
Grzegorz Galezowski
3.7K posts

Grzegorz Galezowski
@GGalezowski
coder, OO designer, amateur guitar player and catholic dogmatic theology enthusiast. Author of free TDD book: https://t.co/ugwtQLYyT7
Kraków Katılım Haziran 2012
709 Takip Edilen447 Takipçiler
Grzegorz Galezowski retweetledi

@unclebobmartin Think we need some more concrete measures. Feelings can be misleading. If AI-assisted coding improves people's productivity, it must show up in some numbers somewhere, I suppose
GIF
English

@AdamTornhill If used well, AI can make us understand more code quicker. Not sure why this is talked about so little.
English

@resharper Is it possible to bring own model to Junie or connect with e.g. Copilot subscription or is it only for Jetbrains provided models?
English

ReSharper 2026.2 EAP 2 is available now.
The Junie preview continues with bug fixes for the initial integration as we keep working toward a more open AI agent ecosystem in Visual Studio.
Learn more: jb.gg/rs-junie-acp-i…

English

@resharper ok, got it, I needed to relaunch setup wizard and update again from there 👍
English

@GGalezowski Hi, Grzegorz! Could you please clarify what you mean by connecting Junie and AI Assistant? Junie gets activated with the first prompt in the AI Assistant chat tool window when selected as the agent of choice in the bottom left corner.
English

ReSharper 2026.2 EAP 1 is out with some Big News!
This preview introduces Junie, JetBrains’ LLM-agnostic AI agent, as the first step toward bringing more AI agents into Microsoft Visual Studio ⛓️💥
Learn more and give it a try: jb.gg/rs-junie-acp-i…

English

@davidfowl @pd96dev Seriously this is hardly something to be proud of.
English


@plainionist The whole point is that these prompts are unclear to a degree. I could input the final correct code into the llm and ask it to echo it back, sure, but I am guessing this isn't how we view LLM's usefulness.
English

@vladikk apologies for calling out like this. I'm rereading Balancing Coupling and can't wrap my head around this sentence in the Model Coupling section (included in the screenshot is the question on which I am blocked). Is there something I am missing?

English

@AdamTornhill I found it a fast way to build a "something like that" software. Sometimes "something like that" is all that's needed, many times it's completely unacceptable as is.
English


@mkristensen Access to all the refactorings and context actions.
English

@plainionist What do you think about the second edition? Significant improvement?
English

@plainionist There are design approaches where refactoring can be done with fine-grained class level tests BTW. One is described here: youtu.be/6Bia81dI-JE?is…

YouTube
English

@GGalezowski I like my tests to be a safety net - but if I have to touch them every time I do some non-trivial refactoring they can't serve well as such a safety net 🤷♂️
English

@plainionist I don't treat the first one as inherently a problem. Tying tests to module boundaries is a good thing as long as I pick the boundaries consistent with those I care for in my design approach. Some design approaches care about objects & their interactions, some don't.
English

@GGalezowski From my experience there are 2 main reasons: many UTs map to single classes and so mimic the product design and many UTs are written after the product code is written and so mocks tend to mimic implementation details 🤷♂️
English

@plainionist If this happens often it sounds like a lack of symbiosis between design and testing approaches. Tests assume different properties of design than the implementer puts in.
English

Many UTs I have seen are testing single classes and use a lot of mocks 🤷♂️
Now if I refactor the class - split it in two - the UT is broken. If I refactor the implementation, chances are high that mocks are broken.
These effects are inherent to unit tests. Acceptance tests typically don't have those issues
English

@plainionist So if a test is too coupled to implementation details of the tested boundary it's typically a badly written test or the boundary is badly designed. Not inherent to unit tests. UTs are typically criticized for testing at łevel where the design struggles to keep a lasting form. 🤔
English

@vladikk are you maybe considering publishing some kind of workbook companion for the Balancing Coupling in Software Design book? I am rereading this book and I think I am getting the hang of the models, but could use more of curated examples on different abstraction levels.
English




