise grade security features to equip these AI agents with company knowledge in a secure manner. I highly suggest you give the report a read.

Crashing the AI Dev @Workato event tonight. Say hi if you’re here 😁

Anybody built skills for creative work like writing scripts, coming up with video ideas? Something that stops AI from giving me the same old cut off, advertisement-like sentences?

Fable is the first model that had to limit its capability to share with the public. Does this mean further model improvements will be limited to enterprise customers? But is enterprise ready to adopt AI in their knowledge work? I’ve been at @Box for a week now and I am already seeing how significantly different agents behave when they have access to company knowledge.

Enterprises have tolerated unstructured data governance failures for years. The blast radius was manageable because humans were the ones accessing it. Agents are a potentially bigger challenge. Our CISO Heather Ceylan shared why AI agents make your unstructured data problem impossible to ignore, and why you have to build the governed content layer before you deploy the agents—not after.

This makes creating profiles for each agent so much more accessible when you are running hermes on a VPS. No need to SSH in every single time to run those hermes terminal commands!

Also slash commands work on the top level slack channels but not in slack threads. @Teknium

Personal AI agents deal with your stuff like your calendar, your docs, and your preferences. But Team AI agents aren't that easy. Now the agent has to understand permissions, stale docs, conflicting versions, private notes, approval flows, and the fact that half the useful context lives in comments someone wrote 8 months ago. The next big jump for agents probably isn’t just a smarter harness. It’s giving them the right company context without breaking the trust model companies already depend on. @Box is already working with clients to enable this on an enterprise scale.

Hey Claude, I want you to create a loop telling subagents to create loops for automatically creating loops. I'm so on top of this AI trend.

Make your digital brain for AI "work safe" with @Box One way to get your CIO to approve your personal AI assistant for work 👀

The Box MCP server now supports binary file upload and download using signed URLs instead of base64 encoding through the LLM context window. The difference in practice: a 175KB PDF corrupts with base64. A 20MB PDF fails entirely. Box handles both in under 30 seconds with full enterprise governance built in. Read the full technical breakdown here. 👇 blog.box.com/upload-and-down…

Personal AI agents are easy. Team AI agents are where everything breaks. Permissions, stale docs, conflicting workflows, hidden decisions, context spread across 20 tools. The future of enterprise AI won’t be won by the smartest chatbot. It’ll be won by the best context layer.

This shows how collaborating with teams at work is completely revolutionizing by using AI and @Box as the "company brain". Until now agentic workflows have been optimized for personal workflows but more often than not, work involved other people with their own workflows. Shared knowledge base for AI running on different environments is essential for collaborative AI workflows.

I look like a 40y old professor here but this is awesome.