ariesta.id - Profile | ThreadSky | a Reddit-style client for Bluesky

Thank you DeepSeek. "Worry time", a new term for me.

submitted 46 days ago • 0 comments

Most people don't set reasonable expectations. I usually forgot to set expectations too, after not interacting with LLM for a while

submitted 46 days ago • 0 comments

It seems that OpenAI Deep Research is so useful because of its search capability. I'm not really impressed with OpenAI's high-end models outside Deep Research. Imagine if they are truly open and people can connect the search indexing with their own source priority list

submitted 75 days ago • 1 comment

My current LLM choice: - large context, image, video: Gemini via AI Studio - code: Claude (web) - search: DeepSeek web, Perplexity - writing: Llama (GPU rent -- just because I have unused credits)

submitted 90 days ago • 0 comments

Anthropic co-founder Jack Clark muses on whether AI systems will give an unfair advantage to people with a fiercely curious nature importai.substack.com/p/import-ai-...

submitted 91 days ago • 15 comments

Yeah, if you don't like "AI", at least criticize it productively Outrightly dismissing it is not productive

submitted 94 days ago • 0 comments

DeepSeek just released R1, an open weight "thinking" model like OpenAI's O1. The difference is: - people said O1's better - O1 is closed, R1 is open weight - R1 is MIT-licensed, can use it to train/finetune other models - cannot see O1 thoughts, can see R1's api-docs.deepseek.com/news/news250...

submitted 97 days ago • 0 comments

Come on Bluesky... If I don't keep the other-side account, I wouldn't find out about DeepSeek R1 open weight release. Even my LinkedIn timeline is better.

submitted 97 days ago • 0 comments

I decided to dedicate more time to my master's research and resigned from a full-time job I hoped to get some time off from reading emails, but then I made the mistake of turning on Google Scholar alerts for the researchers I follow

submitted 99 days ago • 0 comments

Does ChatGPT use 10x more energy than a standard Google search? https://engineeringprompts.substack.com/p/does-chatgpt-use-10x-more-energy #AI #climate

submitted 101 days ago • 0 comments

Markdown Is All You Need

submitted 110 days ago • 0 comments

Blessing in disguise from the WordPress drama. Maintaining a blog with MkDocs, Docker, and Coolify is a lot simpler. Learning Docker is a big hurdle though. New post: ariesta.id/blog/2025/01...

submitted 110 days ago • 0 comments

Just learnt that there are two types of internal link: relative path and absolute path. This can mess up my blog ariesta.id/blog/2025/01...

submitted 112 days ago • 0 comments

An attempt to maintain a link blog. First post: ariesta.id/blog/2025/01... inspired by simonwillison.net/2024/Dec/22/...

submitted 115 days ago • 1 comment

Feels good deploying this with Coolify, Docker, Flask, and Google Login. I'm tempted by the one-click deployment on Coolify that I went through docker configuration hell. Luckily there are Claude, Gemini, Llama, StackOverflow Yes looks is last priority github.com/ariesta-id/t...

submitted 115 days ago • 0 comments

Why is it so hard for LLM to suggest --no-ff and --orphan for my git needs? Sorry for my broken English but I thought it's a common use case and LLM can easily get what I meant.

submitted 116 days ago • 0 comments

Just a reminder that none of the people who make LLMs, no matter how smart, actually know what specific tasks LLMs will be good or bad at. We are barely benchmarking these systems at all on any sorts of tasks. You should explore in areas of your expertise to try to figure it out for your use cases.

submitted 134 days ago • 6 comments

Sometimes our anthropocentric assumptions about how intelligence "should" work (like using language for reasoning) may be holding AI back. Letting AI reason in its own native "language" in latent space could unlock new capabilities, improving reasoning over Chain of Thought. arxiv.org/pdf/2412.06769

submitted 139 days ago • 5 comments

Entropy is one of those formulas that many of us learn, swallow whole, and even use regularly without really understanding. (E.g., where does that “log” come from? Are there other possible formulas?) Yet there's an intuitive & almost inevitable way to arrive at this expression.

submitted 140 days ago • 22 comments

There are only 14 (of 38) provinces in Indonesia of which monthly minimum wage (year 2024) is higher than ChatGPT Pro monthly fee Indonesia is a G20 country

submitted 144 days ago • 1 comment

A $200 ChatGPT. Without open source LM, I would be depressed.

submitted 144 days ago • 0 comments

🐄 Seaweed slashes cattle methane A @pnas.org study shows that a seaweed-based feed additive can cut methane emissions in grazing beef cattle by 37.7%, offering a promising path to climate-smart agriculture. www.pnas.org/doi/10.1073/... #SciComm 🧪 🍎

submitted 146 days ago • 18 comments

The most realistic reason to be pro open source AI is to reduce concentration of power.

submitted 150 days ago • 6 comments

There is a genre of tweet on this site that is just re-writing the sentence “Remember, the people you don’t like are bad people and you’re a good person for hating them” and it always gets over 10,000 likes but then this person gets 10 likes on their amazing nudibranch and I do worry about us…

submitted 150 days ago • 61 comments

I think a lot of the backlash about social media datasets is because people never hear about cool things that regular researchers create or discover. The only thing we see is billionaires selling us blurry reflections of own work.

submitted 150 days ago • 3 comments

You know what opting out of being in the datasets means right?

submitted 152 days ago • 8 comments

People forgot why Facebook was controversial: it's the tracking of the user behavior in the internet, including the likes, who you're connected with, what posts you are engaged. Then, people are afraid the data is used to create a personal-specific manipulative algorithm LLM though...

submitted 152 days ago • 1 comment

The thing is, there's already a dataset of 235 MILLION posts from 4 MILLION users available for months. Not sure why @hf.co is a target of abuse zenodo.org/records/1108...

submitted 152 days ago • 7 comments

Being dismissive of "whatever AI" doesn't help. You have to be specific. What kind of AI? Current "AI"s are usually based on data. What kind of data? Are you concerned with the training, with the architecture? Or with the applications? Using the word "AI" in publications is a mistake

submitted 152 days ago • 0 comments