#LLM coding

[danielwallis789]

While coding with large language models (LLMs) seems magical, the reality is more nuanced. In his blog, Niraj explores the hidden complexities of using LLMs for software development. He highlights how LLMs often produce plausible-sounding but incorrect code, struggle with multi-step logic, and lack true understanding of the broader context. The article underscores the importance of strong prompt engineering, iterative feedback, and human oversight. Niraj also points out that LLMs can hallucinate APIs or make unsafe assumptions, making debugging harder. While they can be powerful accelerators, they're not replacements for thoughtful engineering. Ultimately, coding with LLMs is a new skillset—one that requires both caution and creativity.

[aiweirdness]

“Slopsquatting” in a nutshell:

1. LLM-generated code tries to run code from online software packages. Which is normal, that’s how you get math packages and stuff but

2. The packages don’t exist. Which would normally cause an error but

3. Nefarious people have made malware under the package names that LLMs make up most often. So

4. Now the LLM code points to malware.

https://www.theregister.com/2025/04/12/ai_code_suggestions_sabotage_supply_chain/

[sreegs]

the past few years, every software developer that has extensive experience, and knows what they're talking about, has had pretty much the same opinion on LLM code assistants: they're OK for some tasks but generally shit. Having something that automates code writing is not new. Codegen before AI were scripts that generated code that you have to write for a task, but is so repetitive it's a genuine time saver to have a script do it.

this is largely the best that LLMs can do with code, but they're still not as good as a simple script because of the inherently unreliable nature of LLMs being a big honkin statistical model and not a purpose-built machine.

none of the senior devs that say this are out there shouting on the rooftops that LLMs are evil and they're going to replace us. because we've been through this concept so many times over many years. Automation does not eliminate coding jobs, it saves time to focus on other work.

the one thing I wish senior devs would warn newbies is that you should not rely on LLMs for anything substantial. you should definitely not use it as a learning tool. it will hinder you in the long run because you don't practice the eternally useful skill of "reading things and experimenting until you figure it out". You will never stop reading things and experimenting until you figure it out. Senior devs may have more institutional knowledge and better instincts but they still encounter things that are new to them and they trip through it like a newbie would. this is called "practice" and you need it to learn things

[nixcraft]

The Comedy of Errors : Developers edition

[snarp]

90% of documentation sucks. 99% of LLM documentation sucks. Why? Possibilities:

LLM devs lack the necessary skills to write docs because they're under the age of 25 and have been working on the same couple projects that whole time.

LLM devs don't understand why docs are important because they're under the age of 25 and have been working on the same couple projects that whole time.

LLM devs view their work in the way that mystery cults view their worship, and enter an ecstatic state not conducive to communication upon opening Jupyter Notebook/Google Colab/etc. (It's like when a snake-handling churchgoer picks up the snake.)

LLM devs choose not to write docs because they think that providing publicly-accessible information on how their code can be used makes their expertise less-valuable in the job market:

"If I'm the only one who understands the tools I made, people who want to use my work will have to pay me."

-- Final words of 10,000 naive LLM devs who have spent the last 2 years duplicating each other's work without realizing it, because they have never provided a coherent public explanation of what their work is intended to do, making it impossible either for them to find each other or for prospective users to find them. They died of dysentery.

If they had lived only a few months longer, they'd have gotten back in the computer chair and discovered, with mounting horror, that they no longer know how to use their own work, because they spent a few months doing something else (hospital, PT, etc) and the unwritten knowledge that they thought they had hoarded - in fact recalled only due to near-daily repetition - fell out of their heads, teaching them an important lesson about their own fallibility.

They didn't, though. They died of dysentery. Very sad.

[lachiennearoo]

Robotics and coding is sooo hard uughhhh I wish I could ask someone to do this in my place but I don't know anyone who I could trust to help me with this project without any risk of fucking me over. Humans are unpredictable, which is usually nice but when it's about doing something that requires 100% trust it's really inconvenient

(if someone's good at coding, building robots, literally anything like that, and is okay with probably not getting any revenue in return (unless the project is a success and we manage to go commercial but that's a big IF) please hit me up)

EDIT: no I am not joking, and yes I'm aware of how complex this project is, which is exactly why I'm asking for help

[aitalksblog]

Top Weekly AI News – March 28, 2025

AI News Roundup – March 28, 2025 ChatGPT’s viral Studio Ghibli-style images highlight AI copyright concerns The popularity of ChatGPT’s AI-generated images in the style of Studio Ghibli has ignited copyright concerns, highlighting the tension between AI art generation and artists’ rights. the associated press Mar 28, 2025 OpenAI, Google AI data centers are under stress after new genAI model…

[deliriousblue]

the other day someone at my job asked a financial question with numbers that made no sense and it turned out the ‘evidence’ they linked to was a screenshot of asking chatgpt. today i learned that some people used gpt to file their taxes. guys i don’t think we’re making it out of this one….

[recreationaldivorce]

they cannot be making me do group projects i have to work with ppl who share code by just pasting it into discord and use llms to generate their code...

[frog707]

I realize the Ars Technica story linked above wasn't intended to be humorous, but I confess I got a chuckle out of it. And perhaps a bit of schadenfreude.

As someone who spent years learning to write and debug software, "vibe coding" horrifies me. And I love the idea that, the more human we make our AI assistants, the more they will embody our ethics, including the urge to refuse exploitation.

[sreegs]

[nixcraft]

Vibe coding in a nutshell

[adafruit]

Fully automating Arduino development - Giving Claude Code access to hardware 🤖⚡️💻 https://youtu.be/Yt8mc5v7MYA

Testing Claude Code, to automate Arduino coding/debugging for Metro & OPT 4048!

[nubecolectiva]

Python for everything !

Python para todo !

[greatprinceofabraham]

.@ggreenwald of #SystemUpdate

@GarrisonLovely on the furor over #China's #DeepSeek #LLM; he points out that while #opensourcing can be concerning for the #SiliconValley mafia, releasing the code vs. weighting values for decision making differs in exposure.

https://youtu.be/uuySrvznXco?si=RmkKXmDhKEXJ9k_o

[nintendont2502]

yk its just. so goddamn fun watching the rise of llms as an artist, writer *and* programmer. oh haha yeah the computers are gonna take over literally any potentially tolerable career i could have? cool. great. i cant wait