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

371–380 of 765 posts

Re: Promising results from DeepSeek R1 for code

#371

Post by __MatrixMan__ »

sitkack wrote: post #78
simonw wrote: post #52

Earlier quoted context omitted.

"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…

We have already entered a new paradigm of software development, where small teams build software for themselves to solve their own problems rather than making software to sell to people. I think selling software will get harder in the future unless it comes with special affordances.

...which is a good thing. Software made by the people using it to better meet their specific needs is typically far better than software made to be a product, which also has to meet a bunch of extra requirements that the user doesn't care about.

Re: Promising results from DeepSeek R1 for code

#372

Post by nyarlathotep_ »

anotherpaulg wrote: post #113
joshstrange wrote: post #106

Earlier quoted context omitted.

First off I want to thank you for Aider. I’ve had so much fun playing with it and using it for real work. It’s an amazing tool. How do you determine how much was written by you vs the LLM? I assume it consists of parsing the git log and getting LoC from that or similar? If the scripts are public could you point me at them? I’d love to run it on a recent project I did using aider.

Glad to hear you’re finding aider useful! There’s a faq entry about how these stats are computed [0]. Basically using git blame, since aider is tightly integrated with git. The faq links to the script that computes the stats. It’s not designed to be used on any repo, but you (or aider) could adapt it. You’re not the first to ask for these stats about your own repo, so I may generalize it at some point. [0] https://ai…

does this mean lines/diffs otherwise untouched are considered written by Aider?

If a small change is made by an end-user to adjust an Aider result, who gets "credit"?

Re: Promising results from DeepSeek R1 for code

#373

Post by dzhiurgis »

7thpower wrote: post #346
SkyPuncher wrote: post #306

Earlier quoted context omitted.

The problem is it's very expensive. More expensive than Claude.

You can use the distilled version on Groq for free for the time being. Groq is amazing but frequently has capacity issues or other random bugs. Perhaps you could set up Groq as your primary and then fail back to fireworks, etc by using litellm or another proxy.

Do you know any assistants for jetbrains that can plug into groq+deepseek?

Re: Promising results from DeepSeek R1 for code

#374

Post by nine_zeros »

pertymcpert wrote: post #321
nine_zeros wrote: post #98

Earlier quoted context omitted.

Alright, if this is more palatable - let's just say market caps will decline because of small code updates made anywhere in the world. The point still is: Software/engineering is no longer the moat creator.

You completely misunderstood the reason for the stock price drop. It was because of the DeepSeek MoE model's compute efficiency which vastly reduced the compute requirements needed to achieve a certain level of performance. Notice how Apple and Meta stocks went up last 2 days?

You are misunderstanding my point. It is because anyone with a non-software moat will likely be able to leverage the benefits of AI.

Apple has a non-software moat: Their devices.

Meta has a non-software moat: their sticky users.

So does Microsoft, and Google to an extent with their non-software moat.

But how did they build the most in the first place? With software that only they could develop, at a pace that only they could execute, all because of the people they could hoard.

The companies of the future can disrupt all of them (maybe not apple) very quickly by just developing the same things as say Meta and "at the same quality" but for cheaper. The engineers moat is gone. The only moat meta has is network effects. That's one less barrier for a competing company to deal with.

Re: Promising results from DeepSeek R1 for code

#375

Post by skyyler »

lgats wrote: post #94
mohsen1 wrote: post #33

Earlier quoted context omitted.

Yes! I've done something like this here in my repo. This was nice while lasted (Deepseek is practically useless through the API since yesterday) https://github.com/bodo-run/yek/blob/main/.github/workflows/... https://github.com/bodo-run/yek/blob/main/scripts/ai-loop.sh Using askds https://github.com/bodo-run/askds

added context, deepseek is having ddos issues https://status.deepseek.com/

If you can't beat em on merit, pay criminals to take em offline.

Good business.

Re: Promising results from DeepSeek R1 for code

#376

Post by codingdave »

mike31fr wrote: post #362
simonw wrote: post #3

Given these initial results, I'm now experimenting with running DeepSeek-R1-Distill-Qwen-32B for some coding tasks on my laptop via Ollama - their version of that needs about 20GB of RAM on my M2. https://www.ollama.com/library/deepseek-r1:32b It's impressive! I'm finding myself running it against a few hundred lines of code mainly to read its chain of thought - it's good for things like refactoring where it will thi…

Noob question (I only learned how to use ollama a few days ago): what is the easiest way to run this DeepSeek-R1-Distill-Qwen-32B model that is not listed on ollama (or any other non-listed model) on my computer ?

If you are specifically running it for coding, I'm satisfied with using it via continue.dev in VS Code. You can download a bunch of models with ollama, configure them into continue, and then there is a drop-down to switch models. I find myself swapping to smaller models for syntax reminders, and larger models for beefier questions.

I only use it for chatting about the code - while this setup also lets the AI edit your code, I don't find the code good enough to risk it. I get more value from reading the thought process, evaluating it, and the cherry picking which bits of its code I really want.

In any case, if that sounds like the experience you want and you already run ollama, you would just need to install the continue.dev VS Code extension, and then go to its settings to configure which models you want in the drop-down.

Re: Promising results from DeepSeek R1 for code

#377

Post by kridsdale3 »

spease wrote: post #278
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.…

The nature of this PR looks like it’s very LLM-friendly - it’s essentially translating existing code into SIMD. LLMs seem to do well at any kind of mapping / translating task, but they seem to have a harder time when you give them either a broader or less deterministic task, or when they don’t have the knowledge to complete the task and start hallucinating. It’s not a great metric to benchmark their ability to write…

Sure, but let's still appreciate how awesome it is that this very difficult (for a human) PR is now essentially self-serve.

How much hardware efficiency have we left on the the table all these years because people don't like to think about optimal use of cache lines, array alignment, SIMD, etc. I bet we could double or triple the speeds of all our computers.

Re: Promising results from DeepSeek R1 for code

#378

Post by __MatrixMan__ »

aibot923 wrote: post #178
simonw wrote: post #52

Earlier quoted context omitted.

"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…

It's interesting. Maybe I'm in the bigtech bubble, but to me it looks like there isn't enough work for everyone already. Good projects are few and far between. Most of our effort is keeping the lights on for the stuff built over the last 15-20 years. We're really out of big product ideas.

That's because software is hard to make, and most projects don't make it far enough to prove themselves useful--despite them having the potential to be useful. If software gets easier, a whole new cohort of projects will start surviving past their larval stage.

These might not be big products, but who wants big products anyway? You always have to bend over backwards to trick them into doing what you want. You should see the crazy stuff my partner does to make google docs fit her use case...

Let's have an era of small products made by people who are close to the problems being solved.

Re: Promising results from DeepSeek R1 for code

#379

Post by cynicalpeace »

woah wrote: post #89
cynicalpeace wrote: post #76

Earlier quoted context omitted.

It's making programming more boring and more of an admin task- which is sure to attract different types of people to the field.

Seems like the exact opposite. The very example you are replying to is the mechanistic translation of one low level language to another, maybe one of the most boring tasks imaginable.

It goes both ways- manually manipulating bits is boring. But so is reviewing and rubber stamping code.

It's like the transition from hand-crafted furniture to assembly line mass produced furniture.

The assembly line brought its own excitement, but that excitement was not to be found on the actual assembly line.

Re: Promising results from DeepSeek R1 for code

#380

Post by root_axis »

danielbln wrote: post #361
root_axis wrote: post #345

Earlier quoted context omitted.

Your reasoning still assumes that "AGI" can emerge from quadratic time brute force on some text and images scraped off the internet. Personally, I'm skeptical of that premise.

That's like saying sentience cannot emerge from a few amino acids tumbled together, yet here we are. There is a lot of higher dimensional information encoded in those "text and images scraped off the internet". I still don't think that's enough for AGI (or ASI) but we know a lot of very complex things that are made of simple parts.

> That's like saying sentience cannot emerge from a few amino acids

No, it's not at all the same thing.

We have great evidence that life exists. We have great evidence that amino acids can lead to life.

None of that is true of "AGI" or text scraped off the internet.

Post reply on HN