First, the the elephant in the room.
Yes this newsletter paused for a minute.

In March my son Jack arrived early. Beautiful kid, proud parents.
So the obvious, I’ve been distracted changing diapers and appeasing an annoyed toddler (Tom) for a ‘few’ months.

Good to be back.

Your Weekly “10 Minute Agent”

Set up Claude Desktop and add your first MCP connection. This is your AI supermachine’s foundation.

Look guys, even developers find this stuff kind of hard. AI workflow building and AI agents are tricky, hart to define, etc. The tools are changing fast. But Trust me, you (yes you) can get hands on and can start creating your own solutions.

I really grappled with ‘where to start.’ After deliberation, I think laying a foundation with Claude Desktop + MCP (which requires editing a JSON file, adding Node to your machine and dealing with an API key) is the right point.

If you follow along each week, we’ll get more and more creative in what we build.
Power awaits.

THIS WEEK, DUMBED DOWN

THE GRAVITY OF OPEN AI

If you’ve opened LinkedIn in the past 24 hours, you’ve heard the good news about OpenAI’s newest model, GPT-5. Budding influencers giving their takes before even having access a-plenty.

Open X, and you’ll read how it’s changed everything.

Open Reddit’s AI communities, and you’ll read that it’s a massive let-down.

Personally, I wrote most of this newsletter before the announcement so I’ve not even tried it. But it’s awesome. We’ll see.

Averaging things out, it’s likely somewhere in the middle. Not sentient AGI taking over the world, not a setback for OpenAI either.

This week saw tons of OpenAI news. A continuing trend of news/PR dominance. Some of that below in our select news section, but a few hits:

But our favorite thing of the week wasn’t even in that list.
We just think Google’s new 3D ‘world’ model is wicked cool. Look at it here.

SELECT STUFF IN THE NEWS

  • Claude Opus 4.1 launched, topping o3 in coding benchmarks. A quiet release, but it beat OpenAI models (not sure about GPT-5) on several key tasks. Claude has become the devs’ quiet favorite. Not flashy, just solid. I use Claude models a ton. → link

  • AWS launched AgentCore for building production-ready agents. Lives inside Bedrock. It lets you plug together memory, tools, retrieval, and reasoning in one serverless framework. Honestly, this feels like the toolkit many teams were waiting for. → link

  • Google unveiled AI Agents for data science workflows. These agents can analyze, chart, and explain data like a teammate. The demo felt like the early days of Google Sheets—barebones, but full of promise. → link

  • Microsoft’s new "Click to Do" agent UI rolled out across Windows. It’s a low-key helper that lives in your OS and helps with everyday stuff—summarizing emails, opening the right files, etc. It’s like Clippy got a UX overhaul and a PhD. → link

MORE AI AGENT EXAMPLES

  • Wells Fargo deployed AI agents across operations. Customer service, internal ops, document workflows. It’s a broad rollout. This kind of adoption is eye-catching. → link

  • IQVIA launched clinical trial agents. If you’ve ever peeked into the bureaucracy of pharma, you know how slow it gets. These agents apparently manage trial performance, protocol questions, and prep. It's pharma ops relief in agent form. → link

  • Block built natural-language agents for querying internal data. Basically, any team member can ask questions and get charts. Kind of like giving SQL to people who never want to see SQL.

  • Walmart is trialing agents in stores for scheduling and logistics. Local agents, not mega models. Think assistants that know the store, not the whole world. Cool.

But tell us what you think / want.

Confused about a topic? Have a great agent or AI workflow story (we may tell it). See a gap in the space? Hungry for big reports? Spreadsheets? Videos? We’re (I’m) all ears and eyes. shoot us a message.



Thank you.

Keep Reading

No posts found