Brian Baldock
brianbaldock.bsky.social
Brian Baldock
@brianbaldock.bsky.social
Microsoft Senior Program Manager | Cybersecurity, Identity and AI
Pinned
Just published a blog on self-hosting an AI chatbot with Docker, vLLM, ChatUI, and GPU acceleration. Covers the full setup, troubleshooting, and lessons learned.

Check it out here: https://buff.ly/40WeBlT
#AI #Docker #SelfHosting #HomeLab
Deploying Local AI Inference with vLLM and ChatUI in Docker
Building a Self-Hosted Chatbot with an Nvidia GPU and Open-Source Tools
buff.ly
The AI space is finally shifting from single, siloed copilots to connected ecosystems.
Microsoft just launched Magentic, a marketplace where AI agents can collaborate, automate, and scale across industries.
👉 thenewstack.io/microsoft-la...
#MSFTE
Microsoft Launches Magentic Marketplace for AI Agents
Magentic Marketplace is a simulation environment for agentic markets, a new project from Microsoft to explore multi-agent collaboration.
thenewstack.io
November 7, 2025 at 6:06 PM
The hardest part of the LightningCopilot project wasn’t building the LWC.
It was getting Entra ID, Power Platform, and Copilot Studio to all agree on who you are — silently.

Federated credentials, custom scopes, Static Resources instead of CDNs… a dozen small things that had to line up!
November 6, 2025 at 2:31 PM
AI agents are moving from buzzword to baseline.

Most companies plan to make them core to their AI strategy within a year.
Want to learn how to build one the right way?

Check out the Microsoft Copilot & AI Agents Summit (Nov 26–27).
Real examples, and practical frameworks!

#MS-FTE #Microsoft
October 30, 2025 at 3:31 PM
Finally published the LightningCopilot walkthrough, how to run a Microsoft Copilot Studio agent inside Salesforce Lightning with Entra ID SSO + MSAL.
It’s a proper step-by-step, not another “just drop an iframe” post.
October 29, 2025 at 10:37 AM
Remember: agents drift when they reread their mistakes. Trim history, externalize state, feed less back. Reliability rises.
#ThursdayThoughts #AI #LLM #AIAgents
October 16, 2025 at 2:05 PM
Majority voting barely improves long-horizon reliability. Thinking tokens, ie: "reasoning" before acting, win every time. Pay for the right compute.
#ThursdayThoughts #AI #LLM #AIAgents
October 9, 2025 at 2:04 PM
My desk looks like a small tornado passed through, coffee mug, cables, notebooks, and a few rogue spinner magnets that didn’t survive another fidget session.
Some people need calm to focus; I apparently need creative chaos.
Curious, what’s the one item on your desk you can’t live without?
October 8, 2025 at 1:49 PM
Agent build tip: Plan externally, execute in small, verifiable steps. Cleaner pipelines.
#ThursdayThoughts #AI #LLM #AIAgents
October 2, 2025 at 2:04 PM
Entra rollout made easier.
Check out these step-by-step deployment videos for Microsoft Entra, covering hybrid join, Conditional Access, and more. Perfect for IT admins in the trenches.
October 2, 2025 at 1:00 PM
Ghost citations, part 2: my 7-step checklist to keep AI-assisted content faithful, accurate, and complete. Move fast—without burning trust.
Link in first reply.
September 29, 2025 at 1:04 PM
Drift happens when agents read their own bad outputs. Cut context, watch error rate, escalate sooner. Human in the loop!
#ThursdayThoughts #AI #LLM #AIAgents
September 25, 2025 at 2:04 PM
Step 1: Create the PIM‑enabled group.
Entra → Group → Enable PIM → add Eligible Member assignments (time‑boxed).
Full guide (pics): aka.ms/defpim
#EntraID #PIM #LeastPrivilege
September 23, 2025 at 2:03 PM
Quick reminder — the Entra ID snackable video series is live, thanks to a big team effort across FastTrack and several partner groups at Microsoft.

They’re short, practical sessions covering everything from Conditional Access to Token Protection. I pulled them into one blog, link below.
September 22, 2025 at 2:06 PM
An “ethical AI” report shipped with 15+ fake citations. Why? Hallucinations + humans pasting smart-sounding lines. I share a simple playbook to stop ghost citations and keep trust.
Link in first reply.
September 19, 2025 at 2:45 PM
💭 Track horizon length, not just accuracy. Tiny per‑step gains can mean big jumps in how far agents run without help.
#ThursdayThoughts #AI #LLM #AIAgents
September 18, 2025 at 2:02 PM
Still giving standing access in Defender XDR?
Use PIM‑for‑Groups + URBAC for JIT, time‑boxed permissions.
Clean guide from Matt Novitsch → aka.ms/defpim
September 17, 2025 at 2:02 PM
The FastTrack, Product Marketing, GTM, WWL Studios, and Microsoft Learn teams built out a library of Entra ID “snackable” deployment videos. These are short, focused sessions covering everything from Internet Access to Identity Governance and Privileged Identity Management.

#Microsoft365 #EntraID
September 16, 2025 at 2:04 PM
Quick reminder: I built a PowerShell mini-module to simplify Viva Engage policy management (Copilot + AI Summarization).
Org-wide disable → group enables.
buff.ly/lfgVxuh
repo in the first comment 👇
#LazyAdmins #Microsoft365 #M365Copilot #VivaEngage #PowerShell
Manage Copilot in Viva Engage
Learn how to manage access policies for Copilot in Microsoft Viva Engage with this comprehensive admin guide and mini PowerShell module
blog.brianbaldock.net
September 15, 2025 at 10:06 AM
Hey #LazyAdmins 😎 I put together a PowerShell mini-module to manage Viva Engage access policies for Copilot + AI Summarization.
The flow: start with an org-wide disable, then enable for groups you choose. Clean, predictable, repeatable. blog.brianbaldock.net/admin-guide-...
Manage Copilot in Viva Engage
Learn how to manage access policies for Copilot in Microsoft Viva Engage with this comprehensive admin guide and mini PowerShell module
blog.brianbaldock.net
September 4, 2025 at 3:57 PM
When was the last time you looked at your Microsoft Secure Score?
Here’s why I treat mine like a kitchen cleaning list, and how you can too.
blog.brianbaldock.net/securescore
#CyberSecurity #Microsoft365 #SecureScore
A Cook’s Guide to Microsoft Secure Score: Continuous Security
Enhance security with Microsoft Secure Score: a prioritized checklist for continuous improvements in identity, devices, apps, and data
blog.brianbaldock.net
August 18, 2025 at 1:50 PM
When I worked in kitchens, we had a rule: “Always be cleaning.”
Now I use the same mindset in security with Microsoft Secure Score.
New post: blog.brianbaldock.net/securescore
#CyberSecurity #Microsoft365 #SecureScore
A Cook’s Guide to Microsoft Secure Score: Continuous Security
Enhance security with Microsoft Secure Score: a prioritized checklist for continuous improvements in identity, devices, apps, and data
blog.brianbaldock.net
August 11, 2025 at 9:22 PM
ICYMI: My latest post on identity design for AI agents is still fresh.
OBO flow vs service creds, secrets, and audit trails, all covered.
Still relevant. Still useful.
blog.brianbaldock.net/recipe-desig...
#CyberSecurity #AgenticAI
Creating Identity for agentic AI
Learn best practices for secure identity management in agentic AI systems, ensuring accountability and zero trust in dynamic, scalable environments
blog.brianbaldock.net
July 29, 2025 at 10:06 AM
New post: Designing Identity for Agentic AI.
How do agents securely access systems like SQL Server?
I break down two flows (OBO + service creds) + give you a practical checklist.
Check it out 👉 blog.brianbaldock.net/recipe-desig...
#ZeroTrust #AIIdentity
Creating Identity for agentic AI
Learn best practices for secure identity management in agentic AI systems, ensuring accountability and zero trust in dynamic, scalable environments
blog.brianbaldock.net
July 22, 2025 at 10:06 AM
Last week I shared what breaking into my own systems taught me as a defender. From missed file permissions to my first reverse shell—it was eye-opening.

What moment shifted your view on defense?

Catch the full post here: blog.brianbaldock.net/a-better-def...
July 14, 2025 at 3:04 PM
New blog just dropped: How breaking in made me a better defender.
Offensive security taught me more than I expected; reverse shells, overlooked basics, and how fragile assumptions really are.
What changed your view on defense?
🔗 blog.brianbaldock.net/a-better-def...
July 7, 2025 at 10:02 PM