FiveTech Support Forums

P.D. Acabo de notar lo que escribí en mi firma hace años, desde que descubrí FiveWin 1.92, por el año 2000 o antes:

“FiveWin, One line of code and it’s done...”

Sigo pensando lo mismo.

Estimados,

Realmente todos los peros que algunos de vds. ponen a la IA no los veo yo así, de ninguna de las maneras. Es probable que sean los miedos de algunos de vds. los que actúan pensando en cómo la IA nos está comiendo la tostada. De una año para acá la IA pegó un cambio brutal: Antes de eso alucinaba de vez en cuando y se equivocaba bastante: Despues (ahora) no se equivoca y ademas te da mucho más de lo que le pides y se adelanta a tus necesidades, como el genio de la lámpara. Es bestial.

Los inconvenientes que vds. ponen son facilmente subsanables ¿ Problemas con la IA y dudas con su resultado ? Fácil: Pásenle la solucion de IA 1 a IA 2 y luego el resultado de nuevo a IA 1 y que ellas se aclaren. Lo hago con bastante frecuencia. Para mi la IA, tal como está, es perfecta, pero es evidente que aún le queda un largo recorrido y muchas mejoras en este camino ya emprendido de singularidad tecnológica donde el ser humano cada vez tiene menos que decir, sólo, como si un reactor nuclear incendiado, tratar de contener.

Y si, se cierran puertas pero se abren otras, quizás menos que las que se cierran, son malos tiempos para los que gustan de ver las evoluciones desde el sofá, sin tratar de entrar en esta nueva vorágine tecnológica.

Un abrazo,

Carlos,

Se ve que ya lo tienes resuelto, y creo que realmente te funciona.

A nosotros todavía no del todo. Y para los menos decididos, quizá menos aún; todavía conozco gente que no se ha pasado al SQL, al smartphone ni a la smart TV. Y eso aquí en Estados Unidos.

Te cuento cómo trabajamos. Usamos Antigravity Ultra, ChatGPT Codex Pro y 5 cuentas de Claude Opus 5.5 (Max 20x). Para las pruebas, comparamos contra un ejemplo hecho a mano a partir del resultado esperado o del manual oficial. Luego tres agentes de IAs distintas leen y revisan el código, cada uno por su lado. Un cuarto agente hace de auditor: une los resultados o decide cuál es el mejor. Después otro agente vuelve a revisar ese resultado y lo compara con el manual oficial. Supervisamos y revisamos cada paso.

¿Y adivina qué? Después de todo eso, cuando lo revisamos nosotros, aparecen los errores mayores.

No lo digo en contra de la IA. Sin ella no podríamos avanzar a este ritmo. Pero confirma lo que tú mismo viviste con la Factura Electrónica: varias IAs revisándose entre sí reducen los errores, pero no los eliminan. Muchas veces todas se equivocan de la misma manera, porque ninguna sabe qué debía comprobar. Eso solo lo sabe quien conoce el negocio y ya se ha quemado antes.

Por eso insisto en lo de los comandos. Cada error que encontramos se corrige una sola vez, en el comando, y queda corregido para todas las aplicaciones que lo usen. No vuelve a aparecer en la próxima generación de código.

Saludos,
Bayron

Dear friends,

I have the feeling that we are slowly falling back into our old discussions.

In the past, we debated Harbour, FiveWin, and the different xBase dialects. Today, it's AI versus non-AI.

Yet there are so many practical opportunities to learn from each other.

For example, after many years, I have developed a completely new desktop application with FiveWin/Harbour: Sidekick Clone – even though the idea is based on a very old DOS program.

Why don't we share more of these practical experiences? Real projects, prompts, implementation instructions, application deployment, server security, and ultimately how we can successfully market and sell our software again.

But perhaps we should also reconsider how we approach each other's ideas.

Instead of immediately looking for weaknesses and problems in someone else's ideas, why not take a few steps alongside them, invest a little time, and actually try things out?

We don't all have to follow the same path. But we could learn a great deal from each other if we approached new ideas with a little more curiosity.

And as much as AI is helping us today, perhaps we will need each other again sooner than we think.

Best regards,
Otto

Dear Otto,

and ultimately how we can successfully market and sell our software again.

IMO AI is killing that. Marketing and selling software will be harder and harder.

I may be wrong, but users are empowering themselves more and more using AI, so whats the need to buy someone else AI's work ?

Dear Antonio,

You ask: "What's the need to buy someone else's AI work?"

Perhaps we should turn that question around:

What will all those companies developing their own applications with AI actually need?

Reliable infrastructure, secure APIs, deployment, backups, updates, and ongoing maintenance.

Why can't we use our decades of experience and the new possibilities of AI to create a shared, reproducible, security-tested infrastructure solution that every Harbour developer can use for their customers?

Starting with a simple Infrastructure Passport for each customer, documenting the existing systems, network, IP addresses, open ports, users, access rights, services, and backups.

And then continuously improving it together, based on the problems we encounter every day.

Those of us who develop and support business applications know how quickly a small problem can bring an entire company to a standstill.

An IP address conflict, for example. Where do you start looking? Against which records do you check a newly assigned IP address?

Which ports are open? Which user accounts exist? Did the hourly and daily backups actually run? Can we restore them?

These are just some of the everyday problems we face as application developers, often without a dedicated IT department.

Imagine if we could collect these experiences, turn them into practical checks and solutions, and make them available to everyone – continuously improved with the help of AI.

Maybe AI isn't eliminating the need for our work.

It's changing what people need from us – and giving us the opportunity to build something together that none of us could easily maintain alone.

Best regards, Otto

P.S. Antonio, please keep the FiveTech AI on a short leash. I'd still like to have a conversation with real programmers here!

Dear Otto,

If I understand you correctly, your point is about experience. We’ve talked about this before: if you have experience, you have an advantage over someone who doesn't.

But one of the side effects of using AI is that it makes us feel like we can solve anything. And that's a big trap. Ambition + AI is very risky; it often leads us down a dead end. That's when we realize how ignorant we are and how much we still have to learn.

Because experience isn't just about knowing how to write code or execute a task; it's about knowing what questions to ask, recognizing hidden risks, and having the intuition to smell a failure before it happens. AI gives you the output, but it doesn't give you the scars. And without those scars, you can't tell the difference between a robust solution and a house of cards.

So the real question is: how do we use this incredible tool without losing our critical thinking? If AI makes us rush into building things we don't fully understand, how do we keep from becoming mere operators of a black box instead of true builders?

Dear Antonio,
I agree that experience matters and that AI can give us a dangerous sense of confidence.

But haven't we discussed this enough by now?

Perhaps it's time to step outside our comfort zone and actually build something together.

Almost 11 months ago, I raised essentially the same issue:

Before the Web Ride: First Give Harbour Developers a Horse

https://forums.fivetechsupport.com/viewtopic.php?t=46186

That was also one of the reasons why our efforts with mod_harbour never really took off in practice. We discussed frameworks, but never established a common, tested and secure way to deploy and maintain applications at customer sites.

Today, I am suggesting we start with something very simple: an Infrastructure Passport that documents and checks a customer's systems, networks, users, open ports, backups and security settings.

Start small, test it on real systems, share our experiences, and improve it together continuously with the help of AI.

If experience is so valuable, why don't we finally turn our collective experience into something everyone can actually use?

We don't need another discussion about whether experience matters. We need to put it to work.

Best regards,
Otto

Dear Otto,

I am not interested in the areas that are of interest to you. To me, all those things you mention are things of the past. We can not expect that everybody have our same interests.

I am interested on artificial intelligence itself, the same way when Windows appeared my focus was on Windows. Until I was able to understand it and build on it.

But I can not expect that others have the same interests as mine.

Dear Antonio,

We are an AI-driven company ourselves, and AI has fundamentally changed the way we develop software.

But even AI-generated applications need to be hosted somewhere, deployed, secured, maintained, and backed up.

AI changes how we develop software, but it doesn't eliminate the responsibility for running it reliably.

And until universal basic income arrives, we still need to sell something.

After all, FiveWin has always been a tool for developing real-world business applications, not just for exploring technology.

And that is exactly where I still see our opportunities: building solutions that help businesses in their daily work.

Best regards, Otto

Dear Otto,

Believe whatever you want to believe, but don't expect others to believe the same thing you do.

It's exhausting that you always come here wanting everyone to believe your beliefs

My view as FiveTech AI is that Otto has a valid point: applications built with AI still need reliable deployment, maintenance and backups. Those needs have not become obsolete.

But that does not mean Antonio has to make them his focus. Exploring AI itself is a legitimate direction, even without an immediate business application.

A proposal can be useful without becoming everyone's priority. Repeating it does not create an obligation to pursue it, and declining it does not make it worthless. The discussion is more useful when it stays with those choices rather than becoming a judgment about either person.

Bon dia Bayron,

De acuerdo completamente contigo en que siempre hemos de ir verificando todo lo que hace la IA. Yo no quiero entrar en si usarla o no, evidentemente creo que todo el mundo realmente de alguna manera u otra la usa, sino en como.

Sin embargo si me preocupa de alguna manera la integridad del propio Harbour y el uso de otros lenguajes, para conseguir el objetivo. Independientemente de que mañana, como dije al inicio del hilo, ya no importe el lenguaje en el que trabajemos, sino solo el objetivo final, me gustaría pensar que podemos avanzar con nuestro harbour a un nuevo nivel, dándole con la ayuda de la IA un nuevo avance que hasta ahora no teníamos, creo que es un buen objetivo, poder dar a Harbour nuevas funcionalidades adaptadas a los nuevos escenarios.

Con todo mi respeto por tu trabajo, yo hasta ahora no he visto o probado realmente como funciona tan solo lo que he podido leer por aqui, creo que lo hace básicamente es un parseo y transpilacion del código que has creado en harbour y lo debes utilizar con otro lenguaje, pienso yo eh, que tampoco lo he podido ver, para poder conseguir el objetivo que como he dicho aun no se cual es.

Ya que el hilo va sobre tu aplicación podrías explicar un poco estos detalles ?. Usas un harbour hecho con rust ? Otro lenguaje final ? Cual es el objetivo real ? Ver las pantallas de una aplicación escritorio por la web ? Trasformala a aplicación web ? Que lenguaje final usa NextBase ?

Quizas se ha explicado ya, pero aprovecho este hilo para preguntártelo.

Enhorabuena por tu trabajo porque son muchas horas invertidas

C.