#ai in software engineering

[habilelabs]

Let's check how artificial intelligence impacts the tech industry and where we can see its implementation in the current and future states of software engineering.

[nixcraft]

Vibe coding in a nutshell

[paradoxcase]

The original history of the Mechanical Turk, which was a supposed mechanical man who could play chess with you from the Middle Ages which actually turned out to be a human chess player in a mechanical man suit, has officially repeated itself and we now have Mechanical Turk, 2025 edition (don't talk to me about Amazon Mechanical Turk, that's a whole other post for a whole other day):

This company claimed they had some AI tool that could build apps for you, but apparently what they actually had was 700 software engineers who would build apps for you. Microsoft invested almost a half a billion dollars into this company, and then when they found about this, they pulled out and now the company is declaring bankruptcy.

This is being portrayed as some kind of fraud, and it kind of is, in that they were basically profiteering off of AI hype, but like, being able to hire 700 software engineers to build an app for you is not nothing. It's not worthless. They probably make way better apps than any AI process could make.

The main thing I am taking away from this is that Microsoft is willing to spend half a billion dollars on an AI that builds apps, but is not willing to spend half a billion dollars on 700 human beings that build apps much better than the AI could. Doesn't that just tell you everything about the software industry right now?

[leng-m]

I found out today that my workplace is now asking interview candidates how they feel about AI. My work has become an absolute staunch supporter of generative AI lately and is pressuring everyone to incorporate it into their daily work. Chatting with a few workmates, I am baffled that I'm the only one who seems to have any reservations about this. (I talked with my manager today and he wasn't even aware of all the ethical issues with it! Shocker!)

I want to clarify that I'm not categorically against all AI. I majored in compsci and I did my masters on machine learning. But the things I wanted to build with this skill set are things like anomaly detection and pattern recognition. I get excited about tools like spam filters and accurate search engines. There's a place for generative AI, like with GANs, but I believe that the way it's being wielded now by the industry is just soul-sucking.

My company has heavily hinted that there's no future for anyone there who isn't willing to jump on the genAI bandwagon; rather shocking as it was up until that point, the most employee-friendly company I've worked at so far. I'd hate to lose the comprehensive health benefits it provides at a time I'm dealing with a chronic illness. And even if I did leave, the darlings of the Toronto tech scene had made it clear genAI is their future -- everyone else is going to be copying them. I don't think there's any job that's safe from the AI infiltration.

So I want to ask you guys. If you're at a company that's now forcing you to use genAI, how are you using it in a way that still lets you sleep at night?

[ottopilot-sfw]

Vibe coding something. This shit is supposed to take my job? All y'all are fucked. We are all fucked.

[nando161mando]

Software engineer lost his $150K-a-year job to AI—he’s been rejected from 800 jobs and forced to DoorDash and live in a trailer to make ends meet

https://www.yahoo.com/news/software-engineer-lost-150k-job-090000839.html

[smak-annihilation]

yeah sorry guys but the machine escaped containment and is no longer in my control or control of any human. yeah if it does anything mortifying it's on me guys, sorry

[querade]

things i experienced with my delightfully nerdy older brother during the holidays

Older brother: So, what happens when they run out of dinosaur to make Dino nuggets? Dad: They’ll go extinct. Him: Hang on - do we have enough DNA for gene sequencing dinosaurs? Could we actually make Dino meat in a lab? Dad: Course we can! Haven’t you ever seen Jurassic Park? ___________________________ Him, a software engineer, getting off a three hour meeting about developing an AI program, and then spending five minutes trying to toast a bagel in a toaster oven bc the picture at the bottom explaining how to do it was too complicated ___________________________ Me: what do you know about the Higgs-Boson particle? Older Brother: Well, I know what a boson is but every time I’ve actually studied it I’ve felt like a boson myself. ____________________________ Dad: Get in, the car has butt warmers! Him: OOOH, will they make my butt look hot?

_____________________________

reblog to make him happy

[antifataylorswift]

The developers i work with are asking chatgpt questions about my infrastructure that are in our docs. Unsurprisingly they are getting bad advice for our particular setup, and are bothering my team with a bunch of questions about if the advice is right and what would happen if they ran those commands.

[learn-ai-free]

OpenAI Releases Codex: A Software Agent that Operates in the Cloud and Can Do Many Tasks in Parallel

OpenAI has released a research preview of Codex, a cloud-based software engineering agent that's not just another code completion tool. Codex is a cloud-based software-engineering agent that turns on isolated sandboxes, pulls your repo, and chips away at features, bug fixes, test suites, and even pull-request boilerplates—often in parallel.

What is OpenAI Codex? 📌

→ Cloud-based software engineering agent

→ Can write features, answer codebase questions, run tests, and propose Pull Requests for review

→ Each task runs in its own isolated cloud environment

→ Provides detailed terminal logs, test outputs, and citations

→ Users can create AGENTS.MD files in their repository to instruct Codex on project-specific commands, testing procedures, and coding standards

→ Powered by codex-1

How to use Codex: 📌

→ Users can access Codex through the ChatGPT sidebar

→ Assign coding tasks by typing a prompt

→ Each request is handled independently

→ Codex can read and edit files and run commands like test suites, linters, and type checkers

→ Task completion generally takes between one and thirty minutes

Once done, Codex runs its changes within its sandboxed environment, which users can then review, ask for more changes, open a GitHub PR, or pull the changes into their local setup.

↗️ Full Read: https://aiagent.marktechpost.com/post/openai-releases-codex-a-software-agent-that-operates-in-the-cloud-and-can-do-many-tasks-in-parallel

Codex: Availability 📌

Codex is currently rolling out to ChatGPT Pro, Enterprise, and Team users, with access for Plus and Edu users planned to come soon.

[munaeem]

The Future of Software Engineering: What 2025 Holds

The landscape of software engineering is rapidly evolving. It is shaping up to be as exciting as it is transformative. Gone are the days when coding was solely the domain of dedicated programmers tucked away in dimly lit offices. Today, we’re entering a new era. Artificial intelligence is changing how we approach software development. Real-time computing is also a key factor. Additionally, the…

[nixcraft]

Imagine being this stupid to drink Kool-Aid and giving a remote LLM tool full access to your codebase, and, in many cases, not maintaining backups or using proper Git with permissions. How these guys are getting hired to write code is beyond me.

[mentalisttraceur-software]

I predict that an easy big across-the-board LLM reasoning improvement is to prompt the LLM to argue against every point it makes.

Ideally you have it then argue against those arguments, and so on. Have some stopping condition like "until you cannot think of counter arguments that are better than the previous arguments." But I think even just one iteration of arguing against itself will be an improvement on average.

With existing models, you probably need to do this as two messages, so that the model first gives you the best reasoning it can think of, instead of the easier path of predicting more easily refutable arguments to refute, since those are almost certainly going to be more common in the training data.

[paradoxcase]

Try not to sound like a sci-fi supervillain in your job listing challenge: impossible

[dissolution-digital]

TBH we should just build an AI platform that tells everyone above engineering manager level that everything is chill so we can do our fuckin jobs in peace