ChatGPT vs. IBM watsonx Code Assistant Portfolio

I’d definitely recommend ChatGPT to anyone as a great introduction to generative AI and as a starting point in research, writing, brainstorming, or general questions or judgement questions. It can be a great tool to use when you don’t necessarily need an accurate answer. For example, I wouldn’t let it calculate my taxes, but I’d use it to ask some general tax questions, then ask for sources and then verify by checking those sources. I also love ChatGPT for writing and questions - it’s great for emails, creating templates and outlines, and for generating spreadsheet formulas.

Incentivized

I would recommend for understanding your Mainframe components not for the GenAI piece involved from just my experience. The explanations were not up to the quality we wanted but its deterministic side provided a lot of value for different members of my team. The visuals would be great. I am not sure where it currently stands

Incentivized

• Saves time by generating content about a specific topic very quickly
• Allows us to quickly learn information online (from various sources or even a single lengthy article) into more summed up digestible paragraphs (and even bullet points)
• Can autogenerate content on a vast amount of topics

Incentivized

• It can automatically revamp specific parts of the COBOL code and very useful when we want to maintain the existing codebase but improve its structure. I can highlight a block of COBOL code and use Watsonx Assistant to suggest ways to simplify and optimize it.
• Legacy codes, mostly written in COBOL, are cryptic and difficult to understand. Watsonx Assistant analyzes the code and provides insights into its functionalities and dependencies. A great help when working on older applications where understanding the codebase is crucial.
• A step-by-step approach to modernize our applications slowly and steadily, so that we can control the process better. I don't have to change everything at once. Instead, I can focus on specific COBOL modules and automatically convert them to Java.

Incentivized

• Wish it had support for better slides generation. Sometimes we found ourselves using chatgpt to outline a presentation but build it ourselves or use a tool like Gamma
• Maybe a chepear $10 plan. In some countries the US dollar can be expensive and $20 goes a long way.
• I wish you could make projects with more files. They limit it. Or make the limit based on the content, not the number of files per se

Incentivized

• IDE integration is not always optimal
• I wish it could it could generate the code of an entire WEB site based on user specifications
• I with there was a functionality to generate code based on design documents such as Sequence Diagrams, Database Architecture etc.

Incentivized

ChatGPT is a powerful assistant. As long as you understand what it is you're looking for in its results, it can save you a lot of time due to its ability to do the heavy lifting for you. This frees your time up to enable you to concentrate on other tasks.

Incentivized

Most of the time is up. Seldom do you find a down service. It also has improved in token generation (the speed at which it prints answers) so it's usability is pretty much great all the time. Images do take a bit to generate but nothing that breaks anything. New additions like Projects, custom prompts, and some privacy settings improve experience

Incentivized

Their responses are always professional , although they delay sometimes

Where ChatGPT is better: ChatGPT has significantly more use cases - it's much more versatile. Some aspects of ChatGPT's user experience are better than Claude's. I prefer ChatGPT's results presentation compared to Claude's. Where Claude excels: Claude is a more skilled writer than ChatGPT. Some aspects of Claude's user experience are better than ChatGPT's. Its image, audio, and video translations are better than ChatGPT's.

Incentivized

Security is very important in the mainframe world. At Watsonx, we work in the trusted Z environment, which has strong security rules, stricter than those of other cloud-based solutions. My domain is primarily mainframe modernization and Watsonx Code Assistant for Z is specifically used to understand and work with COBOL, the language used majorly in mainframe environments, not any general-purpose language that used in various platforms. It understands the nuances of COBOL and Assembler specific to the Z environment, something crucial for my work.

Incentivized

• Positive: Speeds up research time (freeing up 1-2 hours per day)
• Positive: Provides complex content in digestible formats
• Positive: Allows us to get super granular, including asking it coding questions
• Positive: Allows us to ask learning-style questions where it can further break something down for us

Incentivized

• While manual review and adjustments are still needed, it's a 50-70% reduction in manual coding. Think about it - a project estimated to take a year is done in 4-6 months.
• We've been able to introduce new features and improvements more quickly by updating our technology faster. One relevant example is we recently released an important update to our main product 45 days earlier than planned.
• It has been a smart move and it's really paid off for our company. We've cut down a lot of time we used to spend doing things manually. We now spend our resources more wisely, work faster and finish projects sooner and as a result, we've reduced our development costs by 25%.

Incentivized