Hi James,
You mentioned it: Project Bob „looks great".
But looking great is not really an advantage, if you need to trust your assistant that things are really true or not.
My experience with the last „big thing" of IBM - maybe someone remembers WatsonX Code Assist for i - I know it is a long time ago since IBM told us nearly the same things about it like they do now with Project Bob... - was that it also looked great but all of my tests lead towards very strange results.
Question: "Why do you think this field is a key field of this file?"
Answer: „Because it sounds like"
No joke. The field had nothing to do with the file....
That is only one experience. I made a lot and got feedback from colleagues and they all said the same... But Bob is totally different - one day ��
When I can test it we will see what it really can and what it can't.
The videos are a little confusing to us because they talk about /36 Applications and in the background you see as the source a QCLSRC and QRPGSRC. ???
I am working with /36 Applications and migrate them for 35 years now but never saw one in a QCLSRC/QRPGSRC.
In another video they show how to „convert" an „old" application to total free and show examples with Sub procedures in fixed format and a lot of /free code which is then converted.
In real world we often find a lot of programs that don't even have sub routines. A lot of RPG developers out there still don't really know what a sub procedure is and how powerful it is so you won't find a lot of code which looks like that example but you will find a lot of code that still uses tons of GOTOs and all of the *INxx Indicators which are not documented at all.
I am really excited to see the examples they use. If these are real world applications or again only that „hello world" examples everybody can analyze and look forward to test it with the source code of our customers.
Another topic is the pricing because nobody knows how much it will cost in total. A lot of people forget that using LLM's in the cloud can be a very cost intensive experience....
My Team and me are working in this area for more than 2 years now and we invest a lot of time into this and all we found out is, that you can't teach a LLM real RPG (in all flavors) in perfection if you don't do something in front of it.
That means: The more you are extracting the data from your code with classic methods, the better results you get.
So we are working together with different vendors to get out the best results for our customers and right now we found two solutions which really work with real world applications and these are Ozgar and another one, I am not allowed to mention right now but if you are interested we can talk about that in a private chat.
Both go different ways like all the others and they are going also different ways like the other one.
Both are able to work with local models and one of it works only with local models but uses more „non AI" features, that it uses AI.
Three things are for sure: Both solutions already work in production NOW - not one day or maybe never but now, both are payable and both create a lot of benefits for the customer.
We also investigated a lot of other „tools" and „solutions" - the „" are because most of them are more or less an old tool that is enhanced by a chat window and most often that even has no access to the existing data and is only a simple RAG based solution that brings only a little benefit if it does at all...
Before you buy something like this you should test the possibilities of the LLM suppliers they deliver, like the „Assistants" of ChatGPT which do a great job when it comes to generate RPG code, if you „feed" them with IBM Reference Documentation and your own. I do that since more than 2 years now and it already worked then and today it works really well.
Claude (which is the most important model of Project Bob for now - and by the way - one of the most expensive models �� ) also generates a very good RPG code (total free) and you can use it directly.
The really cool thing about Project Bob from my perspective is that it integrated into the editor itself. So you can use it like you already use CoPilot within you Typescript, Python, Rust or whatever Code.
But this is something we can also do with Code4i and other Models....
My suggestion to you is, that you should test the tools you are interested in with your (real world) applications and see what they can really do for you.
If you are happy with the result, then everything is fine, go on use them and give some feedback to all of us.
You shouldn't expect too much of all these solutions because they are helpers. They don't do your job (thanks god) but they help you doing it better and faster.
Mit freundlichen Grüßen aus der Südwestpfalz
Markus A. Litters
Geschäftsleitung
IBM Champion
E-Mail: mal@iandme.rocks
Web: https://www.iandme.rocks
i and me – IT Training and Solutions GmbH | Schlossstr. 20 | D-66953 Pirmasens
Original Message:
Sent: 11/12/2025 8:19:00 AM
From: James Cole
Subject: Project Bob IBMI competitors
Hi All,
I know Project Bob is going to be a game changer for IBM i, not to mention across other platforms and use cases. There is so many other posts and information on it and it looks great. Therefore, as there are already vendor products in/coming to the market infused with AI. I just wondered if anyone had any experience any of these and what feedback people had both pros and cons (i appreciate they will personal preferences) however, for me you can't beat hands on experiences :-). I'm thinking that they could be useful use cases for Project Bob like diagramming an application flow etc. Are they integrated in VS Code etc or using their own interface?
I'm trying to leave it open ended so that we can expand the conversation.
Many thanks in advance
------------------------------
James Cole
------------------------------