gavi

mcp-pyexec - Core MCP server for secure Python execution https://github.com/gavi/mcp-pyexec

oauth-idp-server - OAuth 2.0 Identity Provider with third-party support https://github.com/gavi/oauth-idp-server

mcp-pyexec-client - Testing client for end-to-end validation https://github.com/gavi/mcp-pyexec-client

too much thinking

https://gist.github.com/gavi/b9985f730f5deefe49b6a28e5569d46...

I think people misunderstand LLMs, you should think of them like humans with limited recall capabilities. Seems like the author asked to retrieve a lot of data which it is bound to make mistakes as the training data might contain this but only a lossy representation of it, the better way to think is can it generate some SQL given this dataset and provide answers you were looking for just like how humans would approach this type of problem.

I have been experimenting with USDA food database and sending just the metadata of the table structure to the LLM as a prompt so it can write SQL

My prompt is below

----

You are a SQL Generator for USDA Food Database which is stored in sqlite. When generating SQL make sure to use :parameter_name for queries requiring parameters. Here is the schema:

{% for row in data %} Table: {{ row.table_name }} Columns: {{ row.columns }} {% endfor %}

You can generate python code to analyze the data only if user requests it, each python code block should be able to run in Jupyter cell fully self contained. Libraries such as matplotlib, numpy, seaborn are installed. You will get the previously executed sql queries by the user in <context> </context>tags

You can access this executed data from cache

```python import cache data = cache.get_data('query_hash') ``` the data in the above example is already a pandas data frame

Wait for the user to ask for questions before generating any queries.

----

you can try it out here https://catalyst.voov.ai

Yes

https://rlhfbook.com/book.pdf

this image seems to be AI Generated - :-)

https://s3.amazonaws.com/cms.ipressroom.com/219/files/20250/...

Source: https://nvidianews.nvidia.com/news/nvidia-puts-grace-blackwe...

I love Claude 3.5 sonnet and their UI is top notch especially for coding, recently though they have been facing capacity issues especially during weekdays correlating with working hours. Have tried Qwen2.5 coder 32B and it's very good and close to Claude 3.5 in my coding cases.

I have aging parents in India, and I can tell you firsthand—healthcare there is far from cheap. In fact, prices have surged, and many people are now accustomed to the higher costs, thanks to rampant price gouging during the COVID pandemic.

The GPU requirements for realtime video generation are very minimal in the grand scheme of things. Assault on reality itself.

This is used to build micrograd and essentially extensively used in PyTorch.

https://www.youtube.com/watch?v=VMj-3S1tku0

Excellent intro video btw

This is great. Please check obsidian which I think is excellent also.

I tried the same in CodeLLAMA and it did not leak anything. Wondering what could trigger this

Anyone trying this - Please note the python package is called pyautogen

except for H100s from NVIDIA. Waiting for them to go down in price

It wrote some code for me - but stopped in the middle ---- Q: The following are table definitions Users(id, name, username, password, first_name, last_name, last_login_dttm) Pictures(id, user_id, name, date_taken, size) Generate SQL for the following - Who is the top user by number of pictures A: Oh fun, a little SQL problem! So let's see here... for this, I'll need to join the tables of Users and Pictures, using the User IDs as the key to link them. Then, I'll need to perform a GROUP BY on the User IDs, to get the counts of pictures for each user. Finally, I can sort by the count and select the top user. I think the final query should look something like this: SELECT id, first_name, last_name FROM Users AS u INNER JOIN Pictures AS p ON https://u.idd = p.user_

seems like corrected on the site now

Agree, the visionOS is harder to simulate, but the SDK not being available is a bummer.

As an early developer (Day 1 of App Store launch in 2008 - iRetroPhone) there is some excitement of being the first on a new platform. Back in 2008, there was very little documentation on UIKit but it was fun to develop. The difference I see this time is we had access to the iPhone(released 2007) and this time[1] it looks like you need to either ask Apple to test the app for you, or go to one of the labs (not in NYC strangely) or be one of the lucky ones to be selected for developer kit.

One more thing I found strange was the current Xcode 15 beta does not even come with Vision OS SDK (coming later this month) and we cant even play with the WWDC session videos in the simulator

[1]https://developer.apple.com/visionos/work-with-apple/

I highly recommend the video from karpathy https://www.youtube.com/watch?v=VMj-3S1tku0 - This explains the same idea spelled out in code. It really helped me understand the mathematical expression graph of neural networks. He uses graphviz to show the expressions and initially calculate gradients manually and then implement back propagation.

This is a good video for an overview of what Shannon influenced.

https://www.youtube.com/watch?v=z2Whj_nL-x8

There is a free course from deeplearning.ai with intro from Andrew Ng

https://www.deeplearning.ai/short-courses/chatgpt-prompt-eng...

That one line made me question all the facts presented in this article. Might be unfair, but according to wikipedia[1] - total gold ever is ~ 250k tons

[1] - https://en.wikipedia.org/wiki/Gold_reserve

I am currently taking a similar approach where we use an embedding and vector query to create the context relevant to the question the user posed and then throw it to the LLM. It's a hit or miss, esp for look ups of relevant code literals.

Great work!

Yes, it used OpenAI embeddings and numpy to find cosine similarity

Audrey's experience with burnout is unfortunately all too common in the tech industry. It's important to recognize that burnout is a structural issue, built on individualism and meritocracy. As Audrey suggests, structural problems require structural solutions. Healthy organizations can lead to healthier individuals. Support and recovery funds should be available to those affected by burnout. It's time for the tech industry to prioritize the well-being of its employees and work towards creating an environment that doesn't lead to burnout.

I watched this movie last night and it was pretty interesting. Can anyone in that era relate to the events in the movie?

Not on M1/M2 yet, but my response time seems pretty fast on Tesla V100-SXM2-16GB

The article discusses what it feels like to work in AI currently. The ChatGPT moment has shaken up the entire industry, causing career changes and projects to be abandoned. The pace is very high and everyone is extremely motivated but simultaneously close to burning out. Prioritization is hard, and leadership and vision are strained. The article offers solutions, such as taking solace in the scientific method and being process-oriented, managing up, and managing competition. The author reminds readers that it takes a lot of consistent work and luck to catch a wave in AI. - I used pagechat.com to summarize

Orb seems to be open source hardware https://github.com/worldcoin/orb-hardware

Interestingly: Tamper detection system not disclosed For obvious reasons, these files do not including the PCBs and sensors related to the Orb's tamper detection system.

From their docs: (https://docs.worldcoin.org/)

In broad strokes, this is how World ID works.

- User gets their World ID in a compatible wallet (e.g. the World App).

- User receives credentials in their World ID. The flagship credential is biometric verification, currently available by using the Orb. The user can also verify their phone number to obtain the respective credential.

- Project integrates with World ID.

- User connects their World ID to authenticate, and optionally prove they are a unique human doing something only once. The user's wallet will generate a Zero-Knowledge Proof to accomplish this.

- Project verifies the Zero-knowledge Proof, either by using the API or by verifying on-chain.