I've been using AI coding agents a lot lately - tools like Claude Code, GitHub Copilot CLI, and OpenCode. And the more I use them, the more I keep discovering new use-cases and ways they can be used.

Most people think of AI coding tools as just that - tools that write code. And yes, they do that, and they do it really well - but they also do much more than that.

Writing Code

First, let's start with writing code, before we dig into other use-cases. Watching the speed at which AI generates code might initially be seen as terrifying. But now, what I'm genuinely finding terrifying - is that I used to write all that myself by hand! The sheer amount of time I spent writing code that just did small, mundane things. Boilerplate, plumbing, wiring things up - all of it. Looking back, it's crazy how much time that took!

I rarely write code by hand any more. I describe what I want, and the agent generates the code. And the code it produces is genuinely good - often much better than what I'd have quickly thrown together myself. And I'm also no longer restricted to just languages I know - I recently created a Garmin app for my watch that used Monkey C!

And when I say "code", I mean all of it - application code, infrastructure-as-code, CI/CD pipelines, Dockerfiles, Kubernetes manifests, Terraform, the lot. From the agent's perspective, there's no distinction between "application code" and "infrastructure code" - it's all just text and tools. No more hand-writing YAML! 🎉

AI also doesn't have the mental resistance that a human has - for example, when faced with various time and business pressures, it's common for a human developer to skip "just a few" of those automated test cases. "Let me just write enough test coverage to get it through the PR review" (sound familiar?). AI doesn't have that constraint - it can generate a lot of test coverage without the mental fatigue / resistance, and also in a fraction of the time. Why wouldn't you take advantage of that?

The Agentic Loop

There's a concept that underpins a lot of what makes AI coding agents so effective, and it's worth calling out explicitly - the agentic loop.

When an AI agent is working on a task, it doesn't just have one go and hand you the result. It tries something, then verifies its own work. If the build fails, it reads the errors and fixes them. If the tests fail, it looks at the failures and iterates. It keeps going round this loop - try, verify, fix, repeat - until the task is done.

Using AI is no longer "type a prompt into a textbox, then manually copy code snippets out of the response" - then the typical back and forth, complaining about AI hallucinating. Hallucination is still a thing - but we just see less of it thanks to this hands-off self-correcting agentic loop.

Calling CLI Tools You Already Have

This one is huge, and I think it's massively underappreciated.

Your machine is already full of powerful command-line tools - kubectl, docker, az, dotnet, terraform, PowerShell, the Aspire CLI, and countless others. You probably use a handful of them regularly, but let's be honest - how many of us actually remember the syntax for anything beyond the basics? And even if you do - it takes time to type - so you're focusing on that, rather than what you're actually trying to achieve.

AI coding agents have access to those commands, and for most of them they'll already know all the syntax and tricks anyway. If it doesn't - it can quickly find out. Then it can just run these tools directly for you - they become an implementation detail, rather than something you have to worry about. You don't need to remember the syntax - you just tell the agent what you want to achieve, and it figures out which commands to run.

A few weeks ago, my Kubernetes cluster was reaching capacity. Rather than manually inspecting deployments and working out what was over-provisioned, I just told Claude Code to use kubectl to explore the cluster and see if anything could be optimised. I also pointed it at the code for the projects I was hosting. It inspected resource requests/limits across all my deployments, identified things that were over-provisioned, and made a bunch of suggestions to clean things up. The whole process would have taken me ages - it took the agent minutes. It also meant I didn't need to increase my cluster's capacity - so immediately saved me money.

Similarly, the Aspire team have recently ramped up what the Aspire CLI can do - it can now pull logs, traces, and metrics. This means your AI coding agent can just call the Aspire CLI to get all of that observability data. Combine that with the fact that the agent also has access to your code, and it can connect the dots between what's happening at runtime and what's in the codebase. That's incredibly powerful.

Debugging and Investigating Issues

Because the agent has access to logs, traces, and metrics - whether that's via the Aspire CLI, a Grafana MCP, or just reading log files - and it also has access to the code itself, it can investigate issues far faster than I can. It can trace a problem from the symptoms in the logs right through to the offending line of code, all in one flow.

But it's not just code bugs. A few weeks ago, my machine was running slowly. I couldn't figure out why. So I told Claude Code to investigate it. The agent used a whole bunch of command-line tools to look at event logs and system diagnostics. It found that my network driver was outdated and was raising a huge number of error events, which was dragging the whole system down. I would never have found that myself - at least not without a lot of frustration and wasted time.

Same goes for CI pipeline failures - I quite often just give the build id to the agent and tell it to investigate and fix the issue. It'll grab the build logs from either the gh CLI if in GitHub Actions, or the Azure DevOps MCP if in Azure DevOps (those are the two I use) - and do the rest itself!

Throwaway Scripts

There are tasks that come up from time to time where you could write a script to automate it, but the effort of writing, testing, and debugging that script just isn't worth it for a one-off job. So you end up doing it manually, which is tedious and error-prone.

AI agents completely change that. They can knock out a throwaway script in seconds, use it, and then you just delete it.

Here's a real example: I wanted to export my newsletters from Beehiiv and convert them into Markdown files so they could be source-controlled. The Beehiiv export gives you CSVs full of HTML - completely unreadable. I told Claude Code to take the CSV and convert all the newsletter issues into individual Markdown files. I also told it to use the Playwright CLI to verify that the converted output matched the original.

The agent then created a Python script full of regular expressions to bulk-convert the HTML. It ran Playwright to verify the output, found mistakes, went back and modified the Python script, and kept iterating until everything matched up perfectly - all without my involvement. Once it was done, we just deleted the script. Job done.

In the past, writing that script by hand - with all the edge cases and verification - just wouldn't have been an option. But with an AI agent, I'm now able to do things that I previously couldn't have justified spending the time on.

Custom Skills and Commands

Most AI coding agents support some form of custom skills or commands - reusable prompts that you can trigger with a shortcut. In Claude Code, for example, you can create skills and invoke them with a slash command.

This has been a game-changer for me. Here are a few I use regularly:

• Podcast show notes - I've got a skill that creates new show notes for my podcast, following my exact format and structure.
• Newsletter editions - A skill that creates a new newsletter edition and researches news items for me. (note that I still manually cherry pick what news items I want, and more often than not, hand-write the descriptions - my newsletter is certainly not AI slop!)
• Git commits - When I run /commit, it asks me if I want to include "Closes #123" for a GitHub issue, and whether I want to push afterwards. It's a small thing, but it's exactly how I want my workflow to work.
• Jira tickets - I've got a skill that creates Jira tickets and fills out all the custom fields that my client requires, asking me questions along the way. No more fighting with Jira's UI, and remembering that client's various custom fields I need to set.
• Plus many others.

Creating these skills is surprisingly easy - it's just a directory with a markdown file called SKILL.md, plus any other additional files you want to include (images, PDFs, etc). Claude Code even has a skill creator skill that helps you build, improve, and verify them. If you don't use Claude Code, you can still use that skill - as skills are just portable markdown, following a standard format.

Tickets, PRs, and Collaboration Tools

Thanks to tools like the GitHub CLI (gh), the Azure DevOps MCP server, and similar - your agent can read and write across all the collaboration tools you use day-to-day. Tickets (JIRA, Azure DevOps, GitHub Issues, etc), PR comments, the lot.

I often get the agent to query a ticket to pull out the description and acceptance criteria before starting work. It's an amazing context primer, and gives the agent a good checklist to work against. As mentioned above, I also use a skill that writes JIRA tickets for me.

On the PR side, the agent can read comments that reviewers have left and either fix them directly, or suggest reasons to push back. Technically, you can also get it to reply in the comments for you - but I probably wouldn't recommend that if the reviewer is expecting a human response. Unless of course it's an accepted practice within the team.

As an aside, GitHub does have native cloud-based AI code reviews too - but those are out of scope/context of this post.

Standard disclaimer!

Before you go and let your AI agent loose pointing at all your production environments - obviously ensure you take the necessary precautions. If you're giving it access to a prod Kubernetes cluster for example - ensure its authentication has the necessary authorisation restrictions - eg. readonly.

It's easy to anthropomorphise these AI agents - and yet, people are happy to only anthropomorphise certain traits, and still expect AI to not make mistakes. AI does make mistakes, just like humans do. The problem is that if AI makes mistakes - it makes them MUCH FASTER. So do be careful.

Personal projects

On a personal note, I'm currently using Claude Code to build lots of stuff that I wouldn't have previously been able to justify...

• An entire fitness product that I'm planning to productise. Before AI, this project would have needed a team of developers doing years of work. Now I'm able to build it in my spare time without having to quit my job, or it impacting my family-time or fitness training. That's how much of a difference this makes.
• Plus so many small utility apps - from a desktop pomodoro timer, to a time blocking app. Things I never would have been able to justify the time creating previously.

Also, lots of non-code stuff - a few examples...

• I coach kids' athletics and have recently been managing a series of competitions that our club was competing in. This is a real pain to do, as there are various activities with limited slots, and I get various requests from parents. There's also a bunch of forms to fill in, etc. Claude Code has made this WAY easier - especially working out the allocations, which can get complicated.
• I wanted a personal-finance long-term forecast solution - which takes into account income, pensions (including future draw-down over time), inflation, future big spends, etc, etc. I just threw all my info into Claude Code, and it generated a system for me, which shows charts and projections, doing all the tax calculations, etc. It's amazing!

Plus many more things that I won't bore you with - but the point is, the use-cases for AI coding agents go way beyond just writing code.

Wrapping Up

I've covered a lot of use cases here, but honestly, this barely scratches the surface. Every week I seem to discover a new use-case that I can use an AI agent for. The combination of being able to write code, run any CLI tool on your machine (or use MCP Servers), generate throwaway scripts, and iterate on their own work through the agentic loop - it all adds up to something that totally changes the way we work.

I should probably briefly touch on cost. Personally, I pay for the Claude Max 5x plan. It's not cheap if you compare it to a subscription like Netflix or Spotify. But if you compare it to having a team of developers at your disposal, it's an absolute bargain. The amount of time it saves me, and the amount of stuff it enables me to build - both personally and professionally - is worth way more than the cost.

If you're using AI coding tools but only for writing code, I'd really encourage you to explore what else they can do. I'd love to hear about the different use-cases you're using AI for! Please let me and the other readers know in the comments below! 👇

I do believe that Redgate has been very customer focused since it’s inception. I’ve worked with them in some capacity since 2002 and I’ve felt this along the way:

The next page has this statement:

We believe that if we do what is right for our customers then we will thrive.

I think that’s been true when we keep this in mind. The (relatively) few times we’ve started to do things for ourselves rather than thinking about customers, things haven’t worked out as well.

I think this sentiment is one that guides a lot of my life. Certainly inside Redgate, but also in the rest of my life. If I do what is best for another, or the world, often that works out well. It doesn’t mean I’m as efficient, profitable, less stressed, or anything else as I could be.

But I’m happier and I thrive.

I have a copy of the Book of Redgate from 2010. This was a book we produced internally about the company after 10 years in existence. At that time, I’d been there for about 3 years, and it was interesting to learn a some things about the company. This series of posts looks back at the Book of Redgate 15 years later.

The post The Book of Redgate: Do the Right Things appeared first on SQLServerCentral.

For anyone recycling buckets of batteries, Linogy's rechargeable Li-ions are long-lasting and self-organizing.

The post Review: Linogy Li-ion Rechargeable Batteries appeared first on Make: DIY Projects and Ideas for Makers.

Hey Windows Insiders, Today, we’re excited to share some improvements to the Windows Update experience that are now starting to roll out. These improvements are the direct result of your feedback. We are continually reading the feedback submitted about the Windows update experience. Personally, I’ve had the opportunity to read over 7,621 direct verbatims over the last few months. Across this feedback there are two key themes that persistently pop out: disruption caused by untimely updates and not enough controls over when updates happen. Across this feedback there are two key themes that persistently pop out: disruption caused by untimely updates and not enough control over when updates happen. The changes we’re rolling out today are focused on giving Windows users more control over their PC experience, while keeping devices secure by design and by default. Let’s get into what’s new!

More control

Updates are important part of keeping your PC secure and running smoothly. But at the wrong time, they can also critically break your flow. To make this a better experience, we are focused on giving you more control of updates in four key ways:

1. Skip updates immediately during the out of box experience (OOBE)
2. Extend update pauses as many times as you need
3. Always-available options to shut down and restart without updating
4. More insights on available updates so you can make more informed installation decisions

1. Skip updates immediately during OOBE

Earlier this year, we added the control to immediately skip updates during device setup—giving you the option of landing on the desktop faster and getting updates later or getting updates right away and landing on a PC that has all the latest features and fixes. If you choose to skip updates, the latest features and security updates won’t be available until you take the update(s). With immediate choice built into setup, you decide when updates happen. [caption id="attachment_178903" align="aligncenter" width="1024"] OOBE screen showing ability to update later[/caption] *Note – this is not applicable to commercial devices where the out of box experience is being managed or in select cases where updates are required for the device to be functional.

2. New controls for pausing updates

The Pause updates experience for Windows Update now puts you more in control. First, with a new calendar experience, you can choose a specific day of the month you want to pause until, up to 35 days, enabling you to plan around expected travel, conferences, exams, or even just busy weeks. When 35 days just isn’t long enough, we are also enabling you to extend the pause end date as many times as you need.  This means you can now re-pause for up to 35 days at a time, with no limits on how many times you can reset the pause end date. [caption id="attachment_178902" align="aligncenter" width="1024"] Showing ability in Windows Update to extend update pause[/caption]

3. Shutdown, Restart on your terms

Restarting or shutting down your PC should always be simple, predictable, and on your terms – even with updates waiting to be installed. We’re improving this experience by clearly separating power actions from update actions. With this change, the Power menu will always show the standard Restart and Shut down options, meaning you will always have a choice to just restart or shut down your device without having to install the pending update.  At the same time, update‑specific choices like Update and restart and Update and shut down will still be available when applicable. This gives you four clear options—and full control over what your device does next. [caption id="attachment_178904" align="aligncenter" width="1024"] Showing ability to restart and shut down without installing pending updates[/caption] If you choose Restart or Shut down, Windows will perform exactly that action, without unexpectedly starting an update. If you’re ready to install updates, you can explicitly choose one of the update options. After a restart, Windows will attempt to restore previously opened applications faster, allowing users to return more easily to what they were working on. This change is about is about making the Power menu more predictable, so when you need a quick restart or want to power off before heading out, Windows does exactly what you expect.

4. More insights on available updates

One of the key pieces of feedback after moving to simplified update titles at the end of 2025 was an ask to better understand driver updates. Often, driver updates would have similar, if not identical, titles. To help provide you with more insights, we have added the device class to the driver title - ensuring pending or installed driver updates clarify whether they apply to display, audio, battery, extension, HDC, or other applicable driver update classes.

Fewer disruptions

There are few things more frustrating than sitting down to use your computer, only to find that it requires an update. Worse, is when this happens multiple times in a given month. We know this has been a major pain point for Windows users, so as of today, we’re unifying the update experience to reduce the number of reboots you see every month. We are starting by coordinating driver, .NET, and firmware updates to align with the monthly quality update, reducing update experience to a single monthly restart. Windows quality updates include monthly security updates, emergency out-of-band (OOB) updates, and optional non-security updates if initiated by the user. For Windows Insiders in Experimental and Beta you will note weekly updates, Persistent Seekers in retail will see bi-monthly updates, and retail users who have not opted to get any updates early will see monthly reboots. For users checking out the Settings > Windows Update page, you will see all of these updates collapsed into a single Available updates section. [caption id="attachment_178905" align="aligncenter" width="1024"] Showing all updates available under Available updates[/caption] Updates will download in the background, then will wait for a coordinated installation and restart. This installation and restart will align with the next Windows quality update or other update that you manually approve. Users can always acquire all or specific updates earlier if desired by initiating download, install, restart (if applicable) for available updates. If none of these actions are taken, updates will be downloaded in the background and applied alongside the next scheduled Windows quality update.

Keeping you secure, by default

In line with Microsoft's Secure Future Initiative, Windows is grounded in keeping devices secure by design and secure by default, getting Windows devices onto the latest security update shortly after its released. However, we know sometimes users hit issues when attempting to take these security updates due to bandwidth constraints and update failures. Over the last few months, we have made steady progress in reducing the download and overall time it takes to apply a Windows update and will continue to work toward reducing overall update time over the course of this year. These improvements are particularly impactful for devices that spend less time online or in areas with poor connectivity, leading to higher rates of update success. Further, we are ensuring devices stay secure by default through automatic recovery for update failures – taking additional steps in the background to help the update complete successfully without user intervention. This means your device will automatically attempt to recover from installation failures in real time – causing some updates to take longer to complete, but ensuring they have a higher success rate.

Stay protected, with more flexibility

Updates are a critical part of helping keep your device secure and protected, and with these changes you now have more flexibility to take these updates on your terms. As always, we recommend taking these updates shortly after they are released to keep your device and your data secure. We are excited for these changes to reduce disruption and provide you with more control to all now be available, with many beginning to roll out to devices in the Dev Channel and the new Experimental channel today, which also starts rolling out today. We will have more to share around how these features will light up for commercial customers and the controls that will be available for admins around them soon. None of this would have been possible without your feedback, so please keep it coming! Thank you! Aria Twitter (X): @AriaUpdated

Hello Windows Insiders, Today is the day we’re beginning to move to the new Experimental and Beta channels as announced earlier this month.  As a reminder, most Insiders, those who are in the Dev and Beta channels today will transition to the new channels over time as follows:

• Beta Channel > Beta 
• Dev Channel > Experimental

For those in the Canary Channel, you will also be transitioned to specific versions of the Experimental channel, depending on whether you are in the 28000 series Canary channel or took the optional 29500 series update. The transition takes place as follows:

• Canary Channel 28000 series > Experimental (26H1)
• Canary Channel 29500 series > Experimental (Future Platforms)

Transition timeline

To ensure quality and reliability of the program changes, we will be rolling out the changes in a phased approach. Starting today, we will begin moving users in the Dev Channel to Experimental. Delivering on our promise to allow Insiders in Experimental to enable new experiences before they roll out to your device, if you are in the Dev Channel and do not see the new Experimental channel UI, you can enable it yourself by going to Settings > Windows Update > Windows Insider Program > Feature flags and toggling on the new experience. We will continue this rollout over the next few weeks, expanding to Canary Channel 28000 series to Experimental (26H1), Canary Channel 29500 series to Experimental (Future Platforms) and Beta Channel to the new Beta experience. We will announce when we begin starting the rollout for these channels.

Changes to the Beta experience

Based on your feedback to have clearer definitions between channels, we have committed to updating how we use the new Beta experience to reflect what is coming to retail in the following weeks. With this, there will be a realignment of experience when Beta Channel Insiders move to the new Beta experience. While Beta Channel Insiders can generally expect a similar experience after this transition, some may notice feature changes. Although these feature differences will be minor, users who are looking for the best continuity of all existing features should consider moving from the existing Beta Channel to Dev Channel, in preparation for the move to Experimental. While users can still move from Beta to Experimental after the changes are implemented on your device, moving beforehand ensures the best continuity of experience for those features.

Updating WIP build information communications

We will also be changing how we publish the update information for each build, moving to the Windows Insider Program Documentation Hub. Moving release notes to this site provides additional benefits including:

• Easier navigation between different channel and build release notes
• Dark and light mode support
• Better localization support
• Enhanced deep linking capabilities

We will still announce when we’re releasing new builds here on the Windows Insider blog and will link to each build’s release notes in the Windows Insider Program Documentation Hub. We will also highlight some of the more notable new features coming into WIP for the first time in the blog, but Insiders should check the release notes for more detailed build information.

New builds this week

With that, today we are releasing new Windows 11 Insider Preview Builds. Starting today, all Insiders can find the release notes for your device based on the new channel system, even if you haven’t moved yet. This is to make finding build information as easy as possible during this transition. See your channel release notes here:

• Beta (Including Beta Channel): Build 26220.8283
• Experimental (Including Dev Channel): Build 26300.8289

For those on other specific build versions, here are today’s new builds and release notes:

• Experimental (26H1) – Including Canary 28000 series: Build 2820.1873
• Experimental (Future Platforms) – Including Canary 29500 series: Build 29576.1000

As a reminder, you can always find your build number in the watermark on bottom right-hand corner of your desktop. Thanks, Stephen and the Windows Insider Team