Also: Target’s RAG pipeline, the 13-person B2B buying group, Uber on AI dev productivity, and prompt injection as role confusion.

ICYMI

Original content published since last time.

• War for Talent – Software Defined Talk #578 – “This week, we discuss Fable follow-up, Gartner’s AI Governance MQ, and the war for AI talent. Plus, Matt and Brandon review ‘Disclosure Day.’”
• Why not both – build vs buy when AI is the one spending the tokens – Tanzu Catsup

Related to your interests

• Here’s what we learned about AI projects from enterprise buyers so far – Enterprise AI enshitification: “[T]he term I heard repeatedly when talking about OpenAI and Anthropic was ‘Uber-ed.’ Everyone knows that enterprise AI is heavily subsidized to gain users and share. Uber did the same thing. We also know how that story played out as prices went higher as an ecosystem gains leverage. OpenAI and Anthropic are just newfangled plays on the Uber model. If you want another analogy you can substitute Uber for Netflix. Either way, we all know the drill: Things that look good today usually mean you’re screwed tomorrow.”
• From the school of culture eats technology for breakfast
• AI Large Language Models: new report shows small changes can reduce energy use 90% – “Smaller models are just as smart and accurate as large ones: Small models tailored to specific tasks can cut energy use by up to 90%. Currently, users rely on large, general-purpose models for all their needs.”
• 🤖 Inside Target’s LLM-Based System for Semantic Matching in Marketing Forecast Pipelines – Target reframes campaign forecasting as a retrieval problem, using a RAG pipeline of embeddings plus LLM ranking to surface similar past campaigns instead of maintaining brittle rules. It hit 100% coverage at top-3 matches, keeps analysts in the loop, and improves long-tail coverage and interpretability.
• ChatGPT for Pros – How people use ChatGPT for things that are no programming – write-ups and interviews from OpenAI.
• Uber’s journey of measuring AI impact on developer productivity
• B2B Social Media Influencers Have More Influence Than Ever – “Digital natives (Gen Zers and millennials) make up the largest portion of the typical B2B buying group (which now consists of 13 members, on average), and these and other buyers increasingly turn to social platforms to learn, evaluate, and validate decisions. In fact, social media is now second only to conversational AI as the most common self-guided source of information that buyers cite as meaningful.”
• Stop Building Innovation Labs – Yes: “The environments capable of creating meaningful change are almost always the ones that shorten the distance between assumption and evidence – ensuring that learning happens before institutions become too committed to decisions that are difficult to reverse.” But: “The question has never really been whether governments need experimentation. It is whether they are willing to create and protect the conditions that allow experimentation to shape decisions before systems fully commit themselves to the wrong trajectory.”
• Crafting an agent team that still includes me – From the never write a line of code again school of AI: “LLMs and agents basically know everything but your context. There’s a role for you in setting up the full context–instructions, tools, examples, policies, skills, and such–your agent needs. It’s also up to the human to set a goal for the agent to loop on. And unless you completely trust the quality of the output, we have a role in reviewing (and owning) the result.”
• AWS just put $1 billion into forward deployed engineers. Here’s why it matters for enterprise teams. – Enterprise AI needs a lot of custom fitting and one off integration, so you need a lot of post-sales engineering. On the one hand, this is lock-in in the making, but lock-in to your own, stack. You build a stack that is unique to you. On the other hand, it means you get exactly what you want.
• From Container Image to Production: Container Service in VMware Cloud Foundation 9.1 – VMware Cloud Foundation (VCF) Blog – I hope this finally fixes about a decade or obsessing about Kubernetes, finally giving an easy way to tick it off the list: “We have heard from many organizations that sometimes they ‘just want to run a container,’ or they can’t hire the talent they would like to run workloads on Kubernetes” // You get a fine blinking cursor, ready for a platform so your developers can get back to making apps.
• 🤖 From Instance to Fleet: Enterprise-Scale Service Management in Tanzu Platform 10.4 – Tanzu Platform 10.4 makes fleet-wide service lifecycle ops (backup, restore, upgrade, patch) first-class from Tanzu Hub. It also extends the service binding model to VKS apps, giving them the same governed, marketplace-driven path to data and messaging services.

AI Summaries

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

• 🤖 Enterprise AI Buyers Pivot From Model Shopping to Optionality, Cost Discipline, and Architectural Control
• 🤖 OpenAI’s Codex Data: Agentic AI Use Grew 5x in 2026, Spread Past Developers, and Shifted Work From Asking to Delegating
• 🤖 Prompt Injection Isn’t a Bug, It’s Role Confusion: LLMs Identify Roles by Writing Style, Not the Tags Meant to Secure Them

Logoff

Days of doing, but what was done?