Also: harness engineering teams, the Mario meeting, LinkedIn engagement data, and one million passports.

ICYMI

From the blog since last time.

• DIY Platform – don’t do it – my presentation at PlatformCon 2026.
• The Headless Lifestyle – Software Defined Talk #579 – “Anthropic’s 18-day Fable timeout, whether AI is actually killing jobs, and OpenAI’s trillion-dollar IPO problem.”
• find your customer
• why big companies build their own software, and hire consultants to do it
• here come the normies
• Harness engineering is the new DevOps
• training your replacement
• potential fatigue
• How NatWest is using AI
• 🤖 Ten do’s and don’ts for LinkedIn posts, according to actual research, based on pre-slop posts
• get rid of that crap blocking your YouTube shorts
• don’t keep PII
• Reducing token usage
• “this is the voice of American business that is being channeled through me”
• executives don’t really know how their companies work, AI layoff edition
• 🤖 Findings and CIO Retention Guide: Gender, Priorities, and Exit Triggers in High Tech
• Elusive Enterprise AI ROI: bring in the FDEs

Related to your interests

• Most AI Work Can Wait – Use lower-tier, cheaper models to do more of the work.
• Why AI Coding Costs Are Spiraling: Context, Not Usage, Is the Real Culprit – “Futurum data shows 55.4% of enterprises (n=820) identify AI agent reliability and hallucination management in production as a top challenge, and 50.4% (n=736) actively monitor accuracy and hallucination rates as a formal inference metric.”
• Taste Is in the Spec (Cooking Is Not the Recipe) – “The big problem is that the slop will be in how the software works. The flow. The defaults. The thousand tiny decisions about what to make easy and what to make possible. An agent has no opinion about whether this should be one screen or three. It has no taste. It will cheerfully generate a technically-correct, fully-tested, beautifully-documented experience that is miserable to use – and it will do it ten thousand times faster than the faceless enterprise vendors who trained us to accept misery in the first place. So we will get ten thousand times more of it.”
• Did a lottery winner become “the company of the future”? – Right place, right time: “They won the lottery. Now, they’re smart, and they didn’t squander it. They’re building ecosystems and platforms on top. I’m not against Nvidia. But to call a lottery winner the ‘best company of the future’ says nothing about sustainability, processes, priorities, or scale. How many customers does Nvidia have? Five, maybe – the big hyperscalers, every one of which is desperate to fire them. How many chips does it ship? Tens of millions, each costing tens of thousands. Very profitable. But Apple ships hundreds of millions of A-series chips a year, and more. The silicon Apple puts into people’s hands is orders of magnitude more prolific.”
• GenAI.mil records almost 1.7M users, plans new model additions
• Can Agentic AI Fix IT Incident Management, or Will Complexity Outpace Automation?
• 6 use cases for agentic AI in major IT incident management
• entogo – “entogo turns short keywords into instant redirects. Type gh, land on GitHub. No more digging through bookmarks, retyping long URLs, or hunting through browser history for a page you visit every day.”
• The problem with quarter zip

Wastebook

• “Prince is saying that you should be more worried about losing your mind than you should be about mortality.” You too, must die.
• “All of the planets are deserts, populated only sparsely by aliens races that look like they escaped a BBC show from the 80s.” Ben Brown

AI Summaries

I wanted to read these, but I didn’t make the time, so I asked the robot to summarize them.

• 🤖 OSWorld 2.0: Frontier Agents Complete Just 20.6% of Hour-Plus Real-World Computer Workflows, Failing on Hidden State and Verification, Not GUI Control
• 🤖 FCA’s Mills Review: Agentic AI Will Rewire Retail Finance by 2030, and the Regulator Needs an “Agentic Supervisory Model” to Keep Up
• 🤖 AI Fiction in the Wild

Logoff

I’m on my first work trip for a while. I’m off to WeAreDevelopers in Berlin to give the don’t build your own platform talk.As noted, I’ve had terrible success at getting talks accepted this year. Maybe this fall will be better; I’ve put together a few new ones, and already have several accepted.

Related, this is a really good talk on platform engineering. Just 15 minutes! As ever, I’m jealous of what Syntasso does in the thought-leadership realm.

Want to subscribe to this newsletter and get it in your email? Do that here. You’ll just get this type of link and post round-up, not everything posted on the weblog.