Javier Tarí

So it has done it, or you still don't know if the result is good?

Be sure to use Opus at least for the plan if you wish to share with me in private, I could try myself

Are you copypasting with a chat AI? And, as today, AI coding agents are junior developers, not senior devs. So they can get stuck on some stuff. As I said, my opinion as of today, is AI coding agents are amplifiers for us

First of all: I'm not arguing on AI applied to coding; that's for eeach one to ponder and decide by itself. My argument is having or not a "place" in DelphiPraxis to talk about Delphi+AI, both on using AI in ourDelphi programs, and to amplify our coding abilities using AI. The study is not relevant for my proposal, but anyway: it's quite crappy is old (along February-June 2025) they were using Claude 3.5/3.7, and the big advancemente for Claude was 4.0 - we are at 4.1) Only 16 devs They were untrained on using AI to develop (c'mon, waiting around?) Claude Code has been a breakthrough; they were using it Cursor Pro. Claude Code was launched May 22th "But in one experiment, their tasks took 20% longer" "AI tools don’t always boost productivity" "found that when 16 software developers were asked to perform tasks using AI tools, the they took longer" "most of them selected code editor Cursor Pro or Claude 3.5/3.7 Sonnet" "Other developers lost time ... waiting around for the AI to generate results."

I disagree. The question was specifically about compiling with Delphi. Don't you think compiling with Delphi is a question for a forum website called "Delphi-PRAXIS"? When a question is about using i.e. MyQSL in Delphi, should it be asked here, or in MySQL places?

That's exactly why I asked for "somewhere" in this forum; so we could share all the right and wrong ways to do this, without "disturbing the peace" of those that consider AI coding anathema and evil I intend to post the few I've learned, sharing my findings and hoping to know what others have experienced To the point: My first solution: a bat file to compile a .dproj from Windows, and then a bash .sh to run that windows bat file My latest solution: a subagent to use the bash command for compiling, review the results, and answer to the main agent with: Success (0 errors, warnings, hints) Failure: X errors (and list of errors) Must work on it: 0 errors, but has a list of warnings and hints, and the main agent must fix them The subagent also changes the extension to .dproj and checks for file existence; informing that "To compile you must have a dproj file" in case it does not exist Without that, sometimes the main agent creates a .dpr but not a .dproj

Yes, you got it wrong. I was refering to the the big advantage of having the AI agent compile it itself, review the errors, fix them, and so on, until it has no errors. It means, as an example, that it would autofix hallucinated classes/methods/... Right now, to me, an AI coding agent is an amplifier of my coding capabilities, not a substitute of a developer. But it can empower me, making me able to deliver better code in less time. Better, because I can spend some of the time recovered to make better code

if you pay monthly, it's 20$/month. If you pay a full year in advance, it's 200$ the year. Quite usual monthly/yearly discount: 10 months upfront for one year Om the other hand, I would not subscribe to anything IA for a full year... I'm very happy with Claude Code, but if next week might appear a new, far better product; this field changes extremely fast

Is that what you do? If not, what did I posted that offended you so much? Do my actions deserve that answer? Because I thought I was just sharing my experience as Delphi developer with AI agentic development, in particular with Claude Code, hoing that this might be useful for others, and then I find myself under attacks

Software development is my source of income, so when I increase the development done (keeping the quality), it is money for me. It's quite obvious that Claude Code is increasing my monthly income for much more than 200$ I don't know in which country do you work, or if you work for yourself or are employed, but each one will have his maths

I pay it monthly. Started with the 20$ one, then upgraded to the 100$ one, and finally to the 200$ one Two days after subscribing to the 20$ one, it was clear for me that I needed it and wasn't going to wait a few hours when I hit the limits, so upgraded to the 100$ one, till one month ago

I follow daily the telegram group, but not this one. And I'm trying to do what I can to boost AI agentic coding in Delphi; I'm on it by some months now, and my experience with Claude Code is quite good, but seems to me that maybe less than 1% of Delphi Devs are using this way of amplify coding

I never program using a chat agent; I use always Clade Code. For your use caso, I would use a chat agent to make a research and "prepare a .md document for an IA coding agent so it can implement it just following the document"; then save it in a folder, launch Claude Code there, and ask it to read the documentation and propose a plan Claude Code should be configured so it can compile the project it is making, and instructed to not consider a project ready until it compiles without errors, hints or warnings

This repository has been done 100% using Claude Code. I've not edited myself even one single line (In my opinion, if you don't know what's Claude Code, and you are a senior programmer, then you are losing money) Even the documentation and the readme have been written exclusively by Claude Code. The only file I've done is the .gitignore. What is this "Automated splash window with fully alive automated progress bar"? The splash screen runs as a second program, so it will always be animated and alive, even if your program is not responsive Your program only sends messages as "Initializing flowers", "seeding whatever", "Randomizing stuff" when it cans The splash screen will show a live progress bar, some animations, and your status messages At the same time, it learns how long it takes from program start to each message, and as soon as it has started at least 3 times, the progress bar will use the times learned, so it will progress correctly It keeps learning, and stores the latest 30 times your program runs, trying to be as precise as it can with thr progress bar I know the aesthetics are not good at all, but you can adjust that at your heart's content Where is the repository? JavierusTk/Splash: Automated Splash screen that keeps alive and autocalculates time left

This is a Claude Code CLI screen while it's working following the indications of the plan we just made to abstract a generic splash screen drawing, to separate the logic. It's not copying lots of code; it's just "typing" implementations. I'm not asking "make for me a bla bla bla" and waiting for a full program. To be copyrighted, it must have a minimum length

No, I don't. Not in my uise case. I review all the code before commiting it, and I'm never accepting big chunks of code that might come from who knows where. Mainly because I must understand all the code in my codebase. I can't think how the code I'm integrating might be copyrighted I can't talk for other use cases, obviously. I also have licenses for all and every commercial component or library I'm using, although there are some developers that download them "from somewhere" And going back to the generated code: We all should remember that the training text does not form part of the AI "memory". The AI has billions of vectors, and that's it... no text chains anywhere

I'll present myself: I'm 59 years old, developing with Pascal since my 18, and with Delphi since D3 appeared. I am the main developer of an ERP which started in 1987 on MSDOS, TurboPascal, all the way till today, D12.3 (soon D13) I code all day long, only in Delphi. There is a few of Python and what not, but I don't code a single line of it myself (you got it: the agent does it) I don't do Vibe coding as in "make a shiny thing for me". Never tried, to be honest I do AI agentic coding, as in "We are going to work with this repository (which we already documented). This is the explanation of what we will implement; analyze it and make the initial implementation plan so we discuss it"

Yestenight I heard about that; the news said it was requested 1.5 but offered and agreed to pay 3 That corresponds to the books they used at the very start: they scratched torrent sites and downloaded and used what fitted them After that, they have purchased one copy of every book/whatever they wanted to use to train, digitalized it, and destroyed the original, so they only had one single copy, equivalent to the purchase Although I find it a curious even interesting talking about it, I fail to see how is it relevant to AI agentic coding I'm sorry for you never having a working, complete solution from an AI. I feel your frustration. And it explains why Anthropic valuation has evolved from 4B$ in may 2023, to 183B$ september 2025 My experience with Claude Code is awesome, and evolving nicely for the latest... 4 months? Feels way longer than that, but it's only that

I'm not talking of predictions of a distant future: I'm talking of my daily experience with, my own work on my own code. According to what I'm doing as of today: They will be unemployed because right now AI is able to program by itself a lot of Delphi code, and even refactor a lot of production code, just with human supervision on the final result, and a single senior developer can supervise a lot of agents coding in parallel. And if your experience is other: if your experience does not include Claude Code, then your experience is zero; you have been visiting the wrong planet

No, they would be unemployed. Except thos of us that will be retired, and not need to worry about it

True, but imo that should change. As an example: Spring4D should be massively used, given the benefits for Delphi development, but it's not. Crazy, isn't it? In my opinion, the topic is so important that it should be encouraged. I believe there are very few of us using that kind of tools on the daily workflow, and imo that's a bad thing

I don't know as of today, but until one-two weeks ago, the only real way of doing it was Claude Code, properly set so it could compile Delphi projects, which would let it find and fix the errors it was doing, so it would end with a running project. I've used it mostly to develop independent projects (.dpr and .dpk), but also let it work in some of my main application (an ERP) packages, my framework-related ones, with very good result I believe it's very sad and even a shame that the Delphi community is not heavily invested on this. We are mostly senior developers, so we should be the ones geting more advantage of this. Adn it could be a way to contributing more to our community, as making code and tools would be easier and faster, as we don't need to code them ourselves, just supervise it

And where should go "Using IA coding agents for Delphi development"? Does not fit within your proposals. And "Off-topic" would be nonsense, as it's a fundamental Delphi coding primitive. Although it seems only those of us already using it are aware of that. Which leads me to ask: Have you tried it¿ I don't mean copypasting on a chat; I mean using agents as i.e. Claude Code, for Delphi. If you haven't it would explain your point of view on the topic

And how did that fixed the poblem? is it "fixed" because MARS is not in a BPL but in your own exe?

Where should be Delphi AI coding with agents be posted?