Three dates to have on your radar:
📅 May 17 — GPT-5.3-Codex auto-rollout
📅 v1.125 — Edit Mode fully removed
📅 Weekly — VS Code now ships stable releases every week
Full breakdown on the blog: bighatgroup.com/blog/copilot-weekly-2026-03-23/
VS Code 1.112 highlights:
• MCP server sandboxing (limits blast radius of local tool servers)
• /troubleshoot command for debugging agent logs
• Autopilot mode now in Stable
• Edit Mode deprecated — full removal in v1.125
Weekly release cadence means this pace continues.
GPT-5.4 mini hit GA across all tiers — Pro, Pro+, Business, Enterprise.
OpenAI's fastest agentic coding model. Best time-to-first-token. 0.33x premium request multiplier.
Enterprise admins: you need to explicitly enable it in Copilot settings.
GPT-5.3-Codex is GitHub's first LTS model — guaranteed through Feb 2027.
⚠️ Auto-rollout deadline: May 17, 2026. All Business/Enterprise orgs get it as the base model unless you opt out.
Start your internal model review now.
The big one for enterprise governance: every coding agent commit now has an Agent-Logs-Url trailer.
Click any agent-generated commit → full session log showing what it did, why, and what tools it used.
The audit trail that was missing is here.
Configurable validation tools: admins can now enable/disable CodeQL, Advisory Database, secret scanning, and Copilot code review per-repo.
All free. All enabled by default. No Advanced Security license needed.
Granular security control without slowing down teams.
Startup is 50% faster across all invocation methods — issue assignment, Agents tab, and @copilot mentions in PR comments.
Makes the agent viable for smaller tasks where a 60-second cold start was hard to justify.
The coding agent now finds code by meaning, not just text matching.
2% faster task completion, no config needed. For large monorepos where naming conventions vary across teams, this is a real improvement in context discovery.
Biggest week for GitHub Copilot's coding agent since launch.
5 updates in 4 days — semantic search, 50% faster startup, configurable validation, live logs, and commit-level audit trails.
Plus GPT-5.3-Codex LTS, GPT-5.4 mini GA, and VS Code 1.112.
Thread 🧵
#PSADT and #AI combine to make packages. This MCP server can do more but I had to sink quite a bit of time working out the hallicinations and other bugs. Version 0.8 building packages below.
Learning how to launch Amiga software from a PiStorm using Caffeine OS & WHDLoad. The left monitor shows a game running on the Amiga with video out through RGB. One on right is OS running on the Raspberry Pi with HDMI out. #amiga #amiga500 #pistorm #retrocomputing
December 10, I will talk about #Windows Update for Business (WUfB) and #Autopatch. Join me online!
www.meetup.com/calgary-micr...
December 10, I will talk about Windows Update for Business and Autopatch. Join me online!
www.meetup.com/calgary-micr...
I've got a company where we've mainly been in stealth mode but things are coming to market. One of our experiments was to teach AI about patching and automate some reporting and even a patching podcast. Come see what PortalFuse has been up to.
portalfuse.io/blog/portalf...
Brian goes through the history of EUC and does a great job of positioning it in the world of AI and Microsoft's larger strategy around AI from the desktop to the cloud.
I consider watching this as time well spent.
Not sure if this happens with physical hardware but #Intune Autopilot with Hyper-V and two NICs is troublesome. On my Windows 24H2 laptop with 24H2 guest, I could not get an initial Internet connection to deploy. Problem was Autopilot only checked the first NIC and not the other. 🤪
I've been quiet but super happy about how many of us are reconnecting here. It's so refreshing to have a feed with content you want and not be bottled to death. I swear my Web3-only X profile probably has at least half of my followers as bots.
