A New RAD Studio Workflow Is Coming

,

This sounds interesting. I have signed up but it isn’t a very friendly time (1am AEST 29 May) - may have to catch the replay.

Something new is coming to RAD Studio, and it is built for the way Delphi and C++Builder developers work.

In this live event, you will see first-hand how Embarcadero is bringing a new level of intelligence into application development. You will learn how this new workflow helps developers move from idea to implementation faster, work through code more smoothly, and reduce the friction between writing, building, reviewing, and improving applications.

However you currently develop your applications, this webinar will give you a practical look at what comes next.

https://lp.embarcadero.com/NewWorkflow

@ianbarker Any hints? :smiley:

No hints but I promise I’ll make sure the replay is available for you all. I’ll speak to Malcolm at Code Partners and maybe organize another live thing later the same day at an Aussie/Kiwi friendly time for a Q &A.

11pm on the Thursday for us in Perth

There are some internal details that would have a huge impact but have not been publicised yet.

And of course they have unavoidably made a number of assumptions, so the usability is also a question: some may love it and some – hate it, because it’s generally rather intrusive and can get in the way more than help.

So, at this point it’s just a marketing pitch, trying to upsell (it’s a separate component, extra subscription fees). It’s not a lot, and it’s probably worth it in the long run, but there are so many alternatives and the software is rapidly depreciating as we speak, so it’s not a certain “must have” IMHO.

Lastly, it’s regrettably a bit too late. I.e.: I have already settled on some specific process after long trial and error and will not switch, not immediately anyway.

In my opinion, Embarcadero solution is superior to RemObject’s, though.

Alex

That’s not something we often hear when comparing Embarcadero offerings to their third party equivalents.

Please elaborate.

True enough.

The main thing for us was that RemObjects have their own server doing some arcane behind the scenes magic, while at the same time exposing all your code not only to whatever the behind-the-scenes LLM is but also to this RemObjects server itself, which did not impress us at all. It also reduces your control/flexibility, adds latency and has its own very real availability risks. This may be changing from version to version too, I suppose. I figured they have some nice tricks to make it work better, which they must protect from prying eyes.

Embarcadero’s track record may also be a bit patchy here, but its 1) deeper integrated, 2) from the compiler vendor itself, 3) flexibility and [at least some] control are all good things. Provided it passes the test of actually working, which it hopefully will.

Interesting that Emb did not take RemObjects’ demo path, with a more complicated task, but went on doing the simplest things – it could be telling.

But if it has built-in, functional and documented extensibility options for more MCP tools, and exposes all underlying config, per LLM, I’ll vote for it (including with my $$$). From my experience, tools (or rather their quality, which is hard to see from outside) are much more important than they are given credit for.

Alex

Here’s the replay:

And there is a short promo video on the product page:

I have only watched the promo video but it is good to see.

My initial question is why it is an additional cost. Is this to not directly impact the two commercial third party products Chucki and CodeBot?

My guess is it’s two-fold. It’s new so there’s probably bugs he didn’t want to encounter during the webinar and but I also think it was strategic too.

At this point most Delphi developers who are going to fully embrace AI coding, are likely already writing their Delphi code using Claude Code, Cursor and a bunch of other tools I don’t know about and rarely using the IDE.

This is the light touch AI assistance for the rest of us (and I’m guessing that’s not a small number). If you’re still hand coding in the IDE, myself included, you’re probably not crazy about becoming as Vincent described himself on the Symposium Zoom chat, “a Claude code babysitter”.

I’m watching the webinar as I write this and I think I can see myself using it, at least 2 of the 3 main features anyway.

You can’t say no to improved code completion so that’s the first feature I know I’ll use. Secondly the chat window and it’s ability to interact directly with the project files will simplify my current ad-hoc workflow of occasionally copying and pasting code to/from AI web chats when I want a second opinion on something.

The “suggestions” feature is the one I’m less enthused about, but that’s more to do with me than the feature itself. It seemed usable enough in the demo.

So I’ve just finished the webinar. The first half, the pre-recorded parts, are all worth watching.

Once it gets to the Q&A, if you find yourself yelling “Get to the point Ian!” :wink: here’s some timestamps you can skip ahead to.

  • 54:22 Marco gives details on the models used
  • 59:00 the CEO addresses why Kai is a separate cost
  • 1:06:06 Marco discusses code privacy.
  • 1:12:15 Marco discusses Kai with third party components
  • 1:16:33 Marco discusses how to download the trial and get a license key

The first two of those timestamps contain the most meat.

:rofl:

Sorry. I’d had 3 hours sleep in 48 hours and a lot of caffeine. I thought I was actually quite reserved compared to normal!

Also, insider secret, this was actually my THIRD webinar of the day (and the only one about Kai). It’s been an epic week.

BTW - if you want to watch the webinar pre-recorded video content without the live Q & A (and none of me in it, although I did the video editing and creation) you can find it here: https://youtu.be/ktAkIaoAwI4

You were, that’s why was only yelling at my screen instead of throwing things at it. :scream:

My favourite bits were when you tried to end the webinar about 3 times and the others wanted to keep going. Usually they have to cut your feed to stop you talking.

I have trained them too well :face_with_peeking_eye:

I checked my Apple Watch, its not 2024. Heard of Cursor, Claude Code, C#, React/NextJS lol…

Seems to be a pattern, too little, too ugly, too late, too expensive, less capable.

Luckily with the power of AI its so easy to convert even huge Delphi projects to modern platforms and move forward without even hiring any new devs.

Luckily, the power of AI also extends to letting them train on huge Delphi projects, and speed up development of those projects. And it works perfectly well outside the IDE for those who prefer that. :+1:

Regarding the 3 minute introduction, I felt like I was watching a flight safety video. I wonder what was in my green tea today…

I watched the Kai presentation. I can see the benefits, but I’m an old dog, and learning a new trick like team programming with an AI in Delphi is not going to feel natural immediately.

On the other hand, some of the generated code and assistance was astounding, and I did utter a few expletives as I was reading. Increasingly, LLMs are interfacing directly with my personal RLM (Rude Language Model).

Sure, of course we have.

Which of those do you use?

I use Cursor now for all my work. Its 10x - 100x on what I could achieve before and honestly its not just about code - its research, strategy, time saving on approaches and design, its everything.

All my web work moved to React/NextJS and some business systems moved to C#

Once you move you’d never go back, but as a previous Delphi contractor pretty much every Delphi product that was going to survive the next 10 years in Australia has taken a similar migration path since ~2018, thus the close to zero Delphi jobs in the market for a year.

It was clear years ago that Rad Studio needed to be retired and all effort needed to be put into making the form designer + toolchain VS Code compatible. Had this path been taken Delphi would be working natively on Mac, Linux & Windows and been a first class citizen in Cursor which would at least have given it some hope of survival.