How Accurately Do Large Language Models Understand Code?

https://arxiv.org/html/2504.04372v1

"This paper presents the first large-scale empirical investigation into the ability of LLMs to understand code. Inspired by mutation testing, we use an LLM’s ability to find faults as a proxy for its deep understanding of code. This approach is based on the insight that a model capable of identifying subtle functional discrepancies must understand the code well."

It appears that coding LLMs are vulnerable to misleading code comments, misleading variable names and misleading dead code. They still have shallow understanding of code, based on syntax and tokenization designed for natural languages, instead of analyzing code semantics. Writing a lot of incorrect comments can confuse them

via @dotnet : Emit logs on-demand with log buffering

https://ift.tt/vdfE09D
#DotNet9 #LogBuffering #Logging #SoftwareDevelopment #TechInnovation #ErrorHandling #PerformanceOptimization #ASPNetCore #MemoryManagement #DiagnosticTools #LoggingStrategies #Production…

I must say I was involved about midway through my software development career in an early form of AI. At the time we called it rules-based system, and then expert system. It actually seemed intriguing and was custom-molded for the solution I was trying to architect at the time. It also was very constrained and didn't have the "ick" feel of the reach of AI nowadays.

Is there anyone from that timeframe that can share their experiences? I hope I'm not seeing it through rose-colored glasses.

https://en.wikipedia.org/wiki/Expert_system

Senior #softwareengineer out there: how do you keep up on coding and other hands-on skills when your daily job is basically writing documents, attending meetings and “work through others”?

Retoot to reach more people, pls!

Red Flags to Watch for When Evaluating Developer Productivity Tools

https://jellyfish.co/blog/5-red-flags-to-watch-for-when-evaluating-developer-productivity-tools/

Andy Pryor is 'Rethinking User Workflows with Generative AI' July 25th at Nebraska.Code().

https://nebraskacode.amegala.com/

Track Errors First

https://www.bugsink.com/blog/track-errors-first/

Nested loops are often discouraged, but they make perfect sense in scenarios like handling multidimensional data or generating patterns. In this @realpython article, Amanda Adoyi explained what nested loops are, shared practical examples, highlighted common pitfalls and offered tips on how to optimise them.

https://realpython.com/nested-loops-python/

The two sides of hard work and commitment...

Which side are you currently on?

Hard work is a massive asset when channelled in the right direction...

But hard work can go against you if what you're putting the hard work into is unsatisfying and lacks purpose…

The path to success is often full of twists and turns, with things to learn around every corner…

If anyone's interested, the 2025 StackOverflow Developer Survey is now open: https://stackoverflow.blog/2025/05/29/not-just-a-vibe-the-stack-overflow-developer-survey-is-really-here/?utm_source=Mastodon #SoftwareDevelopment #surveys

Machine Code Isn't Scary

https://jimmyhmiller.com/machine-code-isnt-scary

Currently working on a little #rasterizer after watching the fantastic video from Sebastian Lague on yt.

Decided to write mine in #rust with some other techniques than the ones in the video, like using Barycentric Coordinates and calculating an AABB of an tri before iterating over the pixels. The video below showcases the rasterizer trying to process 250 random triangles. (Windowing and the FPS counter are done with #raylib)

Some artifacts here and there due to Z-ordering issues, but I hope to fix this soon.

End goal is to write it in such a way that I even might someday can use the rasterizer inside an custom rust #kernel / #os for basic 3d graphics, ofc to run
#doom on it!!

Why I Wrote the Beam Book

https://happihacking.com/blog/posts/2025/why_I_wrote_theBEAMBook/

OpenAI's Codex AI Coding Agent Gains Internet Access and Better Voice Tools

#AI #OpenAI #Codex #AI #DeveloperTools #VoiceAI #AISDK #SoftwareDevelopment #Programming #FutureOfCode

https://winbuzzer.com/2025/06/04/openais-codex-ai-coding-agent-gains-internet-access-and-better-voice-tools-xcxwbn/

#VisualStudio 2022 version 17.14 is here!

This update introduces agent mode for GitHub Copilot, enabling it to perform complex, multi-step tasks when editing code.

Dive deeper into what's new on #InfoQ https://bit.ly/43Gyxcw

Trying a new, slightly simplified recording setup for an experimental format based on my talk about #bonsai and #SoftwareDevelopment.
Not sure if you'll ever see one of the videos - as I've said, it's experimental, and I'm not sure if I can pull it off in a way that I like.

Are you struggling to retain new information?

Sleep plays a crucial role in transferring short-term memory to long-term memory, which is essential for learning new skills.

Make sure you get 7-9 hours of sleep to improve your memory and learning abilities.

Here are my top tips for better sleep?

 Earplugs
 A good quality mattress
 A well-fitting sleep mask
#developers #coding #softwaredevelopment #softwareengineering #wellbeing #mindset #mentalhealth

In this article, Vitalii Honchar explained how to build AI-powered apps that can chat with uploaded PDF files. He showed how to implement Retrieval Augmented Generation (RAG) using FastAPI for the API and LangChain to interact with OpenAI.

https://vitaliihonchar.com/insights/python-rag-api

У микрофона Анастасии Балакиревой, «АЛРОСА ИТ», рассказывала про «Форсайт. Мобильная платформа», происходило на https://teched2025.fsight.ru/ Как эта женщина говорит, с какими манерами и говором можно полюбоваться в коротком видео https://rutube.ru/video/bcfa5e8279d8d365c82737f8e7be8f61/

К самому по себе «Форсайт» это мало имеет отношения, который занимается той системой для BI-решений (бизнес аналитики), что звалось «Прогноз Платформа» до 2018 года.
Хотя, как ни сталкиваюсь с ними, так вечно лезет эта их специфика провинциальной компании из промышленного кластера Перми.
Вроде и офис маркетинга с продажами у них давно в Москве и управляются якобы из Москвы, но постоянно вылезает то одно, то другое. Даже судебные разборки о банкротстве «Прогноза» эти люди проиграли, и в результате стали называть свою платформу аналитики совсем иным образом, в то время как права на изначальное отошли к другому юрлицу.

Большая часть лиц мелькающих вокруг бренда «Форсайт. чего-либо» — это люди из САПРАН (SAPRUN). Интегратора давно погрязшего в различных стрёмных схема разворачивания и внедрения всего подряд, якобы специализируясь на SAP-решениях. А когда началась волна импортозамещения в РФ, годах в 2018-2019 менеджмент из САПРАНа перекинулся выступать от лица Форсайта. Именно «Форсайт. Мобильная платформа» целиком делалась в САПРАНе, а не бывшей части Прогноза и почти никакого отношения не имеет к BI-решению для бизнес-аналитики.

Оба юрлица входят в холдинг #ITG (INLINE Technologies Group), у которого полно разных активов, включая производство банковских карт пластиковых и POS-терминалов с решениями для эквайринга.

Держитесь от этой группы компаний подальше, нечего там делать, ни стару ни младу.

#Форсайт #САПРАН #SAPRUN #TechEd2025 #TechEd #SoftwareDevelopment #BI @russian_mastodon@mastodon.social @ru_ng@mastodon.social @rur@social.sley.nl @Russia@3zi.ru

My AI Skeptic Friends Are All Nuts
https://fly.io/blog/youre-all-nuts/
#softwaredevelopment #software #aiml #code