pplonski86

Is codex working well with python notebooks?

Wow! how do you make marketing for so many projects?

Lately I was trying ask LLMs to generate SVG pictures, do you have famous pelican on bike created by flash model?

This is amazing. Porting library from one language to one language are easy for LLMs, LLMs are tired-less and aware of coding syntax very well. What I like in machine learning benchmarks is that agents develop and test many solutions, and this search process is very human-alike. Yesterday, I was looking into MLE-Bench for benchamrking coding Agents on machine learning tasks from Kaggle https://github.com/openai/mle-bench There are many projects that provide agents which performance is simply incredible, they can solve several Kaggle competitions under 24 hours and be on medal place. I think this is already above human level. I was reading ML-Master article and they describe AI4AI where AI is used to create AI systems: https://arxiv.org/abs/2506.16499

Working on new backend for Mercury framework. Mercury makes it easy to serve Python notebooks as web apps. You dont need to know HTML, CSS, JS to build beautiful looking web app from Python notebook. I'm writing new backend to support ipywidgets, anywidgets in Mercury. https://github.com/mljar/mercury

Amazing! Great work. Congratulations on launch.

Few questions: 1. Can it work with tabular data, images, text and audio? 2. Data preprocessing code is deployed with the model? 3. Have you tested use cases when ML model was not needed? For example, you can simply go with average. I'm curious if agent can propose not to use ML in such case. 4. Do you have agent for model interpretation? 5. Are you using generic LLM or have your own LLM tuned on ML tasks?

If you are looking for Python notebook that is beginner friendly, I'm working on desktop app called MLJAR Studio. It packs Python and JupyterLab in single executable - installation is stright-forward. Additionally, there are extensions that makes working with Python much easier for beginners: for example, variable inspector and packages manager.

I think you can easily load JSON data into Pandas DataFrame. Then you can create visualizations and compute statistics. Python might be usefule for that.

It is much easier to implement chat bot that intelligent workspace, and AI many times doesn't need human interaction in the loop.

I would love to see other interfaces other than chats for interacting with AI.

I think it is good, and it fuels innovation. We need to try even strange ideas, because maybe some of them will work.

We write and run tests to build trust in our code changes. But maybe tests aren’t the only way to achieve that trust.

When I was younger, I had a friend who was a senior software engineer. I remember he would make changes to production systems without even running the application locally or executing any tests, and yet his changes never failed. The team had a high level of trust in all his code changes.

Could you please recommend some history books?

Boring tech is boring because it is reliable. You don't need to fight constantly with it. You can forget about it, and focus on your work. Maybe reliable equals mature.

Agree. The software should be good and reliable. Then we can think about small and simple improvements like that.

I would say that software should be first smart, then it can be fun.

It is about the tools that you are using. The better the tool, the faster the job is done. I always get the best machine I can afford and never regret it. I'm working with data; I have a desktop machine with 256 GB RAM and 36 CPUs; it was expensive, but I can train huge gradient boosting tree models in a reasonable time.

> You write because you think, because you observe, because you need to put it somewhere.

This resonates with me. Many things get more precise when I write down my thoughts or explain them to others.

For my company, I use the blog to explain how my software is working, I'm using blog for that as well. It is a good marketing channel.

> Still, imagine it as a real-but-relatively-junior virtual coworker. Now imagine 1,000 of them. Or 1 million of them. Now imagine such agents in every field of knowledge work.

I really don't want a future in which I had to supervise 1 million of real-but-relatively-junior virtual coworkers. I would prefer 1 senior over 1 million juniors. I'm in the programming industry and I don't think that coding might be scaled.

It takes some effort, but there’s a lot to gain. When I track what I eat and keep my daily calories in check, I feel much better. If I’m unsure of the exact calorie count, I’ll estimate a bit higher - around 1.2x.

My dream is to create product that I don’t need to promote. I would like to have such market fit that people will search for product by them self.

I'm working on AI assistant in Python notebook. It aims to help with data science tasks. I'm not using it to do a full analysis. It will fail. What I ask is to create a code snippet for my next step in the analysis. Many times I need to manually change the code, but it is fine because LLM speed-up my coding a lot. And it is really fantastic in writing matplotlib code for visualization. I don't remember all matplotlib syntax to change axis labels, add annotations or change style, and LLM really can handle it good, in impressive speed.

Amazing results! Beating AutoML with single model is not easy :)

Could you please explain like I'm five what is doing a trick? You have model pre-trained on large set of small datasets and you leverage it to boost performance?

Training is fast, few seconds, but what is time needed to compute predictions?

How large is the model?

Congrats on launch. I watched the demo video, the UI looks nice. Is it possible to write raw Python or SQL to query data? Do you have automatic interval refresh of data? What databases do you support?

Thank you for comment. Version control is crucial for software development. Data science has a lot of experimentation involved. In my opinion, not everything should be in version control system. My typical workflow is to experiment heavily with Python notebooks, without version control. If I have analysis procedure ready then I apply version control for Python notebooks or re-write code to Python scripts or modules. I think it all depends on what problem you are working on, and the experimentation level.

I'm the founder of MLJAR company, and we are working on AI assistant Python notebook. It is desktop app that aims to help starting in data science.

I love it. Small data is beautiful :) Many times I heard about people running clusters to analyze datasets that can fit in single machine :)

Congrats on V1. Could you please provide more details behind service? How does it work? What is the most unique feature comparing to other ai search engines?

Congrats on launch. I see many people learning on YouTube, that might be useful tool. Probably you can easily integrate it with LLM and generate summary and quizzes based on videos

I got similar experience. I really prefer switch CUDA version with whole PC machine. What is more, the speed and memory of hardware improves quickly in time as well.

There are many LLMs that have tools (function calling) available in their API. Some time ago, I was building AI agent that has profiler connected by function calling, and it works exactly as you described. First produce the code, then call profiler, and last step was to optimize code.

I'm in similar situations, I jump between many environments, mainly between Python and Typescript, however, currently testing a new idea of learning algorithm in C++, and I simply don't always remember all syntax. I was very skeptical about LLMs at first. Now, I'm using LLMs daily. I can focus more on thinking rather than searching stackoverflow. Very often I just need simple function, that it is much faster to create with chat.

Why do you think that stock and crypto plugins will be very profitable? There are no tools to combine finance data in easy way?