Hacker News: andreabergia

New comment by andreabergia in "Show HN: A live Python REPL with an agentic LLM that edits and evaluates code"

andreabergia — Wed, 25 Feb 2026 19:50:42 +0000

> So you could presumably also set it up to implement its suggestions (i.e. if it "executes" its proposed definition for `succ` that would actually just update it in the global namespace). Fun stuff.

Yep, I've had that working in some earlier version of the tooling.

> I'd suggest rethinking this; people usually expect the tab key to auto-complete.

That's a fair point.

> At any rate, thanks for writing the description yourself.

I might let an LLM write the code, and even the user's manual, but I write my own blog post and comments. :-)

Show HN: A live Python REPL with an agentic LLM that edits and evaluates code

andreabergia — Wed, 25 Feb 2026 13:37:47 +0000

I built PyChat.ai, an open-source Python REPL written in Rust that embeds an LLM agent capable of inspecting and modifying the live Python runtime state.

A sample interaction:

    py> def succ(n):
    py>   n + 1
    py> succ(42)
    None
    ai> why is succ not working?
    
        Thinking...
        -> Listing globals
        <- Found 1 globals
        -> Inspecting: succ
        <- Inspection complete: function
        -> Evaluating: succ(5)
        <- Evaluated: None
        Tokens: 2102 in, 142 out, 2488 total
    
    The function `succ` is not working because it calculates the result (`n + 1`) but does not **return** it.
    
    In its current definition:
    ```python
    def succ(n):
        n + 1
    ```
    The result of the addition is discarded, and the function implicitly returns `None`. To fix it, you should add a
    `return` statement:
    ```python
    def succ(n):
        return n + 1
    ```

Unlike typical AI coding assistants, the model isn’t just generating text — it can introspect the interpreter state and execute code inside the live session.

Everything runs inside a Rust process embedding the Python interpreter, with a terminal UI where you can switch between Python and the agent via .

It’s very much a prototype, and definitely insecure, but I think the interaction model is interesting and potentially generalizable.

Source: https://github.com/andreabergia/pychat.ai

Blog post: https://andreabergia.com/blog/2026/02/pychat-ai/

Comments URL: https://news.ycombinator.com/item?id=47151278

Points: 4

# Comments: 4

Untapped Way to Learn a Codebase: Build a Visualizer

andreabergia — Fri, 20 Feb 2026 08:52:17 +0000

Article URL: https://jimmyhmiller.com/learn-codebase-visualizer

Comments URL: https://news.ycombinator.com/item?id=47085425

Points: 244

# Comments: 46

New comment by andreabergia in "Ask HN: Share your personal website"

andreabergia — Thu, 15 Jan 2026 10:33:09 +0000

https://andreabergia.com/ - my personal website and blog, which I'm currently redesigning!

A field guide to sandboxes for AI

andreabergia — Sat, 10 Jan 2026 12:05:18 +0000

Article URL: https://luiscardoso.dev/blog/sandboxes-for-ai

Comments URL: https://news.ycombinator.com/item?id=46565029

Points: 1

# Comments: 0

New comment by andreabergia in "Turning 40"

andreabergia — Fri, 05 Dec 2025 19:27:26 +0000

Thanks! That's really reassuring and will definitely help my nights! :-D

New comment by andreabergia in "The value of hitting the HN front page"

andreabergia — Wed, 20 Aug 2025 08:11:38 +0000

I work at ServiceNow, which sells an enterprise platform and a ton of products built on top of it. Internal teams and customers can extend the platform by writing JavaScript, hence we have a JS runtime in the platform (the venerable https://github.com/mozilla/rhino/), which our team works on.

New comment by andreabergia in "The value of hitting the HN front page"

andreabergia — Wed, 20 Aug 2025 06:26:07 +0000

Same here! I am now actually working in compilers, which is one thing I'm really passionate about, but not something I was doing professionaly. I managed to turn a toy project and some blog posts into an actual job at almost 40, so, thank you HN!

New comment by andreabergia in "Gremllm"

andreabergia — Fri, 04 Jul 2025 16:42:57 +0000

  from gremllm import Gremllm

  # Be sure to tell your gremllm what sort of thing it is
  counter = Gremllm('counter')
  counter.value = 5
  counter.increment()
  print(counter.value)  # 6?
  print(counter.to_roman_numerals()) # VI?

I love this!

Gremllm

andreabergia — Fri, 04 Jul 2025 16:42:57 +0000

Article URL: https://github.com/awwaiid/gremllm

Comments URL: https://news.ycombinator.com/item?id=44465980

Points: 129

# Comments: 18

New comment by andreabergia in "Writing C++ in 2025"

andreabergia — Wed, 21 May 2025 07:37:59 +0000

Of course I didn't assume to be 100% productive immediately, and I did spend a few weeks on this experiment, not just some days. My point was more about how even modern C++ felt (to me) dated and annoying to use compared to more modern language, because of the strong (but very valid!) focus on backward compatibility that C++ has. Furthermore, the long history and baggage means that for any thing, there are _many_ different ways of doing it, which doesn't help.

And anyway, it was not a particularly serious (or, for that matter, well written or argued) discussion. As I mentioned in literally the first line, it's just a rant :-)

PS: I am unsure what you mean with "relying on LLM for configuring things says it all". In my experience, this is one area where LLMs _really_ do help a lot - it has been much faster than going through the documentation (which I also did quite a bit of, in particular for conan).