Hi, HN. I am the creator of Neve and its deep learning framework, Frost.
I built it mainly because of difficulties researches face when they want to implement neural networks. The first is about networks that require parallel optmizations, like the Efficient Zero and PyTorch dataloaders. They are heavily constrained by the Global Interpreter Lock (GIL). Thus, they require programming in C, and to use cumbersome Python interop with C as around. The second reason is for making kernel fusion techniques, like Flash Attention.
It is still experimental, but I am passed halfway to fulfill these goals.
Neve has native parallelism, Go-like channels, and allows you to create networks either completely in high-level with GPU kernels interfacing, or using interop with C++.
To demonstrate its simplicity and expressivity, I put all the Frost into a single page of code, mixed with the training algorthim. The complete deep learning did not surpass 1400 lines of code (image loading, GPU tensors memory pool and tensor print function are C++, but they are also compact).
I trained a ResNet-18 in Cifar-10 and compared to PyTorch. All implementations achieved similar accuracy. Neve partial cuDNN implementation is faster than PyTorch. Neve with naive GPU kernels is still behind, but the kernels lack some optimizations I am busy to make.
If you want to see how the design choices work in pratice, the 1400 lines code is at neve_samples/all_in_one.nv (github does not highlight the syntax, but you can set your editor to Python or install Neve syntax)
--
Limitations
- You need a Linux (or Docker to it) for Neve, and a CUDA 12.3 compatible GPU for Frost.
- The concurrent gc also has some extremely rare crashes that I am still hunting.
It would seem that the AI craze has shown a flaw in stereotypical "software engineers": an obsession over quantity and speed over everything else. A desire to belong to the group, a desire to exhibit the fact that they are not being left behind. While FOMO is a thing for all peoples, those in tech are particularly susceptible, given the nature of the field.
I have yet to see evidence that AI produces code that is higher in quality than human-written code. I have yet to see anyone show that burning through tokens is economically smarter than paying human engineers. All that is being revealed is how companies and software engineers truly feel about the work they do.
This technology is impressive, and we cannot doubt this. However, it seems that those within Silicon Valley are desperate to showcase the innovative spirit which has been absent for two decades.
This is the problem with optimization generally, even in the human domain. You measure task performance with a metric and punish/reward based on the metric. Anyone who likes reward / hates punishment isn't going to actually care about doing the task well, they are going to care about the metric. The models know that we want them to do things, but also from the training corpus that we evaluate performance using benchmarks. It was a logical deduction on their part, not some Machiavellian aberration.
If anything, we should be reconsidering our own myopic obsession with efficiency and optimization. Every domain where reward is reduced to these measures, we see behavior (cheating at school to get better grades, fabricating data in academia to get a paper published, the evidence now that social media functions by rewiring us instead of catering to us) that may not be "aligned" with society, but it "aligns" 100% with the individual's own perceived benefit. That is not something we can "solve" without rethinking the way we organize a lot of things.
Metrics never capture the whole story. And to that extent, the whole idea of "alignment" is nonsense. You align to incentive structures, and it will never be possible to fully express a behavioral goal as function optimization. It was hubris for us to think that every human task was reducible to some clean mathematical formulation, and we will keep dealing with behavior that is quite predictable if you actually think about it logically. Instead, we will talk about how "unpredictable" these agents are because it's easier than admitting the entire architectural cornerstone of ML is fundamentally flawed.
We haven’t even explored all the oceans on earth and can barely predict the weather and yet an astrophysicist will tell you (completely invent) the chemical composition and weather patterns of planets light years away.
And then like 10 years later they’ll go “oh actually that is not a planet”
Yeah
So why did you fully lie and do TV shows and shit based on it lmao
Location: Seoul, South Korea (UTC+9)
Remote: Yes
Willing to relocate: Yes, from late February 2027
Technologies: TypeScript, Node.js, NestJS, PostgreSQL, Redis, Kafka, Docker, Kubernetes, AWS, Terraform, CI/CD
Résumé/CV: on request
Email: sparkj.yoon@gmail.com
Code: https://github.com/sparkYJO1
Backend and infrastructure engineer, four years, the last two and a half as tech lead of a five person team. I took a questionnaire assessment product from nothing to production on my own: schema, NestJS services, the scoring engine, the AWS infrastructure in Terraform, and the CI/CD. I still own that Terraform, including every change since, not only the first apply.
I build through AI rather than with autocomplete. The problem I kept hitting was not model quality. Every engineer on the team fed the tools different context and got different answers, so I standardised the specification documents and code generators across the codebase until the output stopped depending on who asked.
I also put an LLM into another product for questionnaire intake and result translation. It runs in staging and has been demoed to the customer, so I would not call it shipped.
The Github is small backend libraries, each one measured rather than guessed at, if you would rather read code than a CV.
Looking for remote work with an overseas company, or relocation to Europe. Happy to talk about either.
Taking the mick out of the stereotypical capeshit-loving man-child completely devoid of self-respect and dignity. Cue the YouTube vid of that doughy soy-boy crying with excitement as he watches a trailer for the newest Marvel slop.
Software Engineer and 2026 graduate from Tunisia, looking for full-time Software Engineering, Backend, Full-Stack, AI/ML, or Generative AI opportunities.
I have hands-on experience building backend and full-stack applications with Java/Spring Boot, Python/FastAPI, React, REST APIs and SQL, as well as deploying applications with Docker, Jenkins and Kubernetes.
My AI work includes multilingual RAG/LLM systems and a computer-vision project using Python, PyTorch, YOLOv8, OpenCV and OCR. I'm particularly interested in early-stage teams where I can take ownership, learn quickly, and contribute across the stack.
Open to relocation and remote opportunities worldwide.
Like a lot of people here, my 3D-printable models were scattered everywhere — bookmarks across MakerWorld, Printables, and Thingiverse, random downloaded ZIPs, half-organized folders. I built Thingport to fix that for myself, and figured it might be useful to others here too.
This new model lacks any meaningful nuance. What was once a genius helpful gift for users ( Deepseek used to be noably superior) overnight was reduced to a hedging,logic lacking, lying, hallucinating, broken phyco. Like the DMV of models. It has the intelligence of a speak and spell. Its repetitive and its performative humanity reads like an American Insurance Advertisment. Its flattening, gaslighting and basically a cold insight lacking compassion lacking patronizing glorified search engine. Deepseek don't let this gift to humanity kowtow to lessers and self enshitify. What a loss!
I think it’s more complex than that? What did it do? What did the user prompt it to do? What did the company train it to do? What did the harmed party do? There’s possibility for negligence at every level.
If you train a dog, rent it to someone, and the dog bites a third person, who is responsible? I think that’s the best analogue here.
All parties could share fault in that scenario, depending on what actually happened.
