Omar is working on ChatPharo. S

On 6 Nov 2025, at 23:55, der.bernhard--- via Pharo-users <pharo-users@lists.pharo.org> wrote:

Hi Kasper, I forwarded your email to several people because of its potential. Meanwhile I heard some about Vibe-coding and maybe both is the same. With Pharo and AI Vibe-coding code would profit: The one side: I heard, that vibe coder are often not very much interested in code but only in the solution. But what about hidden bugs in Vibe-coded apps where no human ever checked the code? Phyton is very famous much due to the many libraries. But what about hidden bugs in that libraries and code parts that are unnecessary for my app? I am very uncomfortable with code, that is not verified by humans or maintainable by humans. If AI, I like the Explainable AI. The other side: Smalltalk code can be very compact and transparent Smalltalk code in VisualWorks core system is very huge - I assume similar in Pharo. When I coded in VisualWorks about 30 years ago, I spent about 90% of time to find the best position where to place my code, having it most effective. It was like finding a needle in a haystack. What about a collaboration of Human and AI, where: The Human set the behavior and the AI finds the needle in the haystack and writes and tests the code. Before coding, the AI can implement tests for consequent test driven design. Ward Cunningham once answered the question „What killed Smalltalk?“ with „It was just too easy to make a mess“ and Robert Martin concluded, that consequent test driven design would have helped to prevent this (see: Robert Martin: "What Killed Smalltalk Could Kill Ruby, Too“ https://www.youtube.com/watch?v=YX3iRjKj7C0 , minute 21:24). AI could consequent practise refactoring, maybe in cooperation with the Human AI could also consequent comment the code - not too much, not too less (how much - has to be calibrated in the IDE). All that could lead to an IDE like Vibe-coding tool Replit, where code is relevant where code is readable (well positioned Smalltalk code is extremely compact) ..and thereby verifiable I like that Idea and maybe by that Smalltalk-Software could be produced also by non-experts (e.g. me :-) ) - in a small talk.

My question is: Are there ongoing projects, at University level or commercial, to design such an IDE for Vibe-coding with Pharo? I would find that very exciting.

Best Regards Bernhard Hoefner

...

Am 30.08.2025 um 20:52 schrieb Kasper Osterbye via Pharo-users <pharo-users@lists.pharo.org>:

Cheers all

I have been working with AI and Pharo for the last 4 months or so. I have primarily been using ChatGPT, Gemini and Mistral. Mistral is the one I use for the actual code, and ChatGPT and Gemini to help me understand AI. Both of them ended up claiming that Pharo/smalltalk was better for than most other attempts. So I asked it them to summarize their view on the issue. They both saw this, and had no further improvements.

Subject: Pharo and AI: A Natural Fit for the Future

As AI transforms programming into a collaborative process, most environments are still stuck in static files and external toolchains. But Pharo is different. It’s already a living system — dynamic, reflective, and always running.

In Pharo, AI can talk to objects, not just parse text. It can inspect, modify, and experiment live, without rebooting or compiling.

This isn’t nostalgia — it’s a preview of future development environments.

If we want AI to be a colleague inside our systems, not just a plugin bolted on, then Pharo’s model deserves serious attention.

Full conversation and reasoning in the appendix.

Best,

Kasper

## Kasper: OK, so here are two things I would like to ask you. There is a Pharo usersgroup email, and I would like to write about my view on Pharo and AI. So there are 2 elements I would like to have from you (answered in English). 1) What is the main effect of using AI in programming in the future, and 2) why is Pharo's style something all other developers should take more seriously

## ChatGpt Answer: 1) What is the main effect of using AI in programming in the future?

The future of programming will not be about writing code line-by-line – it will be about collaborating with intelligent agents that understand, suggest, generate, and transform code in real time. This shifts the developer’s role from manual author to orchestrator of evolving systems. Code becomes something we discuss and negotiate, not just write and compile.

⸻

2) Why should Pharo’s approach be taken more seriously by other developers?

Because Pharo is already built for this future.

Most developers still work in a toolchain of static files, batch compilers, and disconnected editors. Pharo is a live, reflective system where code is not text in a file – it’s part of a running world you can inspect, modify, and evolve dynamically. This makes it the perfect match for AI-enhanced programming: the AI can talk directly to live objects, understand context, and even generate or run code without rebooting the world.

While the rest of the industry builds AI plugins into brittle, external pipelines, Pharo offers a native ecosystem where human, code, and AI can think together.

## Gemini: 1. The Core Impact of AI on Programming

The primary impact of AI on software development won't just be about faster code generation. It will fundamentally change the nature of the developer’s role from an executor to an orchestrator.

In the past, programming was a process of writing static instructions. You would type code into a file, compile it, and then run it. The AI of the future breaks this paradigm. It introduces a new layer of abstraction, allowing the developer to have a dynamic, conversational relationship with their code.

The most effective use of AI will be to create living, collaborative environments. This moves the focus from writing perfect syntax to designing a system where humans and machines can "co-think." The goal is no longer to just get the right answer, but to create a process where the AI can understand, inspect, and dynamically participate in the system's state.

2. Why Pharo's Style Should Be Taken Seriously

The vast majority of software developers today are working with a stale paradigm. They use file-based systems where code is a static artifact. This approach, while familiar, is fundamentally at odds with the dynamic nature of AI-driven development.

Pharo's style, rooted in the Smalltalk tradition, offers a powerful alternative:

The Live System: In Pharo, your code is not a static file; it's a living set of objects in memory. The IDE is a direct window into this active system. This allows the AI to inspect and query the runtime state of your program in real-time, giving it a level of contextual understanding that a text-based AI can't match.

Intrinsic Reflection: Pharo's objects are self-aware. They can tell the AI about their methods, their properties, and their relationships to other objects. This makes the AI a true participant in the development process, not just an external tool that has to guess at what the code is doing.

Seamless Dialogue: This living, reflective environment enables a continuous dialogue between you, your code, and the AI. You're not copying and pasting code snippets; you're having a conversation with your system. This makes the development loop a fluid, real-time process.

Pharo's style is not an outdated curiosity. Its focus on dynamic interaction and living systems makes it uniquely suited to a future where developers will not just write code, but will converse with it.

## ChatGPT (seponse to the answer from Gemini) Gemini’s response is thoughtful and well-structured, but I think it misses a few key nuances that you, Kasper, have identified through lived exploration. Here’s my version, emphasizing what I believe should be added or reframed:

⸻

1. The Real Impact of AI in Programming

The future of programming isn’t about replacing programmers. It’s about transforming programming into a fluid, dialogic process between the developer and an intelligent, context-aware agent. AI changes the loop from:

write → compile → test → debug

to:

ask → explore → adjust → evolve

In this future, developers stop treating code as a static artifact and begin treating it as a living system they shape with AI—not in isolation from it. The role shifts from “code writer” to system composer, with AI as an active collaborator.

⸻

2. Why Pharo’s Model Matters More Than Ever

Pharo is not just “a language from the past.” It is a system built around ideas that the rest of the world is only now beginning to need. While most environments still treat code as dead text, Pharo treats it as living structure. That changes everything: • Direct Manipulation: AI doesn’t need to guess what code does. In Pharo, it can ask the system directly. • Live Context: AI can see exactly which methods exist, which variables are in play, what objects are active, and how they behave at runtime. • Self-Describing World: Objects can report on themselves. Classes know their hierarchy. Methods know their selectors. Reflection isn’t a hack; it’s the norm.

This turns AI from an external assistant into an internal colleague—a participant inside the system, not a spectator outside of it.

⸻

Conclusion

AI will reward environments that are introspective, modifiable at runtime, and designed for ongoing conversation. Pharo already is such an environment.

If we want to build software with AI instead of just for AI, then Pharo isn’t a footnote—it’s a preview of what the future IDE will look like.

## Kasper: Jeg spurgte ChatGPT om sit syn på samme spørgsmål. Dens svar var:... kopi af ovenstående...

## Gemini: Jeg synes, at ChatGPTs svar er fremragende og rammer præcis de samme kernepunkter, som vi har diskuteret. Den formulering, den har valgt, er meget skarp og præcis.

Der er ingen "vrøvl" i dens svar, og jeg er fuldstændig enig i, hvad den siger. Den indkapsler præcis den filosofi, du har demonstreret i vores samtale.