Recording of my AI platform engineering talk

This is a new talk of mine going over how platform engineers can support AI. Well, it’s more about how we don’t exactly know, but we can speculate based on a handful of early use cases. Here’s the slides if you’re into that kind of thing.

Here’s the 🤖 on my key points:

1. Platform engineering for AI is mostly running another middleware service - same infrastructure tasks as always, plus model registries and figuring out who handles AI safety evaluations.

2. Put a gateway/proxy in the middle immediately - the eternal computing lesson we relearn every five years: never let developers talk directly to services or you’re stuck sending angry emails.

3. Three new types of customers for platform teams - developers who want AI coding help, “normals” using internal ChatGPT for business tasks, and the mythical “AI embedded in applications” unicorn nobody can actually find.

4. Platform-as-a-product thinking applies here too - make something developers actually want to use, get lawyers and security on your approval board to avoid the old “center of killing dreams” problem.

5. Experiment rapidly when you don’t know what you’re doing - which is right now for everyone, so build fast feedback cycles rather than one perfect thing that takes forever and turns out wrong.

Check it out, and tell me if you have any things you’ve learned, heard, done’d, etc. on the topic.

Are you enjoying the widgets?

This week’s Software Defined Talk:

This week, we cover Oracle’s OpenAI deal, the RubyGems drama, and Atlassian buying DX. Plus, does anyone still use widgets?

Like and subscribe, hey guys!

Also, don’t forget to check out the sister-show, Software Defined Interviews with Whitney Lee and me. For example, in the last episode we talked with Hannah Foxwell about, I don’t know, everyday AI stuff. New interviews come out every two weeks, usually on Wednesdays at 7:30am Amsterdam time. Subscribe to auto-pilot your mid-week enjoyment.

Relative to your interests

Lots of “measure the ROI of AI is very difficult” vibes below.

• Microsoft Hopes Hastened AI Rollout, Price Discounts Can Fuel Office 365 Growth - Enterprise AI not legible. If you can’t measure it, you can’t ROI it: “It’s easy for an employee to say, ‘Yes, this will help me,’ but hard to quantify how. And if they can’t quantify how it’ll help them … it’s not going to be a long discussion” over whether the software is worth paying for, Thompson said. // And: it’s a “challenge for businesses is that there’s this leap of faith moment where you try to justify it with the return-on-investment calculation, which is hard to figure out.”

• Marc Benioff Said AI Was Easy. A ‘Crazy’ Team at Salesforce Proved Him Wrong - “But after nine months on the market, fewer than 5% of Salesforce’s more than 150,000 customers are paying for Agentforce, according to the company’s disclosures. And more than half of customers that are using Agentforce are still testing it without paying.”

• Development Productivity, Not Developer Productivity - Once you fix the bottleneck of coding, you face the all the other bottlenecks down the line. So, you have to think of how to apply AI, or whatever, to the rest of the SDLC processes. Some real theory of constraints stuff there.

• 🤖 Measuring the performance of our models on real-world tasks - OpenAI introduces GDPval, a new benchmark for evaluating AI models on real professional tasks, showing frontier models nearing expert-level work quality.

• ChatGPT Pulse - Manton reviews ChatGPT Pulse: it might drive traffic to more websites, going around Google: ”There’s something else about how this works that is fundamentally different than current chat-based AI where people are looking for answers. Instead of replacing a Google search, it’s adding opportunities to point to other websites and blogs. Because it’s proactively pushing stories to you that you may never think to look for, it should increase referrers to websites instead of subtracting them. Not enough to offset the lost Google searches, but still notable.”

• Broadcom Is Driving ‘More Innovation’ Than Ever At VMware After Hock Tan’s ‘Tough Decisions,’ Says VMware’s Leader

• 🤖 Honeycomb Launches AI-Native Observability Suite – ADTmag - Honeycomb introduces an AI-driven observability suite with tools for real-time anomaly detection, visual collaboration, and IDE-native debugging.

• 🤖 Harness pitches AI agents as your new DevOps taskmasters - Harness debuts agentic AI for DevOps, promising productivity gains while stressing human oversight and modular adoption.

• 🤖 Amazon blamed AI for layoffs, then hired cheap H1-B workers, senators allege - Senators accuse Amazon and other tech giants of exploiting H‑1B visas after AI-driven layoffs, while Trump’s new visa fee ignites industry backlash.

Wastebook

• “Biting each others’ ankles over who will be the greater fool.”

• I have measured my life in expensing €3 travel agency fees.

• “Everybody talks about the weather, but nobody does anything about it.” - suggested as a quip slide title for IT department harrumphing about modernization and overall digital transformation. 🤌

Conferences

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’m continuing revivifying of my blog. It has most of the links and things you find here, but a few different things. In particular, I’ve given myself the freedom to picture dump there. It can be annoying to see duplication in a newsletter and a blog, but if you don’t care, here’s the RSS to subscribe to it.

Meanwhile, I finally got better audio gear after…15 years? I have a Rodecaster Duo and a Rode PodMic. You can hear the results in this week’s SDT opening. Since then, I think I’ve tuned it even better, so we’ll see. Overall, I really like it. There are other fun benefits like being able to pipe audio into Zoom calls and stuff. With a select few, playing the sad trombone at just the right part in corporate complaining is key to long-term enterprise friendships.

For the audio-heads out there, if you have excellent taste in music, like I do, you’ll really like this remastered classic. I’ve listened to that song hundreds of times, and this version has a new soundscape to it.