Palmik

21 ms·

181.
by Palmik 1y ago
I really don't see the big deal. Gemini also allows on-prem in similar fashion: https://cloud.google.com/blog/products/ai-machine-learning/r...
182.
by Palmik 1y ago
No issues on Linux for me. Much better (faster / snappier) than VS Code and forks.
183.
by Palmik 1y ago
I have been using Zed as my main editor for the past ~5 months and I have been very happy with it. It's actually fast and snappy. I hope they become sustainable. VS Code forks (Cursor and Windsurf) were extremely slow and buggy for me
184.
by Palmik 1y ago
I wonder why the "no-out-of-country logins" block happens after verifying login credentials and not before, which would make more sense to me.
185.
by Palmik 2y ago
There's also Google Vids, also using Veo 2 under the hood. Product confusion :) https://workspace.google.com/products/vids/
186.
The path to open-sourcing the DeepSeek inference engine(github.com)
550 points by Palmik 2y ago|63 comments
187.
by Palmik 2y ago
I did not see this paper on the front page before. It seems like a genuine contribution. And do you think it's better or worse than this: https://hn.algolia.com/?dateRange=pastMonth&page=0&prefix=tr... -- posti
188.
by Palmik 2y ago
Perfect time to accelerate de-dollarization!
189.
by Palmik 2y ago
Exponential growth generally means that the time between each order of magnitude is roughly the same.
190.
by Palmik 2y ago
Except VAT and Sales Tax is typically applied regardless of where the item / service originated from. It also doesn't apply potentially several layers down. That's not to mention things like reverse charge for B2B etc. In oth
191.
by Palmik 2y ago
In your example you already indicated two tasks that you think might be hard for AI but easy for humans. Who said that cooking dinner couldn't be part of ARC-AGI-<N>?
192.
by Palmik 2y ago
What I imagine 1. Use LLM, possibly already grounded by typical RAG results, to generate initial answer. 2. Extract factual claims / statements from (1). E.g. using some LLM. 3. Verify each fact from (2). E.g using separate RAG system
193.
The AI Scientist Generates Its First Peer-Reviewed Scientific Publication(sakana.ai)
2 points by Palmik 2y ago|0 comments
194.
by Palmik 2y ago
In my opinion, if any copyright related rule is applied to books or other paywalled content, it should equally apply any Joe Shmoe's blog or code on GitHub.
195.
by Palmik 2y ago
Incorrect. Llama 1 trained on books3 dataset.
196.
by Palmik 2y ago
A real person wouldn't have to pay to read random blog, Reddit comments, StackOverflow answers or code on GitHub (many open source licenses do not imply license for training). They might have to pay for books, or use a library. Should
197.
by Palmik 2y ago
What I think the parent meant is the interest to sell license to others to train on their data.
198.
by Palmik 2y ago
Funnily, OpenAI also calls for the ban of their free and open-weight Chinese competitors DeepSeek and Qwen.
199.
by Palmik 2y ago
Yes, I know Exa and have used them in the past for some side projects. Great product. But that's still an index. Google also internally uses combination of traditional and sematic/embedding based indices.
200.
by Palmik 2y ago
You realize that an index is still necessary for RAG? Where do you think the results used by Perplexity, Open AI etc. come from?
201.
by Palmik 2y ago
Especially since Grok has history of changing the system prompt to influence its answers about Musk and Trump (yes, just these two specifically) in positive direction: https://techcrunch.com/2025/02/23/grok-3-
202.
by Palmik 2y ago
Why don't you allow OAuth / social login via Google, GitHub and yes, even Discord? Seems like it would make adoption much easier.
203.
by Palmik 2y ago
I don't think (2) would really work out in practice. I have hard time believing that France would use nuclear weapons if one of the smaller states gets attacked. At the end of the day, when SHTF, you're on your own and Putin or so
204.
by Palmik 2y ago
Why these specifically? Why Poland and not Slovakia, Romania or Moldova that also neighbor Ukraine? These countries have much weaker conventional defenses than Poland, further increasing the importance of something like nuclear weapons. I w
205.
by Palmik 2y ago
The example from the landing page does not exactly spark joy: testWorkflow .step(llm) .then(decider) .then(agentOne) .then(workflow) .after(decider) .then(agentTwo) .then(workflow) .c
206.
by Palmik 2y ago
The negative quality impact of quantization is more pronounced for smaller models [0], so I'm surprised this tiny quant works at all. [0] or rather models closer to saturation, which is a function of model params and amount of training
207.
by Palmik 2y ago
I think this is a very good initiative actually, especially for small businesses. Building these tools in-house is not easy, and frankly quite stressful. And while there are commercial solutions, they have questionable privacy profile and a
208.
by Palmik 2y ago
Can you expand on the argument? You might as well have invested the 2021 "value" (regardless of in which currency it was disbursed) into some non-inflationary asset (like bitcoin, gold or some index fund) -- that should make you f
209.
by Palmik 2y ago
I am referring to this one: https://huggingface.co/deepseek-ai/DeepSeek-V3/blob/main/LIC... It is a bit more permissive than Llama's it seems (no MAU threshold it seems).
210.
by Palmik 2y ago
Also open-weights comes in several flavors -- there is "restricted" open-weights like Mistral's research license that prohibits most use cases (most importantly, commercial applications), then there are licenses like Llama&#x

More ›