Welcome to another edition of Talsco Weekly

In This Issue

• IBM i Briefs: ⚙️ IBM’s new Power admin agent, in plain English. 📊 IBM changes the story it tells Wall Street.
• AI: 🔌 Ask your IBM i a question in plain English.
• Career: ⚖️ What becomes of the RPG developer when AI writes code?
• Development: 🔁 The RPG loop bug that processes your last record twice. 📇 From column templates to AI: 46 years of RPG.
• Hiring: 🎯 AI is now part of every IBM i job I’m filling.
• Learning: 🚦 Act, ask, or abstain: when to trust AI on IBM i.
• Security: 🔒 AI found the bugs. Now who patches them?
• Did You Know: 💡 RPG is older than the moon landing.

IBM i Briefs

⚙️ IBM’s new Power admin agent, in plain English

IBM started shipping Power Autonomous Operations on September 23. So what is it, actually?

Think of a junior admin who watches your HMC around the clock, answers questions, and can make small moves without asking.

Ask it to list every IBM i partition with its cores and memory, and you get the list. No clicking through multiple HMCs.

It also acts. If an LPAR runs past 85 percent, it can add capacity. If one sits overprovisioned, it can take some back.

Low-risk changes are auto-approved by default. Bigger ones wait for a human.

Before you turn it on: Four things worth knowing.

It doesn’t see inside IBM i yet. It works at the partition level: cores, memory, hardware errors. OS integration is on the roadmap.

It runs as its own AIX or Linux appliance on Power10 and Power11, and needs a one-way cloud connection. Air-gapped shops wait.

It uses capacity itself: two cores, 32 GB of RAM, 500 GB of storage.

And IBM’s 15X faster claim comes from its own testing on 11 systems.

📊 IBM changes the story it tells Wall Street

IBM’s investor pitch has shifted from its running AI tally to recurring software revenue. The test comes Wednesday, October 21, when IBM reports Q3 against a lowered 6 to 8% software growth forecast.

IBM (NYSE: IBM): $234.12 (Oct. 15 close, up 2.7%) · 52-week range $199 to $332

AI

🔌 Ask your IBM i a question in plain English

A free, read-only AI bridge to Db2 for i.

An open source tool covered by ITJungle this week lets AI assistants like Claude and Cursor query Db2 for i. So what is it, actually?

Think of a translator that sits between your AI assistant and your database. You ask, “Which customers have the most open orders?” It writes the SQL, runs it, and hands back the rows.

It only reads. Anything that isn’t a SELECT gets rejected before it reaches the database.

Nothing gets installed on the IBM i. It uses the host servers you already run, and queries run with the user’s own IBM i authority.

Our Take: This is the kind of AI on IBM i worth trying first. It’s free and read-only, and your existing object authority still decides who sees what.

One catch before you point it at production. The answers travel to whichever AI assistant you use. Make sure it’s one your company has approved.

Then start with a question nobody built a report for. Which customers ordered less this quarter than last?

Career

⚖️ What becomes of the RPG developer when AI writes code?

IBM’s new study points to judgment, not typing.

So what becomes of the RPG developer when AI can write the code?

IBM’s Institute for Business Value recently asked 1,500 HR chiefs and 8,800 employees a bigger version of that question.

71 percent of the HR leaders said the most essential skill now is supervising, validating, and overriding AI’s output. Only 38 percent of employees agreed.

And 60 percent of employees worry AI is eroding their skills. Critical thinking topped the list.

That’s the gap. The job isn’t disappearing. It’s shifting from writing every line to judging every line.

From the Field: I’m hearing both sides of that study right now.

One client told me AI has been genuinely valuable for digging into code and tracking down errors. But it’s not always right. Not by a long shot.

I’m also working with a seasoned RPG developer, 40 years in, who jumped into AI headfirst. He’s never been more excited about his work.

His decades of RPG, Db2, and IBM i know-how are what make AI powerful in his hands. He knows when it’s wrong.

That’s the developer every shop will want. The loop bug in this week’s Development section is exactly what he’d catch.

Decades of RPG behind you? AI may have just made you more valuable than ever. Book a call.

Development

🔁 The RPG loop bug that processes your last record twice

Dow, Dou, and the case for an early exit.

Jim Buck’s excerpt from Programming in ILE RPG covers loop moves every RPG developer uses daily, and one that trips people up.

Iter skips to the next pass through a loop. Leave exits the loop entirely.

Then the classic file loop. Dow needs a priming read before the loop.

Dou can get by with one Read, but skip the Leave at end-of-file and the last record gets processed twice.

From there it moves into top-down design, subroutines, Leavesr, and *INZSR for setup work.

Next up in the series: control breaks and navigating legacy code.

📇 From column templates to AI: 46 years of RPG

The cheat sheet that read your green bar.

Simon Hutchinson of RPGPGM.COM dug a relic out of his desk drawer: an RPG Debugging Template.

Back in the System/36 days, this folded card mapped out the RPG II specifications. Lay it over a green bar printout and you could tell which column held which operation or indicator, and what the code would actually do.

Fully free RPG eventually dropped the columns altogether.

The best part is in the comments. In college, one reader mocked RPG as the “fill in the blanks language.”

Forty-six years later, he’s still writing it, and says AI at his side has renewed his enthusiasm for the work.

Hiring

🎯 AI is now part of every IBM i job I’m filling.

Management to admin to RPG, the ask is the same.

I’m seeing steady demand across the board right now. IBM i managers and architects. Admins covering Power, the OS, and disaster recovery. Senior RPG developers. Direct hire and consulting alike.

One thing shows up in every search: AI.

Some companies are all in. Some are experimenting. Others are just dipping a toe.

That spread is creating some indecision. Hiring managers aren’t always sure what the role will look like.

My take: whatever the role becomes, hire someone who genuinely wants to learn AI and put it to work.

Not sure what your next IBM i role should look like? Let’s define it together. Book a call.

Learning

🚦 Act, ask, or abstain: when to trust AI on IBM i

Three words every AI rollout needs.

IBM’s Thomas Suedbroecker has a sharp follow-up on trusting AI agents. His point: an agent can take too much risk, but it can also take too little.

Act too often, and it makes confident mistakes. Ask a human every time, and you’ve automated nothing.

His answer is three choices. Act when confidence is high. Ask when more information would settle it. Abstain when being wrong costs too much.

The key move is pairing confidence with blast radius. Renaming a variable and deleting production data should never clear the same bar.

New to IBM i: Before you hand Bob or any assistant a task, sort it into one of the three buckets. Formatting a program is an act. Changing a file layout is an ask.

Been on the platform for years: You’ve already seen this framework ship. IBM’s new Power admin agent, up top in this week’s Briefs, auto-approves low-risk changes and sends bigger ones to a human. That’s act and ask. The real question is who decides where the line sits, and whether it’s written down anywhere. What’s the blast radius in your shop if the agent gets it wrong?

Security

🔒 AI found the bugs. Now who patches them?

IBM X-Force tracks a 578% jump in disclosures.

IBM’s X-Force team counted publicly disclosed vulnerabilities across Microsoft, Oracle, Google Chrome, and Apache. From January through September, the total went from about 1,500 in 2025 to more than 10,000 this year.

The surge started in April, around the time AI-assisted vulnerability research took off. IBM stops short of calling AI the cause.

The flaws aren’t trivial, either. Of 223 vulnerabilities traced to frontier AI models, 55% were high or critical, and 91% were remotely accessible.

X-Force’s conclusion: finding vulnerabilities is no longer the hard part. Patching the right ones fast enough is.

Why It Matters: None of those are IBM i numbers. But your IBM i doesn’t run alone. The Windows desktops, browsers, and Apache software in and around it come from exactly the vendors in that count.

And IBM i shops are built to change carefully. Bigger patch cycles on a tighter clock collide with that culture.

So ask the question before the auditor does. Which fixes is your partition missing right now?

If the answer takes more than one query, start there.

Did You Know?

💡 RPG is older than the moon landing

IBM introduced the Report Program Generator in 1959 for the IBM 1401. Its job was to help shops move off punched-card tabulating machines.

That’s where the RPG cycle comes from. It copied the machine cycle of the tabulators it replaced.

Sixty-seven years later, RPG IV still supports the cycle, and *INLR is still a nod to “last record.”

💭 Before you go: something I’m chewing on

AI compresses the doing far more than the knowing.

I’m seeing it everywhere right now. Analysis that took hours takes minutes. Code nobody had time to document is finally getting documented.

But the speed only helps in the right hands. The developer who knows what to ask gets the right answer fast. Someone who can’t judge the output gets a confident mistake, just as fast.

And as everyone picks up these tools, the bar rises for all of us. More output becomes the expectation, not the exception.

I don’t have this all figured out. Am I on to something, or missing it?

What are you seeing in your shop?

Share

Do you know of someone who could benefit from Talsco Weekly? If so, please use the social media buttons to spread the word. Thank you!