Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Promising results from DeepSeek R1 for code

simonwillison.net

Post reply on HN
Print view View original

161–170 of 765 posts

Re: Promising results from DeepSeek R1 for code

#161

Post by attractivechaos »

jeswin wrote: post #31

> 99% of the code in this PR [for llama.cpp] is written by DeekSeek-R1 I hope we can put to rest the argument that LLMs are only marginally useful in coding - which are often among the top comments on many threads. I suppose these arguments arise from (a) having used only GH copilot which is the worst tool, or (b) not having spent enough time with the tool/llm, or (c) apprehension. I've given up responding to these.…

I wonder what prompt they use. Before asking DeekSeek – is there a good post/video that walks through this procedure?

Re: Promising results from DeepSeek R1 for code

#162

Post by LeFantome »

AnotherGoodName wrote: post #120
onetimeusename wrote: post #107

Earlier quoted context omitted.

Just as a side note, at my university about half the CS people are in the AI track. I would guess that number will keep increasing. There is also a separate major that kind of focuses on AI/psychology that is pretty popular but I am not sure how many people are in it. A good number of the students have some kind of "AI startup". Also, although it violates the honor code, I would be willing to bet many students use AI…

The ask from every new grad to be assigned to ai development is unreasonable right now and they are probably hurting their careers by all going the same direction honestly. It’s a small fraction of our development efforts and we usually hire very senior for that sort of role. We still need people that can program for the day to day business needs and it’s a perfect starting role for a new grad yet almost all of them…

Agree and disagree. You do it need a “degree in AI”. However, you need to be using AI in your degree. Really using it.

What are those “day to day business needs” that you think people are going to do without AI?

In my view, this is like 1981. If you are saying, we will still need non-computer people for day-to-day business needs, you are wrong. Even the guy in the warehouse and the receptionist at the front are using computers. So is the CEO. That does not mean that everybody can build one, but just think of the number of jobs in a modern company that require decent Excel skills. It is not just the one in finance. We probably don’t know what the “Excel” of AI is just yet but we are all going to need to be great at it, regardless of who is building the next generation of tools.

Re: Promising results from DeepSeek R1 for code

#163

Post by deadbabe »

simonw wrote: post #52
jeswin wrote: post #31

> 99% of the code in this PR [for llama.cpp] is written by DeekSeek-R1 I hope we can put to rest the argument that LLMs are only marginally useful in coding - which are often among the top comments on many threads. I suppose these arguments arise from (a) having used only GH copilot which is the worst tool, or (b) not having spent enough time with the tool/llm, or (c) apprehension. I've given up responding to these.…

"Jobs are going to be lost unless there's somehow a demand for more applications." That's why I'm not worried. There is already SO MUCH more demand for code than we're able to keep up with. Show me a company that doesn't have a backlog a mile long where most of the internal conversations are about how to prioritize what to build next. I think LLM assistance makes programmers significantly more productive, which makes…

Too much productivity can be a bad thing.

If you’re infinitely productive, then the solution to every problem is to just keep producing stuff, instead of learning to say no.

This means a lot of companies will overbuild, and then drown in maintenance problems and fail catastrophically when they can’t keep up.

Re: Promising results from DeepSeek R1 for code

#164

Post by nwienert »

menaerus wrote: post #119
simonw wrote: post #99

Earlier quoted context omitted.

I continue to suspect that the hiring problems are mainly due to massive over-hiring during Covid, followed by layoffs that flooded the market with skilled developers looking for work. I'd love to see numbers around the "execs don't think they need engineers because of AI" factor. I've heard a few anecdotal examples of that but it's hard to tell if it's a real trend or just something that catches headlines.

Massive overhiring or not, it's the fact that many (skilled) engineers can't find a job. Many companies were shut off during the past few years and market became oversaturated over the night. Whether AI will help to correct the market creating more demand we will see but I wouldn't hold my breath. Many domain specific skills became a commodity.

Where's the fact coming from, as in it's higher than before? I seem to be getting more than ever recruiting emails, and have felt out interviewing at a few places which we're very eager to find staff level talent.

Re: Promising results from DeepSeek R1 for code

#165

Post by simonw »

magicalist wrote: post #160
simonw wrote: post #79

Earlier quoted context omitted.

"Yeah, but LLMs can't handle millions of lines of crufty old Java" is a guaranteed reply any time this topic comes up. (That's not to say it isn't a valid argument.) Short answer: LLMs are amazingly useful on large codebases, but they are useful in different ways. They aren't going to bang out a new feature perfectly first time, but in the right hands they can dramatically accelerate all sorts of important activities…

> "Yeah, but LLMs can't handle millions of lines of crufty old Java" is a guaranteed reply any time this topic comes up. That's not at all what the GP was saying, though: > There's no need to implement half the code, realize the requirements weren't specified properly, and go back and have a conversation with the PM about it. Which is, you know, what developers actually do.

I was responding to this bit:

> This is the kind of stuff LLMs are great at, but it's not representative of a typical change request by Java Developer #1753 at Fortune 500 Enterprise Company #271.

Re: Promising results from DeepSeek R1 for code

#166

Post by LeFantome »

bick_nyers wrote: post #121
jeswin wrote: post #86

Earlier quoted context omitted.

> That's why I'm not worried. There is already SO MUCH more demand for code than we're able to keep up with. Show me a company that doesn't have a backlog a mile long where most of the internal conversations are about how to prioritize what to build next. I worry about junior developers. It will be a while before vocational programming courses retool to teach this new way of writing code, and these are going to be te…

I definitely agree with you in the interim regarding junior developers. However, I do think we will eventually have the AI coding equivalent of CICD built into perhaps our IDE. Basically, when an AI generated some code to implement something, you chain out more AI queries to test it, modify it, check it for security vulnerabilities etc. Now, the first response some folks may have is, how can you trust that the AI is…

If it can point out the things you may need to consider, it is already better at security than most dev teams in the world today. Deep Seek can already do that.

Re: Promising results from DeepSeek R1 for code

#167

Post by simonw »

mjr00 wrote: post #100
simonw wrote: post #79

Earlier quoted context omitted.

"Yeah, but LLMs can't handle millions of lines of crufty old Java" is a guaranteed reply any time this topic comes up. (That's not to say it isn't a valid argument.) Short answer: LLMs are amazingly useful on large codebases, but they are useful in different ways. They aren't going to bang out a new feature perfectly first time, but in the right hands they can dramatically accelerate all sorts of important activities…

Yes, LLMs are very useful, when used properly. But the linked change request is not a good example of how they would be used by a typical software developer. The linked pull request is essentially output from a compiler that's been hardcoded. > Writing individual functions, classes and modules. You have to be good at software architecture and good at prompting to use them in this way - you take on the role of picking…

> If you have enough skill and understanding to do this, it means you already have enough general software development experience and domain-specific experience and experience with a specific, existing codebase to be in rarefied air.

Yes, exactly. That's why I keep saying that software developers shouldn't be afraid that they'll be out of a job because of LLMs.

Re: Promising results from DeepSeek R1 for code

#168

Post by sharperguy »

menaerus wrote: post #119
simonw wrote: post #99

Earlier quoted context omitted.

I continue to suspect that the hiring problems are mainly due to massive over-hiring during Covid, followed by layoffs that flooded the market with skilled developers looking for work. I'd love to see numbers around the "execs don't think they need engineers because of AI" factor. I've heard a few anecdotal examples of that but it's hard to tell if it's a real trend or just something that catches headlines.

Massive overhiring or not, it's the fact that many (skilled) engineers can't find a job. Many companies were shut off during the past few years and market became oversaturated over the night. Whether AI will help to correct the market creating more demand we will see but I wouldn't hold my breath. Many domain specific skills became a commodity.

We had a huge boom due to the low interest rates allowing businesses to pay developers with borrowed money, effectively operating at a loss for years on the basis of future growth. Now interest rates have risen the need to actually be profitable has caused a lot of optimization and lower hiring overall.

Re: Promising results from DeepSeek R1 for code

#169

Post by cchance »

simonw wrote: post #43
punkpeye wrote: post #39

I don't get something. So I tried hosting this model myself. But the amount of minimum GPU RAM needed is 400gb+ Which even with the cheapest GPU providers will be at least USD 15/hour How is everyone running these models?

Using the smaller distilled versions. I'm running this one, which only needs 20GB of VRAM (or regular RAM on Apple Silicon): https://ollama.com/library/deepseek-r1:32b

Keep in mind the distilled versions are NOT shrunken versions of deepseek-r1 their just finetunes of Qwen and Llama i believe, and they are no where near as good as real r1 (the 400g version) or even the 133g quants.

Re: Promising results from DeepSeek R1 for code

#170

Post by cchance »

tantalor wrote: post #25

> it can optimize its own code This is an overstatement. There are still humans in the loop to do the prompt, apply the patch, verify, write tests, and commit. We're not even at intern-level autonomy here.

I mean currently yes, but writing a test/patch/benchmark loop, maybe with a seperate AI that generates the requests to the coder agent loop, should be doable to have the AI continually attempt to improve itself, its just no ones built the loop yet to my knowledge
Post reply on HN