#coding-assistant — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #coding-assistant, aggregated by home.social.
-
RT @Ali_TongyiLab: Wir freuen uns, ein hervorragendes Community-Modell des Entwicklers Qwen3.6-27B-MTP-pi-reasoning-GGUF vorzustellen.
mehr auf Arint.info
#CodingAssistant #CommunityModel #DeveloperTools #LocalAI #OpenSourceAI #Qwen3 #arint_info
-
Before working for 2 years on the Apache APISIX API gateway, I was mainly oblivious to API gateways. It’s only by working with them that I understood their value. Decoupling the client and the server unlocks a lot of options: moving authentication to the API Gateway, securing APIs, deduplicating API requests, etc.
In this post, I want to describe how the same pattern applies to #AI.
#AIGateways work in a similar way.
-
Before working for 2 years on the Apache APISIX API gateway, I was mainly oblivious to API gateways. It’s only by working with them that I understood their value. Decoupling the client and the server unlocks a lot of options: moving authentication to the API Gateway, securing APIs, deduplicating API requests, etc.
In this post, I want to describe how the same pattern applies to #AI.
#AIGateways work in a similar way.
-
Reports of #code's death are greatly exaggerated
-
Reports of #code's death are greatly exaggerated
-
In the world of AI and LLM every mistake, error or bug you run into is already a »classic«.
»You’re running into a classic ›columns depend on async data‹ vs ›columns must be stable for grid state‹ conflict.«
Okay, I feel less alone now, but this still doesn't help me solve the problem, haha.
-
Question for #LLM haters:So I don’t really want to hate on LLMs or generative AI in this post, but I am curious about alternatives to LLMs for code generation. What I particularly have in mind is how LLMs are pretty good at calling up code examples and fitting them into your code base.
Has anyone tried just creating a very large database with millions of code examples or code snippets, perhaps tagged with keywords pertaining to what the code does, and letting people search code by tag and automatically paste it into your file? If StackOverflow has been mined for LLM training data, can’t we just take that StackOverflow training data and parse-out the code snippets and generate such a database?
If LLMs coding tools like #Cursor, #Claude, #Copilot, Cody, etc. are basically doing that using statistical algorithms, couldn’t we do the same thing with ordinary symbolic computation and a clever little editor plug-in with good UI/UX design for the copy-pasting?
Does anyone know if such tools/databases already exist?
I’ll tag @screwlisp and @kentpitman on this one because I am especially curious about what you think about it.
#tech #software #AskFedi #CodingAssistant #LLM #AI #GenerativeAI
-
@regehr
🙏🏻 So true, so well said, so best practice: helpers should “never, ever touch the keyboard when they're helping a student with an assignment. not even once! because as soon as someone else is driving, it becomes real easy for the student to stop thinking and just let things happen.”AND, the corollary or by extension:
“kind of like what happens when we use a coding assistant.”
-
People share the post by a senior #Google engineer that #GenAI produced a distributed system orchestrator as if that was surprising, or a sign of actual capabilities.
The #LLM training data includes many, many very well documented code bases for this type of project. Lots of user experiences. User and API documentation. Reviews. PRs. Patches. ADRs, KEPs, RFCs. Discussion forums, thoughts about alternatives, comparisons. The list goes on.
(1/3)
-
When you're #CodingAssistant responds to the author of a file with "Oof. That's... concerning." It's not good.
It felt so wrong, but came from a project lead so I ran it past the #JetBrains assistant to make sure I wasn't missing something obvious. I hope this was written by #AI
-
Wow ... what a surprise. Or wait, no, it isn't 🤔😑
#cybersecurity #infosec #vibecoding #LLM #AI #malware #amazon #coding #codingassistant #cybersec #devops #programming #developers
-
I have this feeling lately that coding assistants are going to be the thing that kills free open source software.
It seems inevitable that competing coding assistants will end up creating their own libraries (or languages) to lock users in.
Not only will their models be better trained on their own stack, but competing platforms won’t be trained at all (or very little).
And since these platforms don’t give anything back to the community when a user asks a question, the incentive to share will keep dropping.
Open source thrives on being open. This fits in with free platforms like Stack Overflow.
But what happens when (any) new libraries are written or new languages are created and all the questions asked are of a proprietary model that considers the questions as well as the answers a trade secret?