Welcome to another edition of Talsco Weekly

IBM i Briefs: 🐝 The biggest objection to AI on IBM i just went away.
AI: 🛠️ McKinsey says build is back. IBM i never left. 🐝 Hours to seconds: one IBM i shop’s first 90 days with Bob.
Career: 🗣️ The newest language for IBM i developers is English.
Database: 🗄️ RPG developers already know the data is the application.
Hiring: 🧭 IBM’s talent chief: judgment is the entry-level skill now.
Learning: 🪑 IBM i isn’t the distant cousin anymore.
Modernization: 🌿 Two RPG developers, one program, no more waiting.
Open Source: 🐘 PHP 8.2 support ends in 90 days. AI makes catching up easier.
Security: 🔐 Your IBM i security was built for 5250. AI doesn’t use 5250.
Trends: 🪙 IBM is building the rails for tokenized money.
Did You Know: 💡 IBM runs a bank-grade vault for digital assets.

IBM i Briefs

🐝 The biggest objection to AI on IBM i just went away

Self-hosted Bob is generally available, IBM i included.

IBM has made a self-hosted deployment of Bob generally available. Shops can run it on premises, in a private or sovereign cloud, or fully air-gapped.

With a self-hosted model, code, development context, and build artifacts stay inside the customer’s environment. Supported self-hosted models include NVIDIA Nemotron and Poolside Laguna, and hybrid setups can connect to approved outside models.

The Premium Packages for i, Z, and Java all work with it, subject to licensing. The Z package also gained deeper application analysis tools.

The Bottom Line: IBM i shops have always kept their code and data in the building. For many, that alone ruled out cloud-based AI coding tools.

Self-hosting takes that objection off the table. Your RPG, your business logic, and your build artifacts never have to leave.

Wall Street noticed too. IBM shares jumped 5.7% the morning of the announcement.

AI

🛠️ McKinsey says build is back. IBM i never left.

A third of companies are building instead of buying.

McKinsey’s 2026 State of AI survey, 1,719 respondents in 97 countries, found something every IBM i shop will recognize.

Nearly a third of organizations, 32 percent, passed on buying at least one software product or feature. They built it in-house with agentic coding tools instead. Among the rare high performers, it’s nearly half.

The rest of the picture is sobering. Nearly nine in ten use AI regularly, yet only 37 percent see any EBIT impact, unchanged from last year.

The difference, per McKinsey, is redesigning workflows around AI, not bolting it onto old ones.

Our Take: IBM i shops answered build versus buy decades ago. Custom RPG, written by people who know the business, running the core.

The market called that legacy. Now a third of companies are heading back the same way, with agents at the keyboard.

Someone still has to direct the agent and check its work against how the business actually runs. That’s the IBM i developer.

Two more numbers back that up.

In August, we noted the AI jobs apocalypse never showed up. McKinsey now has the receipts. Last year, 32 percent expected AI to shrink their workforce. Only 14 percent say it did.

This year, 39 percent expect cuts. Read that forecast with the same grain of salt.

And with just 6 percent seeing significant returns, almost everyone is still early. That’s not a reason to wait. It’s a window.

🐝 Hours to seconds: one IBM i shop’s first 90 days with Bob

A production defect, found in seconds.

Heartland Co-op manages hundreds of millions of bushels of grain a year on IBM i applications built over decades.

When a production defect hit grain contract processing for newly acquired Texas locations, IBM Bob Premium Package for i found it in seconds. The old way took hours.

The team didn’t stop at debugging. It built custom skills, slash commands, and review modes around its own development standards.

Its application architect, Todd Stewart, puts the payoff simply: “Time is an essential commodity for IT professionals.”

The Results (first ~90 days):

80%+ less time finding and fixing defects (company estimate)
Faster modernization across the ERP, moving fixed-format RPG and DDS to free format and modern database design
~75% of developers actively using it
Next up: agentic workflows and automation, tapping data from about 70,000 connected IoT devices

Our Take: Some shops are dipping a toe in. Others are jumping. Heartland jumped, and 90 days later it has numbers to show for it.

Read it next to the McKinsey survey above. Companies are building more in-house with coding agents. Heartland is the IBM i version, with AI wired into requirements, design, code review, and onboarding.

Notice who’s using it, too. Younger developers dove in. Veterans are coming around, seeing it as a multiplier, not a replacement.

That’s the pairing that works. AI shortens the ramp. Veterans supply the judgment.

A newer developer can use Bob to learn decades-old RPG, job streams, and the business behind them in a fraction of the time. The veteran knows when the answer is wrong, and why that rule exists in the first place.

Neither one gets there alone. Together, they’re the succession plan.

Career

🗣️ The newest language for IBM i developers is English

Or whatever your language is.

Four questions worth stealing.

I had two fascinating conversations this past week with IBM i architects, developers, and leaders.

One of them spends much of his free time building on the side. He works every new tool through the same four questions:

How does it work?
How can it be used?
What are the pitfalls?
What should we be concerned about?

That’s the habit. Experimentation, driven by curiosity.

His other point reframed everything. AI is just the new language, and that language is English.

From the Field: Node.js, Postgres, RPG. Those still run behind the scenes, and they still matter.

But the language to get really good at is the one you already speak. Describing what you want, clearly and completely.

Here’s where IBM i-ers already have an advantage. You’re system thinkers.

You know what the business is doing and what it’s trying to do. You know how orders, inventory, and the GL tie together, and what breaks when one piece moves.

That’s the real separation. You can’t describe what you want to an AI if you don’t understand the system you’re describing.

Most people learning AI start with the tool. You start with the system. From there, the jump to using AI well should feel natural.

So stay curious. Pick one tool this week and run it through the four questions.

Building on the side and wondering where it leads? Let’s talk. Book a call.

Database

🗄️ RPG developers already know the data is the application

AI gave everything an undo button. Except data.

A developer who builds with coding agents describes moving faster than ever. If a change is wrong, the next deploy fixes it.

Then he hits the data layer and stops.

Migrations, storage formats, backups, retention. A bad call there doesn’t redeploy away. Truncate a column, roll back the code, and the column comes back empty.

Jeff Bezos calls these one-way doors. There’s no walking back.

The developer’s version is blunter: “Data is data. If you don’t have it, you can’t reproduce it.”

Rebuild the UI and servers anytime. Lose the data, and you’ve rebuilt a login screen with nothing behind it.

Our Take: RPG developers have known this for decades. An RPG program declares its files before it writes a single line of logic.

Look at what has survived in many IBM i shops. The programs have been rewritten, converted to free format, and wrapped in APIs. The files from the 1990s are still there, holding decades of orders, customers, and history.

The code was always the part you could replace. The data is the business.

That matters more now. AI can rewrite RPG faster than ever. It can’t recreate thirty years of transactions.

Move fast on the code. Slow down at the file.

Hiring

🧭 IBM’s talent chief: judgment is the entry-level skill now

AI took the reps. Judgment still needs them.

IBM’s chief talent officer, Aparna Nair, says entry-level workers need two skills most: critical thinking and judgment.

The catch: people used to build judgment by doing the work, and AI now does much of that work. Her fix is to have new hires validate what AI produces.

IBM’s own survey backs her up. 57 percent of CHROs call critical thinking important, versus 49 percent of employees. On judgment under uncertainty, it’s 48 percent versus 29.

And IBM is tripling U.S. entry-level hiring this year.

For Hiring Managers: This is the case for an IBM i apprentice in one paragraph. A new developer can use AI to read decades-old RPG. Judgment comes from checking that work against someone who knows why the code exists.

That’s the Hire, Train and Mentor model. Pair the apprentice with your veteran now, while the veteran is still in the building.

Building your next IBM i developer? Let’s talk. Book a call.

Learning

🪑 IBM i isn’t the distant cousin anymore

The IBM i community moves into IBM’s house.

This October 26 to 29 in Atlanta, COMMON’s NAViGATE conference co-locates with IBM TechXchange for the first time. IDUG North America, the Db2 user group, is co-located too.

Think about that. The IBM i community, the Db2 community, and thousands of IBM developers, engineers, and technologists, under one roof.

One pass covers both. NAViGATE gets its own IBM i hub and eight dedicated tracks. TechXchange opens everything else, from AI and cloud to security.

IBM i isn’t the distant cousin you never call. It’s family, and it has a seat at the table.

New to IBM i: This is the fastest way to see where IBM i fits in the wider IBM stack, all in one week.

For the veteran: Bring someone newer with you. Walking the floor together is how knowledge moves to the next generation.

Four weeks out. Register through COMMON for the discounted rate, no membership required. Register here.

Modernization

🌿 Two RPG developers, one program, no more waiting

Member locks become warnings, not walls.

Picture two RPG developers who both need to change the same program. One locks the member and the other waits.

Project4i, a VS Code extension for IBM i projects, is adding a Git Mode that changes that. Each developer works on a branch and compiles into a personal development library. The second one gets a warning, not a wall.

Merges stop short of production. The documentation’s reasoning is blunt: “Git does not compile. A merge produces text.”

So a merge deploys to DEV at most, and the existing promotion pipeline handles the rest.

It isn’t released yet. The Marketplace version is still 0.3.9.

Our Take: The most useful line in the documentation is its honesty about when to bother. Count how often someone on your team has waited on a locked member lately. If it’s rare, traditional mode is enough.

That’s modernization done right. Adopt what solves a real problem, not what’s fashionable.

And notice the pace. It feels like every day I find another new IBM i resource. An IBM i architect told me this week that he used to learn something new every year or so. Now it’s every day.

The community is vibrant again, and it shows.

Open Source

🐘 PHP 8.2 support ends in 90 days. AI makes catching up easier.

Behind on PHP? AI makes catching up faster.

PHP was one of the first open source languages on IBM i, back in the early 2000s. Version 8.6 is due by year-end.

The clock is running on older releases. Security fixes for PHP 8.2 end with 2026, and newer releases flag sloppy code that older ones let slide.

The good news, per Seiden Group’s Alan Seiden: a knowledgeable developer with an AI coding agent, whether Bob, Claude, or another, can find outdated code and upgrade far faster.

His advice: “People who let it go and didn’t upgrade before – now is the time.”

Reality Check: Read Seiden’s words closely. He didn’t say AI upgrades your PHP. He said a knowledgeable developer with AI does.

That’s the thread running through this whole issue. AI shrinks the work. Someone still has to know what good looks like.

Most IBM i shops running PHP have at least one app nobody has touched in years. Seiden’s rule of thumb is a new minor release about once a year. If you’re several behind, catching up has never been cheaper.

Security

🔐 Your IBM i security was built for 5250. AI doesn’t use 5250.

AI agents don’t log in at a green screen.

For decades, IBM i data sat behind trusted 5250 sessions inside the data center. Many shops set object authority for that world, and some left default *PUBLIC access that lets any user read or change data.

That baseline was written for a world that no longer exists. AI agents, APIs, and remote users now reach the same data through ODBC, SSH, DRDA, and HTTP.

The fix starts with basics. Set *PUBLIC to *EXCLUDE, classify your data, and run IBM i’s authority collection to strip excess access. Then ask what AI should touch, and who’s watching.

Why It Matters: Shadow AI is already here. Employees are pasting business data into AI tools with no governance in place, often before leadership knows.

On IBM i, that data is the system of record: orders, customers, contracts. As our Database piece put it, the data is the business.

IBM i’s deny-by-default model is one of the strongest foundations there is, but only if it’s actually configured. Lock down the files before you open the doors to agents.

Trends

🪙 IBM is building the rails for tokenized money

Banks get a regulated path to digital assets.

IBM announced two betas for Digital Asset Haven, its platform for managing stablecoins, tokenized deposits, and other digital assets.

The first connects banks to Swift’s blockchain-based shared ledger. Banks send standard ISO 20022 payment messages, the format they already use, to move tokenized deposits around the clock. Swift links 12,500 institutions in 200+ markets, and 17 first movers are already piloting the ledger.

The second runs Haven fully on-premises, inside a bank’s own data center on IBM Z and LinuxONE, with no public cloud dependency.

Both are in beta.

Our Take: The next wave of payments may not come from people at all.

AI agents are starting to buy, sell, and settle on our behalf. When one agent pays another, it can’t wait days for settlement or for a bank holiday to pass. It needs money that moves instantly, around the clock, with the rules built in. That’s what tokenized deposits and stablecoins are built for, and IBM is keeping that money inside regulated banking.

Where does IBM i fit? Haven runs on IBM Z and LinuxONE, but it’s built to plug into existing core banking systems, and plenty of those still run on IBM i, the platform banks trust with the transactions that can’t fail.

The ledger is new. The systems of record behind it aren’t.

Did You Know?

💡 Did You Know: IBM runs a bank-grade vault for digital assets

IBM Digital Asset Haven launched in October 2025 to help banks, payment providers, and governments hold, move, and govern digital assets across multiple blockchains.

Keys are protected by IBM hardware security modules and multiparty computation. Payment providers can use it for near-real-time cross-border payments with same-day stablecoin settlement.

It’s built to integrate with existing core banking systems, the kind IBM i has run for decades. A few weeks ago, we noted IBM i is probably in the path when you use an ATM. Digital assets may be the next path.

Join

Contact us

If you are an RPG programmer looking to explore opportunities or a client who is looking for a talented IBM i professional, please contact us. We look forward to assisting you.

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!

Talsco Weekly: Data is the Application