Explainer: Secure Boot Chain of Trust

Here’s an uncomfortable, seldom considered truth: your operating system isn’t the first thing that runs when you power on your PC. The firmware goes first. Then the bootloader. Then the OS kernel. Malware creators figured this out a long time ago. Get in early enough — before the OS loads — and you can own a machine invisibly, surviving reboots, reinstalls, and even antivirus scans. All this explains why the secure boot chain of trust is vital to modern Windows security.

The threat is real and it’s present right now. BlackLotus, a UEFI bootkit sold on criminal forums, made headlines in 2023 for bypassing Secure Boot on fully patched Windows 11 systems. BootHole exposed a critical flaw in GRUB2’s boot process that affected both Linux and Windows. PKFail (2024) revealed that dozens of device vendors had shipped products using a leaked “do not ship” test Platform Key — meaning the root of the entire trust hierarchy was compromised straight out of the box. Then, in January 2025, ESET researchers disclosed CVE-2024-7344: a Microsoft-signed UEFI recovery application that could silently load unsigned bootkit code — on any UEFI system, regardless of whether Secure Boot was enabled. Microsoft pulled the vulnerable binaries in the January 14, 2025 Patch Tuesday update.

Boot-time attacks aren’t theoretical. They’re happening. Under-standing Secure Boot’s chain of trust is the first step toward knowing whether your defenses are actually holding.

Understanding the Secure Boot Chain of Trust

Think of the chain of trust as a series of checkpoints at the border. Each checkpoint must vouch for the next before anything is allowed through. No vouching, no entry, and the boot process stops dead.

In technical terms: every component in the boot sequence verifies the digital signature of the next component cryptographically before handing off execution. The firmware checks the bootloader. The bootloader checks the OS kernel. The kernel checks drivers. If any link in that chain can’t be verified — wrong signature, no signature, a signature that’s been revoked — the process stops. Your PC refuses to proceed rather than run untrusted code. That’s the whole point. Always safe means never sorry, even if it also means a PC that won’t fire up and run.

The chain only works, of course, if the first link is trustworthy. That’s where the UEFI key hierarchy comes in.

The Key Players: PK, KEK, db, and dbx

UEFI Secure Boot manages trust through four interlocking databases baked into your firmware. Get familiar with them — they come up constantly whenever something goes wrong at boot time.

Key / Database Full Name Role
PK Platform Key Root of trust. Set by the hardware manufacturer. Controls who can update KEK.
KEK Key Exchange Key Authorized to update the signature databases (db and dbx).
db Signature Database Hashes and certificates of trusted bootloaders allowed to execute.
dbx Forbidden Signatures Database Revoked signatures and hashes. Anything here is blocked unconditionally.

The PK sits at the top. Your motherboard manufacturer owns it. Below the PK, the KEK authorizes who gets to update the lists of trusted and forbidden signatures. In practice, Microsoft functions as the de facto Secure Boot Certificate Authority for the consumer PC ecosystem. Nearly every machine you buy ships with Microsoft’s certificates pre-loaded in db — exactly why CVE-2024-7344 was so broadly dangerous. A legitimately Microsoft-signed binary became a usable attack vector!

Worth Knowing: PKFail and the Test Key Problem

In 2024, the PKFail vulnerability revealed that over 200 device models from multiple vendors shipped with a Platform Key originally marked “DO NOT TRUST” — a sample key from AMI’s reference firmware that was never meant to leave the lab. When your PK is public knowledge, the entire root of trust collapses.

How the Chain Is Created at Boot Time

Power on your PC, and here’s what actually happens — fast, invisible, and mostly taken for granted.

  1. The UEFI firmware initializes hardware and activates Secure Boot mode.
  2. The firmware reads the bootloader from the EFI System Partition and checks its signature against db. It also checks against dbx — if it’s there, execution stops immediately.
  3. The signed bootloader (Windows Boot Manager, for example) takes over and verifies the OS kernel’s signature using its own embedded certificates.
  4. The kernel loads and verifies signed drivers. On Windows, this is enforced through Driver Signature Enforcement — unsigned kernel-mode code is blocked by default.

Every handoff is cryptographically verified before it happens. Compromise any link — plant an unsigned binary, exploit a signed-but-vulnerable loader, sneak past a misconfigured dbx — and an attacker owns your machine below the OS waterline. That’s precisely the attack surface that BlackLotus, BootHole, and CVE-2024-7344 each exploited in different ways.

Maintaining a Strong Chain of Trust

Secure Boot isn’t a “set it and forget it” control. Maintaining a healthy chain of trust requires ongoing attention from both Microsoft and from you.

The most important maintenance lever is the dbx — the blocklist. When a bootloader is found vulnerable (as happened with a batch of 2011-era Microsoft-signed binaries in 2023, and again with the CVE-2024-7344 binaries in January 2025), Microsoft issues dbx updates through Patch Tuesday. Your firmware then refuses to execute those specific binaries even if they’re somehow placed on the system. Keeping Windows Update current is how those revocations reach your PC.

Firmware updates matter just as much. Vulnerabilities in the UEFI firmware itself require OEM-supplied updates delivered via Windows Update or manufacturer tools. The NSA and CISA have both issued guidance recommending that organizations periodically audit their Secure Boot configuration — confirming the correct keys are enrolled, the dbx is current, and no rogue Platform Keys are in place (a lesson PKFail drove home hard).

Complementing Secure Boot is the TPM’s Measured Boot capability. While Secure Boot enforces what can execute, Measured Boot records cryptographic measurements of everything that did execute into TPM Platform Configuration Registers (PCRs). Remote attestation tools can then verify those measurements after the fact. Think of Secure Boot as the bouncer at the door; Measured Boot is the security camera logging who actually got in.

Why the Chain of Trust REALLY Matters

Secure Boot isn’t perfect — BlackLotus, BootHole, PKFail, and CVE-2024-7344 all proved that. But “not perfect” is a long way from “useless.” It raises the cost and complexity of boot-level attacks significantly, and when the ecosystem keeps the revocation databases current, it closes known attack paths quickly.

Do yourself a favor: open System Information (msinfo32), find BIOS Mode (should read UEFI) and Secure Boot State (should read On). If either is wrong, fix it. Keep your firmware updated. Keep Windows updated. The chain of trust is only as strong as its weakest, most-neglected link — and that link is usually sitting right between the keyboard and the chair. Here in Windows-World keeping track of key security concerns is darned important. The Secure Boot chain of trust should be at the top of everyone’s list.

Facebooklinkedin
Facebooklinkedin

Intel DSA Remains Driver Install Clickmeister

I just realized that DSA was MIA on my ThinkPad X12 Gen 1 Detachable Tablet. So I installed it, then ran it. It found 3 drivers in need of updates on that device: Wi-Fi, Bluetooth, and (Xe) Graphics. In updating them, I observed that the  Intel Driver and Support Assistant (Intel DSA) remains driver install clickmeister supreme. Let me explain…

Why say: Intel DSA Remains Driver Install Clickmeister?

It’s long been my observation that using DSA requires lots of mouse clicks. This time around, installing the three drivers shown in the lead-in screencap required at least 24 mouse clicks. For the record, those drivers were (numbers at right count clicks for each one):

  • Wireless Bluetooth Drivers (9)
  • 11th-14th Gen Processor Graphics (10)
  • Wi-Fi Drivers (5)

This time around it actually took me 4 additional mouse clicks to get from item 2 to item 3, because I was installing the GPU driver for the first time. Thus, I had to reboot my system, because DSA got “stuck” on “installing” for item 2, and wouldn’t advance to item 3. Sigh. I didn’t count those “extra” clicks in my reported total.

Achieving Intel Driver Update Silence

Believe it or not, that’s the title of a blog I posted on April 27, 2023. That was another time when the sheer number of clicks involved in running DSA hit me hard. It remains noticeable. Today, it struck me as excessive. So I’m formulating this plea to the Intel DSA developers:

Please add a silent mode switch to DSA. Let users tell the tool to run the installs without requiring minutes of babysitting to get through routine maintenance.

I wonder if anybody is listening. Then, I wonder if they’ll respond. Here in Windows-World the silence can sometimes be deafening. Let’s see what happens, shall we?

 

Facebooklinkedin
Facebooklinkedin

Windows Defender May Delete PowerShell Scripts…and More!

Here’s a fun way to start a Monday: you fire up a PowerShell script you’ve run many times — maybe it provisions a batch of AD accounts, maybe it sweeps stale GPOs — and it simply vanishes. No error dialog. No event log entry. Quarantine warnings not provided, either. The file is just gone, like it offended someone. Which, as it turns out, it did.

The culprit? Recent changes to Microsoft Defender’s Attack Surface Reduction (ASR) rules — specifically, tightened enforcement arrived with Windows 11 23H2. And it has only grown more aggressive in 24H2/25H2. If you manage Windows endpoints for a living, this one deserves some notice.

How and Why Windows Defender May Delete PowerShell Scripts

Microsoft has been steadily ratcheting up ASR rules over the past couple of years. Two rules in particular have become dramatically more assertive: “Block execution of potentially obfuscated scripts” and the newer “Block execution from known script interpreter paths” (rule GUID 9e6c4e5a-1037-4377-92f4-2db0f7e629e7). The latter now matches elevated execution paths that have nothing to do with user shell startup, which means your perfectly legitimate admin scripts can get caught in this net.

Here’s the insidious part. Starting with the 23H2 and 24H2 Defender sensor updates, script-blocking ASR rules are now enforced at the kernel driver layer (via WdFilter.sys, Defender’s minifilter drive) — before process creation even occurs. That means scripts launched via WMI, COM+, or scheduled tasks can be silently killed or quarantined without generating an event log entry. You get no breadcrumbs. The script just doesn’t run, and the script file itself may disappear.

This has caused a wave of false positives hitting legitimate PowerShell scripts, SCOM monitoring agents, Active Directory management tools, and enterprise deployment scripts. If you experienced déjà vu reading that, you’re not wrong. In January 2023, a faulty Defender signature update (builds 1.381.2134.0 through 1.381.2163.0) caused the “Block Win32 API calls from Office macro” ASR rule to go haywire and mass-delete Start menu and taskbar shortcuts across enterprises. Microsoft had to ship a dedicated recovery script (AddShortcuts.ps1) and a taskbar repair utility to clean up the mess. Consider this the sequel — quieter but just as disruptive.

How to Recover Deleted or Quarantined Files

If Defender has eaten your scripts, don’t panic. Work through these steps in order:

  1. Check Defender’s quarantine via the GUI. Open Windows Security → Virus & threat protection → Protection history. Filter by “Quarantined Items.” If your script is there, select it and choose Restore.
  2. Browse the quarantine folder directly. Quarantined files live in C:\ProgramData\Microsoft\Windows Defender\Quarantine. They’re encrypted, but they show that Defender took them.
  3. Use PowerShell for deeper inspection. Run Get-MpThreatDetection and Get-MpThreat to list recent detections and see exactly which ASR rule fired. To restore from the command line, use MpCmdRun.exe -Restore -ListAll followed by MpCmdRun.exe -Restore -Name <ThreatName>.
  4. Add targeted exclusions. Use Add-MpPreference -ExclusionPath “C:\Scripts” or configure per-rule exclusions via Intune or Group Policy to prevent recurrence.
  5. Restore from backup. If the file is gone from quarantine entirely, fall back to File History, system restore points, or your backup solution of choice.
  6. For enterprise environments: check the Microsoft 365 Defender portal’s quarantine and Action Center — detections from managed endpoints often surface there even when local logs stay silent.

That leads to what I’ll call a “Pro tip” for admins to consider. Before enabling any new or aggressive ASR rule, set it to Audit mode first (value 2) rather than Block mode (value 1). Audit mode logs what would be blocked without actually deleting anything. Run it for a week or two, review the results in Event Viewer under Microsoft → Windows → Windows Defender → Operational (Event IDs 1121 and 1122), and then flip to Block. This single practice would have prevented most of the heartburn described above.

You Win Some, You Lose Some…

Let me be clear: Defender’s tighter ASR rules are genuinely good for security. Blocking script execution at the kernel level before a process even spawns is a meaningful defense against fileless malware and living-off-the-land attacks. But Microsoft badly needs to improve logging transparency when scripts get blocked at the kernel driver layer. Silent enforcement with no audit trail isn’t “defense in depth” — it’s “debugging in the dark.”

Until that gets fixed, the playbook is straightforward: keep good backups, audit before you block,  and test ASR changes in a staging ring before pushing to production. Remember: your antimalware solution is only as smart as its latest signature update. As the January 2023 shortcut debacle proved, even Microsoft’s own rules can bite the hand that feeds them. I think these just bit me. Don’t let it happen to you!

But Wait! There’s More…

In my usual ElevenForum readover this weekend, I stumbled on a thread that mentioned scripts — and an encrypted password file — disappearing from the poster’s Windows 11 PC. As I responded to that thread “This is deeply disturbing.” It just doesn’t seem right that Defender can cause scripts (and more) to vanish via rule enforcement. You need to steer around this pothole until it gets filled. Not an unfamiliar strategy, alas, here in Windows-World.

Facebooklinkedin
Facebooklinkedin

Revo Roots Out Relics

I’ve been meaning to do this for a while. But this morning, I found the fabled “round tuit” for an app clean-up on Flo6 using Revo Uninstaller. Using that tool, I reduced my count of installed apps from 95 to 83, eliminating an even dozen items. When I claim that “Revo roots out relics,” I’m claiming that the program helps stamp out no-longer-needed (or relevant) apps quickly and easily. Let me offer some details, and an explanation…

How Revo Roots Out Relics…

The intro screencap shows a partial list of all apps installed on Flo6. When I started this clean-up adventure, I was mostly beset with two sets of relics:

  • Leftovers from the ASRock B550 Extreme4 motherboard, which I replaced with an MSI MAG Tomahawk B500 MAX in January (3/12)
  • Leftovers from the Creative Sound Blaster AE-7 I installed earlier this month, but couldn’t get to working (5/12)

The other items were a hodge-podge of odds’n’ends including:

  • AIDA64, yet another system information tool that I don’t even remember installing, and never use
  • Angry IP Scanner: an alternative to Advanced IP Scanner that I tried a few times, before switching back to Advanced…
  • CPU-ID: I don’t need the plain-vanilla one any more, because MSI provides a customized version for the MSI MAG Tomahawk
  • CrystalDiskMark 8.4.0 still installed on Flo6, even though I’m running version 9.0.2. Removed it.

That’s it. Subsequent disk cleanup on Flo6 recovered 6 GB of disk space, too.

App Cleanups Should Happen Periodically

I consider this sort of review and removal part of a good Windows PC hygiene regime. Today was my day to clean up old apps. I’m glad I did. I’ll probably do it again at summer’s end, as I tend to pick up detritus like this over time. Here in Windows-World, if you don’t need it, or can’t use it, why keep it? Out it goes!

Facebooklinkedin
Facebooklinkedin

Canary Jump Sows Predictable Chaos

After recently clean installing the 2021-vintage Lenovo ThinkPad X12 Detachable Tablet Gen 1 I decided to leave it running a production build.  That means I needed a new Canary channel test machine. So this morning, I upgraded the 2025-vintage ThinkPad P16 Gen 3 to that Insider Preview level. Unsurprisingly, this “Canary jump” sows predictable chaos. Let me tell you what happened, and what I did to recover…

Why Say: Canary Jump Sows Predictable Chaos?

No sooner did I reboot into Canary Build 26300.8246 than did all hell start breaking loose. As is my usual practice, I remoted into that PC from my Flo6 desktop — but not for long. In under less than a minute the PC crashed, and threw a slew of interesting “Critical events” as it went down. You can see them depicted in the lead-in screencap.

Of the 6 items in that list, numbers two through five are relevant. That’s because all cluster in the same minute (11:02 AM) and all are related to my remote session failing, then Windows crashing. The X-Rite Color event simply reflects the program’s unhappiness with running in a remote session (it appears again as item 6, when I start my next remote session).

The others are worth visiting in a little more detail:
• Windows stopped working comes from a bugcheck. Copilot tells me this is most likely owing to firmware or driver issues between the laptop and this bleeding edge release
• This provoked the “not properly shut down” as Windows crashed
• It culminates in the “shut down unexpectedly” as Windows turned itself off
At the start of this sequence an illegal memory reference kicked things off. This is why firmware and drivers are suspects: they’re the most likely perpetrators of such untoward acts.

Chaos Cleanup on Aisle 7!

I now understand my cleanup, had it been performed before upgrading to Canary, might have prevented the crash that occurred. I visited Lenovo Vantage, found a new UEFI update, and installed it. I ran Intel DSA, found four new drivers, and installed them. I ran the NVIDIA app, found new driver version 596.36, and installed it, too.

Now the P16G3 laptop seems to be purring right along. I’ll take it as a lesson learned (and probably re-learned) that it’s a good idea to update firmware and drivers before making major OS changes. Now that I actually stop to THINK about it, that makes pretty good sense. Even in Windows-World, it’s better to plan and aim before firing…

Facebooklinkedin
Facebooklinkedin

Timing WinGet’s Update Pipeline

OK then, I just read at WinAero that a new PowerToys v0.99.0 is out. Checking via WinGet upgrade in PowerShell it’s not yet in the pipeline. Nevertheless, the app itself is happy to grab said update from its GitHub repository, as you can see in the lead-in graphic. I’m now conducting an experiment. I’ll be checking hourly as I work at my desk, to see when that new PowerToys version comes into WinGet’s ken. Should be interesting…

What’s Involved in Timing WinGet’s Update Pipeline?

Behind the scenes, lots of things must happen before WinGet catches up, and offers the PowerToys update:
1. MS publishes the new release on GitHub (that’s done)
2. A Pull Request (PR) is sent to winget-pkgs with info about the new version, URLs, hash values, and so forth (usually automated)
3. Pull Request validation runs: automated checks verify installer hashes, check URL resolution, and validate manifest schema
4. Pull request merges into the WinGet source: a maintainer approves the package and merges it into the public database
5. WinGet CDN propagates: the updated database index appears via the winget source in related commands (show, install, uninstall, etc.)

How Long Does It Take?

Because PowerToys comes from Microsoft, its timeline is about as short as such things get. Turnaround normally takes no less than 12 hours, nor more than 48 depending on timing. If a weekend gets in the way the delay can stretch out. Ditto if issues with the manifest show themselves, or if the software being packaged shows a bug. Thus, for example, PowerToys v.0.99 has a Command Palette crash bug, and may be slowed to accommodate suitable hotfix.

We’ll see how this one goes. There’s already a new V0.99.1 version on GitHub (which includes that very hotfix). It’s in the WinGet pipeline now: let’s see how long it takes to get through, shall we?

Note Added 1:05 Later…It’s HERE!

The original post went up at 1:05PM local time. It’s now 2:10PM and a check on the P16 Gen 1 Mobile Workstation produces the following WinGet output: It’s here…

Notice that version 0.99.1 is on offer. That means the PowerToys team got its hotfixes into the package before sending it off to WinGet. Good job, @ClintRutkas and team. I am impressed.

And, now that I’m running it on the suitably-configured X380 Yoga, I see that the PowerToys upgrade also flashes an icon. Impressed again:

Facebooklinkedin
Facebooklinkedin

Firefox Update Fixes Weird Cursor Ripple

I’ve got to admit, I was misled this morning. After updating my NVIDIA Studio driver for the 3070Ti GPU, I noticed a strange “ripple” behavior around the on-screen cursor in Firefox. This occurred as I was scrolling inside today’s new posts and threads at ElevenForum.com. After reloading the graphics driver (WinKey+Shift+Ctrl+B), no change. So I asked Copilot: “Do I need to reboot?” “Nope,” it responded, “a Firefox update fixes weird cursor ripple” thanks to a fix for a DirectComposition code path error when using NVIDIA cards. It worked!

How Firefox Update Fixes Weird Cursor Ripple

A well-advised principle in troubleshooting relies on answering the question “What changed?” That’s what had me ready to blame the new NVIDIA driver as soon as Firefox got wonky. After taking advice from Copilot, I noticed further that the cursor ripple was indeed limited only to Firefox. It didn’t show up in Chrome or Edge, nor in other Windows apps. If it had been the GPU driver, all would have been affected.

Thus, I’m glad I thought to ask Copilot rather than start rebooting or rolling back the driver. Turns out the cause was obvious, indeed, but related to the specific program I was running as it interacted with the NVIDIA driver. Here in Windows-World, it’s wise not to overlook the obvious. But it’s also wise to cast a wider net, so as not to blame the obvious when something else could be — and in this particular case, was — at fault.

All’s well that ends well. I’m happily using my updated system. And Firefox — where I usually work to create this WordPress content — is working correctly now, too. Bonus: updating the browser is much faster than a driver rollback, and faster than a reboot. Good-oh!

Facebooklinkedin
Facebooklinkedin

Avoiding Excess Laptop Power Drain

Over the weekend, I transported one of my ThinkPad P16 Mobile workstations to a neighbor’s house. When I unpacked the unit from its knapsack, I noticed that it was pretty warm (and so was the interior of the bag). “What’s up with that?” I wondered. I’d closed the lid, but hadn’t shut the unit down. That was on me, as a little bit of investigating clued me into what’s needed when it comes to avoiding excess laptop power drain. Let me explain…

Power Options Key to Avoiding Excess Laptop Power Drain

Here’s the skinny: by default, my ThinkPad P16 Mobile Workstations go into S0 sleep but maintain network connection. Here’s how the powercfg /a command reports that state:

Basically, the network connection stays up even though the unit itself is mostly idle. Alas, when moving the unit by walking down the street, I’m entering and exiting WiFi domains at a pretty serious clip. Indeed, I’d bet big money that on Arbor Drive, not a single house has less than one active WiFi domain (I’ve got 4, if you count low- and high-bandwidth channels for SSIDs separately). That kept the NIC kinda busy and didn’t really idle the machine at all.

Hence the item shown as the lead-in graphic for this blog post. If you go into Power Options you can change the “Close Lid” setting to hibernate. That means the network connection goes quiet when the lid is closed, and the usual network traffic that might otherwise wake the machine does likewise.

Moving Around Makes Network Connection Iffy

If a laptop is on the move, making or using network connections is a chancy business. My neighbors may have SSIDs galore, but that doesn’t mean they’re sharing passwords with me. And it looks like most of them are smart enough to require WPA3-Personal these days, so guessing isn’t productive, either.

My point is that the network connection doesn’t really do anything except waste battery power when a laptop is on the move. For transport, that means the “Hibernate when lid is closed” setting shown above makes the most sense. That’s why it’s now my preferred move, before I shove the laptop into a knapsack or briefcase from now on.

All I need to do is remember to switch from hibernate to sleep when I get somwhere I plan to be (or work) for a while, and it’s all good. Here in Windows-World, responding to local conditions is a must, especially when it comes to conserving laptop battery life.

Facebooklinkedin
Facebooklinkedin

MS Kicks Off New Insider Channels

OK, then, today’s the day. Earlier, MS promised to change up its Insider Channel line-up. This morning, MS published an item to the Windows Insiders blog entitled “We’re moving to Experimental and Beta!…” Amusingly enough, this means that the former Canary and Dev channels are collapsing into a single Experimental channel, while Beta stays Beta (at least, for now). So, of course, I had to download and install the latest Beta onto X380 to see what things looked like, as MS kicks off new Insider Channels. Let’s check…

As MS Kicks Off New Insider Channels, I Wait…

As you can see in the lead-in graphic above, the latest Beta version shows up as a “Windows 11 Insider Preview Quality Update” (Build 26220.8283). It took about 5 minutes to download, but install seems poised to take somewhat longer. It’s been 3 minutes and it’s 12% done, so that means ~42 minutes total? Gosh, I hope that’s wrong… And indeed, it got to the “Restart now” button 27 minutes later. By the time I got through everything and back to the desktop total time elapsed was 35 minutes.

To my great disappointment, the view of Insider Channels from the latest Beta release is unchanged. It still shows Canary, Dev, Beta and Release Preview channels. Here’s what I used to (and still) see:

I guess I need to stand up a Canary PC or VM and see if it shows anything different. Here in Windows-World, I’m often reminded of the old French saw “The more things change, the more they stay the same.” I’m sure change is coming, but it hasn’t made it to Beta just yet.

 

 

 

Facebooklinkedin
Facebooklinkedin

Testing Catches Copilot Errors

I’m happy to admit that Copilot makes many of my Windows adventures easier to undertake. It certainly helps a great deal with PowerShell scripting stuff. But just this morning I was reminded that it’s always smart to question — and verify — everything Copilot tells me. In testing a USB4 NVMe enclosure just now, I asked it which of the ThinkPad X12 Detachable Gen 1 left-side USB-C ports is faster, remembering that one is 10Gbps (UASP) and the other 40Gbps (TB4). It replies “the lower port,” so I run CrystalDiskMark on both to check its answer. Once again, I confirm that testing catches Copilot errors. Let me explain.

How Testing Catches Copilot Errors

Simply put, checking what Copilot tells you will either confirm that it’s correct, or show that it’s wrong. As you can see in the lead-in graphic, which places the upper USB-C port left, and the lower one at right, large reads are MUCH faster on the left. That means the upper port MUST be the TB4 one because UASP/10Gbps can’t deliver more than 1,000 – 1,200 MB/s for either reading or writing.

When I showed Copilot the CDM results, it responded as follows:

I had it backwards — apologies for the bad call. The source I pulled from apparently described the ports in board-schematic order rather than physical top-to-bottom position as you see them on the tablet. Your bench test is the definitive answer.

And if you’re seeing typical UASP numbers (~1,000 MB/s sequential) from the lower port right now, that confirms it’s the USB 3.2 Gen 2 controller — exactly what you’d expect at 10 Gbps with protocol overhead.

At least, it doesn’t fight with me when I knock it over the head with contrary results. And when it writes scripts, the PowerShell mostly either works or it doesn’t. In the latter case, Copilot may fumble around for a while but it eventually gets things right.

Trust, But Verify

It’s not unreasonable to believe much, or perhaps even most, of what Copilot tells you. But if you don’t check its work, you’re setting yourself up for trouble. It’s good at finding stuff and presenting theories and explanations. It’s by no means infallible, as I have just shown. Take that into account, when you start leaning on it for help…

Facebooklinkedin
Facebooklinkedin

Author, Editor, Expert Witness