Building your own internal developer platform is a lot of work, buy it instead

Here’s a transcript:

One of the reasons internal developer platforms are so valuable is because they do so much, and that’s something that you run into if you’re trying to build your own platform instead of buy it.

The goal of a platform

The goal of a platform is to remove as much toil wait time, just nonsense work that a developer has to do just to get their applications into production, just to actually run them and configure them.

You want your developers building their apps, not building production or configuring their apps. That’s how you make the business better by changing the applications around, and that’s exactly what a platform is targeting

With your ideal platform, a developer can write their application and then just deploy it without even having to worry about building containers or how it’s packaged, how it’s deployed.

Building your own platform

Now a lot of people think they can get away with building their own platform.

Usually people who set themselves up for this journey don’t realize how much a platform does.

If you look at the Cloud Native Computing Foundation’s

platform reference architecture, you can get a notion of everything that a platform does. Also imagine each of these little components have to be integrated together and also work with the existing systems that you have in your organization.

Now you look at all this stuff, you’ve got middleware, security management, you’ve even got things like the build tool chain, what a lot of people will call the “golden path” or a pipeline in there.

You now own the entire platform

So let’s say you’ve built all of this, this platform.

That’s just the beginning. You now have to update it. You’ve gotta monitor what works and doesn’t work. You’ve gotta a product manage it. Not only update bugs that you have, definitely apply security patches.

Think about that across all of the different boxes that you have. Also make it compliant, pass whatever audits that you have.

But you need to add actual new functionality that your developers need.

Right now we’re in the middle of a great example of that, adding AI functionality.

What does it even mean to add AI to a platform?

(We’ve got some good ideas at Tanzu.)

But you owning the platform, having built it, you need to add that in, understand it, add that functionality in, and totally own that process.

Chances are high. You haven’t thought about how much work it is to build your own platform. Like a lot of teams I’ve talked with, you probably have a smaller scope, like thinking that it’s just putting Kubernetes out there, maybe with some pipeline integration and doing a baseline container image that your developers use.

But it’s a lot of work as you can see, and when you think about it, it’s nothing that’s gonna differentiate your business, make your business run better.

Year Zero

In fact, it’s gonna be a huge distraction and something that’s probably gonna take you at least a year to get out there. It’s gonna hold you back from improving the way your organization runs.

Just like any kind of undifferentiated thing, it’s basic strategy to not do it yourself and instead offload it to someone else.

More in the paper

If you’re considering building your own platform, obviously I don’t recommend it and I think it’s a bad idea.

Now, if you wanna see some other reasons why I think that, you can check out a free paper that we have going over six other reasons and ways of thinking about the build versus buy choice, I think the answer’s pretty clear. You can get that paper for free if you go to cote.io/diy …and, really, buy the platform.

Relative to your interests

Always lots of AI stuff, this week more tempering expectations, and the usual how to stuff.

• Breaking the AI illusion: From adoption to growth - Garbage in, garbage out: “37% of midmarket CMOs believe AI-enabled marketing technologies have potential to help their organizations over the next 12–18 months, according to the Midmarket survey. This includes incorporating tools and workflows to boost content creation and automate campaigns, critical elements of modern marketing success. Still, only 31% of the same CMOs are prioritizing the modernization of their MarTech stacks. This is a crucial metric. Without updated systems and integrated data pipelines, even the most sophisticated AI tools remain disconnected from broader workflows, limiting their value.”

• All IT work to involve AI by 2030, says Gartner - ‘AI’s hidden costs mean Gartner believes 65 percent of CIOs aren’t breaking even on AI investments." And: "Plummer said Gartner doesn’t foresee an “AI jobs bloodbath” in IT or other industries for at least five years, adding that just one percent of job losses today are attributable to AI.’ // Still figuring out the ROI.

• The Three Faces Of Generative AI - “People today use large language models for three central purposes: 1) Getting things done 2) Developing thoughts, and 3) Love and companionship.”

• What if the AI stockmarket blows up? - ‘By our reckoning, the total revenue from the tech accruing to the West’s leading AI firms is currently $50bn a year. Although such revenues are growing fast, they are still less than 2% of the $2.9trn investment in new data centres globally that Morgan Stanley, another bank, forecasts between 2025 and 2028—a figure which excludes energy costs. Meanwhile, the extent to which revenues will translate into profits is murky. A recent study by researchers at the Massachusetts Institute of Technology concludes that 95% of organisations are getting “zero return” from investments in generative AI.’

• How I Use Claude - “Claude is really good at helping here, mostly because thinking quickly saturates: when you’ve thought about a problem for five minutes, you’ve had all the thoughts you’re gonna have, and it’s time to talk to someone else. Claude lets me sample fresh perspectives and possible actions I had not thought of.”

• An ode to “via” - ‘the old ways’ - there are so many of these little things from the 2000’s web/bloggosphere.

• Oxide Friday FAQs - These are clever/amazing “brand marketing” assets. You take advantage of a unique and famous asset you have (Bryan) and get some thought leadership and brand definition out there. Also, they have the product right there the whole time.

• Blade Runner makes its live-action return next year - 2026 in Amazon.

Being a VC - Software Defined Interviews

In last week’s Software Defined Interviews, Whitney and I talked with Rachel Chalmers. I’ve know Rachel a long time. I’d followed her as an analyst and talked her here and there at conferences. Then in a chance meeting in a hotel lobby in San Francisco, she accidentally recruited me to take over her job at 451 Research when she was going off into VC-land. In the interview we talked about lot of VC stuff, and, of course, a few other topics like horses.

Conferences

VMUG London, speaking, September 18th, speaking. SREDay London, speaking, September 18th and 19th. Civo Navigate London, September 30th, London, speaking. Cloud Foundry Day EU, Frankfurt, October 7th, 2025, speaking. AI for the Rest of Us, London, October 15th to 16th, London, speaking. SREDay Amsterdam, November 7th, speaking.

I’ve got a 20% off discount for AI for the Rest of Us: SDI20. You should go the conference if you can, it’ll be good!

Logoff

Next week I’m back to business travel, in London for two of the conferences above. I’m scheduled to take the train for that trip and the next two. This means it’ll be more work to get my status with KLM. Is that bad? Is it good? Traveling by train is so much nicer and, I presume, earth friendly. It’s not always cheaper, which is odd, but I guess the way of the pricing. But, it so much more pleasant.

How I Use Claude - “Claude is really good at helping here, mostly because thinking quickly saturates: when you’ve thought about a problem for five minutes, you’ve had all the thoughts you’re gonna have, and it’s time to talk to someone else. Claude lets me sample fresh perspectives and possible actions I had not thought of."

An ode to “via” - ‘the old ways’ - there are so many of these little things from the 2000’s web/bloggosphere.

Oxide Friday FAQs - These are clever/amazing “brand marketing” assets. You take advantage of a unique and famous asset you have (Bryan) and get some thought leadership and brand definition out there. Also, they have the product right there the whole time.

Breaking the AI illusion: From adoption to growth - Garbage in, garbage out: “37% of midmarket CMOs believe AI-enabled marketing technologies have potential to help their organizations over the next 12-18 months, according to the Midmarket survey. This includes incorporating tools and workflows to boost content creation and automate campaigns, critical elements of modern marketing success. Still, only 31% of the same CMOs are prioritizing the modernization of their MarTech stacks. This is a crucial metric. Without updated systems and integrated data pipelines, even the most sophisticated AI tools remain disconnected from broader workflows, limiting their value."

All IT work to involve AI by 2030, says Gartner - ‘AI’s hidden costs mean Gartner believes 65 percent of CIOs aren’t breaking even on AI investments." And: “Plummer said Gartner doesn’t foresee an “AI jobs bloodbath” in IT or other industries for at least five years, adding that just one percent of job losses today are attributable to AI.’ // Still figuring out the ROI.

The Three Faces Of Generative AI - “People today use large language models for three central purposes: 1) Getting things done 2) Developing thoughts, and 3) Love and companionship."

Training for Legacy - Press Pass

Elisabeth Greenbaum Kasson asked me recently for advice on working with legacy applications. Check out her piece on it. Here’s the full reply I sent to her in email:

When it comes to “dealing with legacy," there aren’t that many good options. We often think of “legacy” as software that must be changed but that we’re afraid to change. If you’re not afraid to change it, you often just think of it as “our software.” Legacy has this connotation of it being risky, scary, or maybe just boring.

If someone wants to go down into the mines of legacy management, the first thing I’d recommend is doing some history work to find out why the legacy system in question was created, what it’s currently used for, and, hopefully, who the current stake-holders/owners are. You’d be surprised - or maybe not! - how often some or all three of those are totally unknown: with unknown stake-holders, I sometimes hear tale stories of IT departments just shutting down systems and waiting to see who calls them.

Understanding the why, what, and who of a legacy system will tell you most of what you need to know when it comes to managing it. Further up the management chain, having a good grasp of and on portfolio management is valuable. Given the why, what, and who, you should be able to prioritize any given “legacy application” relative to another with respect to funding and attention. Is fixing the application that’s used to schedule office party birthday cake orders more important than the application used to re-order plungers for the warehouse bathrooms? You won’t know between cakes or plungers if you don’t do portfolio management.

The other aspect is simply learning the technologies you need - operationally, programming, and sometimes physical management - to keep the thing up and running and to modify it. This might mean learning, for example, about operating systems for mainframes, AS/400, UNIX, older versions of Windows, and sometimes even exotic things like OS/2. There’s dozens of programming languages out there, and you’ll need to learn not only the appropriate “old” language, but how the build, version control, and project management tools around those old stacks function.

For more, I wrote about dealing with legacy in my cloud native journey booklet last year.

Publishing Versus Orchestrating

You can focus on short term delivery, or long term delivery. Do you follow the Twitter drama day-to-day, or just wait every week or two to see what happened? Do you write an article in a day or two, or write a 40 page PDF in a month, 120+ page book in a year.

My brain is wired to find the pay-off in every moment, and to be present and remembering. Long term projects seem like I’m not doing work - all that time just…working, not publishing. It was blogging that did this to me, not Twitter.

(Here, when I say “publish,” I mean deliver something.)

Managers are rarely publishing/delivery frequently. Instead the orchestrate other people dong it: build the systems that everyone works in (the organization and “culture”), make sure there’s capable people in the system (hiring, retaining, career management), give editorial direction (strategy, goals), and check on the status of things (verify that things are happening, fix problems).

When I think about that work it seems like it’d be so unsatisfying: you’re never directly publishing anything! I know people who manage take great satisfaction in…making things happen, helping people, all that. And, of course, you often make more money, which is always nice.

Individual contributors (as they call people like me) often run around saying “we should do this and that…” or the unkind among us “why aren’t we doing such and such - what do all these people do?” That, perhaps, is a portal into orchestrating other people delivering: when you have that thought, rather than take that work on your self (which leads to you doing less overall because you now have too much work to do), why not orchestrate someone else doing it?

Summer. For most of my life I lived in Texas, where the heat of summer melts your face off. Summer was fun because I wasn’t in school, not because it was sunny. Now that I live in a part of the world where summer is mild, I really like summer. I see what all the fuss was about! So, too bad it’s mostly over now.

On this week’s Software Defined Talk episode we discuss the effectiveness of reorgs, Meta’s new AI team, and the Google antitrust ruling. Plus, some strong thoughts on cold brew and bathtubs. Listen in, and you can also watch the unedited video recording if you’re into that kind of thing.

Wastebook

• “Buy once, cry once.”

• “My sister loves all orphans.” RotL, #592.

• “If you’re good at sleep, you like sleep.” Ibid.

• The reviews of this airport are amazing. As one “just fine” review puts it: “You have to have a good sense of humor and low expectations at this airport.” I’ve been a frequent business traveler for 20 years, and I like, and try to live, that idea: have a good sense of humor. If you have that when you’re traveling, you’ll have a good time. If not, you’re just asking to be upset. The people who left bad reviews of that hilarious airport are obviously not a golfer.

• “they are funny, shrewd and clear-eyed about aging, taking in the good (lunchtime drinking, caring less about others’ opinions) and the bad (everything else).” The Economist World in Brief, August 28th, 2025.

• “A demon buried in a glowing glass container.” Warren Ellis.

• An old Italian man at the beach standing, talking with his older friends - men and women - just lets out a huge, sonorous fart. Chao!

• “At least nobody gave him any solid gold statues this time, as far as I know.” With venom dripping. And:

• “Bill Gates was also there for some reason.”

Relative to your interests

• The Tanzu portfolio no longer includes Kubernetes. Hear an overview from the Tanzu GM, Purnima Padmanabhan. // Over the past, I don’t know, 4 or 5 years, the company and now business unit that I work in was the home of Kubernetes in VMware. This made “Tanzu” synonymous with Kubernetes. Now, the VMware Kubernetes products have been moved to the VMware group, out of Tanzu. And, that bundle of Kubernetes stuff is now called VMware Kubernetes Service, VKS. This brings the Tanzu business unit’s focus back to just platform for developers: the Cloud Foundry based PaaS, databases and data services, and now AI middleware. Also, check out Forrester’s portfolio overview.

• New VMware private AI infrastructure rethinks Tanzu, again - “If you want to use Kubernetes as Kubernetes, then VCF, which includes VKS, is what you use.”

• Private AI powers Broadcom’s vision for VCF 9.0 - “We shared we have more than 80 customers now,” for AI stuff, I believe he’s saying. // And: “The key thing with VCF that I think kind of gets missed sometimes in the conversations is everybody is claiming they can do sovereign cloud, but the details matter here,” Wolf said. “The difference with VCF is we run a fully air-gapped environment. The organization owns the control plane. That is absolute control that you have. No matter what happens in the world, you have ownership of the software stack and your intellectual property.”

• VMware Explore US 2025 Breakout Session URLs - William Lam makes a simple, nicely usable list of all Explore sessions and links to slides. Every conference should have a page like this.

• The Great Migration: Why Workloads Are Coming Home to Private Cloud - “55% of enterprises are already running GenAI in a private cloud – especially when it comes to use cases like inference, fine-tuning and RAG.”

• With AI Boom, Dell’s Datacenter Biz Is Finally Bigger Than Its PC Biz - “Thanks to the GenAI boom, Dell’s Infrastructure Solutions Group, which sells servers, storage, switching, and services into the datacenter, is finally – and very likely permanently – larger than its PC business for the first time in its history. (We are not counting the time a decade and a half ago when Dell ate Perot Systems and was also eating software companies to try to create a clone of IBM, much as HPE did at the same time.)” // Weird parenthetical?

• VCF And Private AI Take Center Stage At VMware Explore 2025 - “We believe that while the company has quietly designed and delivered the leading private cloud platform, it should be more vocal and direct in the market. While competitors claim to be the more affordable option, these claims are often based on upfront discounts that don’t fully account for cost over the long term. And while these competitors talk about vendor lock-in, they are in reality simply suggesting that customers move from one proprietary cloud stack to another.”

• Is Your IT Organization A Ponzi Scheme? - “The only way out is to stop borrowing against the future and start paying down the past. Escaping requires sustained platform investment — enough to reach equilibrium where debt stops growing. This means: Refactoring to improve code structure and reduce the cost of future changes. Refreshing technologies before they become emergencies. Rationalizing redundant systems to reduce complexity and risk.”

• Thirty Years On, the Californian Ideology is Alive and Well - The cycle of tech loosing its democratic morals in favor of tech innovation, and, of course, making money.

• China Has a Different Vision for AI. It Might Be Smarter. - AI applications outside of the consumer sector: “In February, the city announced the release of an agricultural AI model, using technology from the Chinese startup DeepSeek, which gives local farmers guidance on crop selection, planting and pest control, according to a local government report. The city’s meteorological service is using DeepSeek to improve the accuracy of weather reports. DeepSeek is also helping local police analyze case reports and decide how to respond to emergencies.”

• Five Vide Coding Lessons for the Enterprise - If you start with a garbage system, you end up with garbage apps. // ‘In an enterprise, the “context window” isn’t just a technical term; it’s the accumulated technical debt, undocumented tribal knowledge, and complex dependencies that hold your systems together. A new developer or a new AI tool can’t simply be dropped in and expected to understand this history. The leader’s job is to provide that context, not expect the tool to figure it out.’

• The End Of Business Apps As We Know Them Is Here - Kate Leggett (Forrester) has a go at defining the enterprise AI stack, bringing in the “fabric” notion.

• Seeing like a software company - When management meddling slows down the business because they need/want to measure and make decide. Also, good sub-plot on “the meeting for the meeting,” prep meetings and using the “back-channel” to get things done.

Conferences

VMUG London, speaking, September 18th, speaking. SREDay London, speaking, September 18th and 19th. Civo Navigate London, September 30th, London, speaking. Cloud Foundry Day EU, Frankfurt, October 7th, 2025, speaking. AI for the Rest of Us, London, October 15th to 16th, London, speaking. SREDay Amsterdam, November 7th, speaking.

I’ve got a 20% off discount for AI for the Rest of Us: SDI20. You should go the conference if you can, it’ll be good!

Logoff

I was on Cloud Foundry Weekly yesterday. I’ve been researching the role of platform engineers with respect to AI. There’s all sorts of things! Nick of the show put together a presentation of what he’s seen talking with our customers and also running AI services on his own. It was great, check it out if you’re interested in this topic.