How much does AI impact development speed? - “AI sped up task completion by ~21%.” // Aside: Love the implication that 21% is approximate. Are we talking a difference in decimals here, or the actually number? Is it between 21.08% and 21.5%, or 19% and 22%?

If you program enterprise apps, it’s likely in Java. And if you Java, you probably use the Spring Framework. Come to the Spring conference by the Spring people, SpringOne, August 25th to 28th in sunny Las Vegas, Nevada. There’s several sessions posted now: you can see there’s stuff from foundational Spring stuff, AI and MCP, to managing Spring in large organizations. You also get access to all of Explore, which is a whole lot of cloud, platform engineering, DevOps, and ops stuff. Here’s my little pitch for why you should come:

Relative to your interests

• The 2025 CEO Agenda: Leading in an AI-Driven World, IDC - “In what areas have you achieved measurable business benefits from GenAl initiatives?”: Increased operational efficiency: 31%; Improved customer satisfaction: 30%; Improved business resilience: 27%. n=278 CEOs.

• Stop pontificating about other people losing their jobs to AI and worry about your own job - More executives need to use AI for a month straight. This will both de-hype enterprise magical thinking and help us come up with more uses that accrue value to the organization. In the meantime, most of the value of AI accrues to the individual. Which, really, is fine with me…maybe executives should not use AI.

• My 2025 system prompt (LinkedIn)

• Where should AI sit in your UI?. Mapping emerging AI UI patterns and how… - embedded or a chat window on its own?

• Microsoft is pushing Copilot, but everyone just wants ChatGPT - The ChatGPT brand is strong: “OpenAI has said it has 3 million paying enterprise customers, and that number is growing fast. Microsoft told employees that ‘multiple dozens’ of customers have over 100,000 paying users, which would work out to a floor of 2.4 million paying Copilot licenses, but the company hasn’t shared an exact figure.”

• How Goodyear is trying to meet its “big, hairy, audacious goal” - time for tech execs to step up to agentic AI, says CDO Mamatha Chamarthi - Helping “coach” sales people is a popular AI use case. You can imagine prompts like “given what we know about this customer and their business, what products should we sell to them and how should we pitch them? Also, how can we maximize volume, profit, etc.?”

• Training LLMs on books judged as fair use - “In a nutshell, the judge says that legally purchased books can be used to train AI, as long as the models do not reproduce verbatim the original copyrighted works. Pirated books, of course, are a separate issue. They are unlawfully acquired! We can’t steal a book from a store, regardless of what we planned to do with it.”

• The Am Dash - A version of the m-dash that’s purposefully human.

• Detecting AI-Generated Text by Uncovering Its Statistical “Tells” -

• Private Cloud vs. Public Cloud in Federal IT Modernization - More on the Broadcom private cloud survey.

• Expert Generalists - “Traditional interview loops still revolve around product trivia - ‘Explain Spark’s shuffle stages,’ ‘How does Databricks Delta time-travel work?’ A candidate who has never touched those tools can still be exactly the kind of person we need: someone who quickly grasps unfamiliar concepts, breaks complex systems into manageable parts, and collaborates across functions. Focusing on a single stack or cloud provider risks filtering out such talent.”

• Some more good talks from PlatformCon: Rabobank’s platform approach, a product management for platforms story from zolar, and then a story from Elia Group (energy company in BE and DE) that I haven’t watched yet.

The robot read it

I often ask ChatGPT to summarize links for me, here’s a summary for you.

Salesforce’s Adam Evans declared that agentic AI is no longer a science project, unveiling Agentforce 3.0 with dashboards, observability tools, and pay-as-you-go pricing, all to entice companies still “trying it out” to finally start paying.

Tracy Durnell named the corporate GenAI aesthetic the “Business Borg”: an ideology of cheap, abundant, soulless content built on dominance, efficiency, and anti-human taste — designed not to elevate culture but to automate its destruction.

Ted Gioia revealed that longform media is surging across books, music, video, and journalism, as audiences rebel against shortform digital dopamine, with Taylor Swift, YouTube, and even The Atlantic thriving by ignoring the experts and betting on depth.

Wastebook

• “Aesthetics are looks that signal values.” Business Borg.

• Mater/Thread with AppleKit , Alexa, Google Home, etc. - what a cluster fuck of lost opportunities that is.

• In generative AI, you can’t use the word “context” because that word represents a fundamental concept. Instead, you could use the word “intent” in some cases.

• “Bandana of the Good Boy.” Leveling up your pets.

AI Native vs. AI Add-on

Brandon and I tried a 15 minute format for a podcast. We discuss where AI should go in apps and, more importantly, what will work out:

This week, we try a shorter format inspired by the Dithering podcast. The conversation digs into the difference between apps built with AI from the ground up and those with AI bolted on after the fact.

I think I end up falling into the position that it’ll be really hard to inject AI into existing apps (like Office and Google Docs, etc.). I don’t know if I like that position, but it seems historically accurate? Maybe the only modification I’d make is that it will take a long, long time.

Take a listen and tell me if you like the format. We’re thinking of doing more narrowed down AI as the main topic.

Conferences

SpringOne, Las Vegas, August 25th to 28th. Explore 2025 US, Las Vegas, August 25th to 28th. Cloud Foundry Day EU, Frankfurt, October 7th, 2025. AI for the Rest of Us, October 15th to 16th, London.

Logoff

I’m trying to get back to doing some D&D agentic AI this week. My plan is to update it to use remote MCP and then also use services inside a running Tanzu Platform instance. Other than having fun, the goal of all of this is to learn, first hand, what doing agentic AI stuff in Java is like. So far it’s instructive.

The Am Dash - A version of the m-dash that’s purposefully human.

Where should AI sit in your UI?. Mapping emerging AI UI patterns and how… -

Stop pontificating about other people losing their jobs to AI and worry about your own job - More executives need to use AI for a month straight. This will both de-hype enterprise magical thinking and help us come up with more uses that accrue value to the organization. In the meantime, most of the value of AI accrues to the individual. Which, really, is fine with me…maybe executives should not use AI.

The 2025 CEO Agenda: Leading in an AI-Driven World, IDC - “In what areas have you achieved measurable business benefits from GenAl initiatives?": Increased operational efficiency: 31%; Improved customer satisfaction: 30%; Improved business resilience: 27%. n=278 CEOs.

Microsoft is pushing Copilot, but everyone just wants ChatGPT - The ChatGPT brand is strong: “OpenAI has said it has 3 million paying enterprise customers, and that number is growing fast. Microsoft told employees that “multiple dozens” of customers have over 100,000 paying users, which would work out to a floor of 2.4 million paying Copilot licenses, but the company hasn’t shared an exact figure."

How Goodyear is trying to meet its “big, hairy, audacious goal” - time for tech execs to step up to agentic AI, says CDO Mamatha Chamarthi - Helping “coach” sales people is a popular AI use case. You can imagine prompts like “given what we know about this customer and their business, what products should we sell to them and how should we pitch them? Also, how can we maximize volume, profit, etc.?"

Private Cloud vs. Public Cloud in Federal IT Modernization - More on the Broadcom private cloud survey.

Expert Generalists - ”Traditional interview loops still revolve around product trivia - “Explain Spark’s shuffle stages,” “How does Databricks Delta time-travel work?” A candidate who has never touched those tools can still be exactly the kind of person we need: someone who quickly grasps unfamiliar concepts, breaks complex systems into manageable parts, and collaborates across functions. Focusing on a single stack or cloud provider risks filtering out such talent.”