William Judge, Ph.D.
@williamjudge.bsky.social
I work on AI for batteries
Claude you just made my day!
I just Claude Code to critique your repository and it comes back with “This is a well-engineered, production ready codebase with excellent testing practices and clean architecture patterns”
I just Claude Code to critique your repository and it comes back with “This is a well-engineered, production ready codebase with excellent testing practices and clean architecture patterns”
August 4, 2025 at 6:46 PM
Claude you just made my day!
I just Claude Code to critique your repository and it comes back with “This is a well-engineered, production ready codebase with excellent testing practices and clean architecture patterns”
I just Claude Code to critique your repository and it comes back with “This is a well-engineered, production ready codebase with excellent testing practices and clean architecture patterns”
Has anyone else found that Cursor has become less useful overtime? For me it has mainly turned into a Google replacement/syntax editor
March 21, 2025 at 1:02 AM
Has anyone else found that Cursor has become less useful overtime? For me it has mainly turned into a Google replacement/syntax editor
Are there any LLM evals with increasingly more vague instructions to solve the proposed question?
Because I have noticed a huge difference in performance of various LLMs depending on how detailed my ask is. This directly correlates with how useful these models are to me in a real world setting.
Because I have noticed a huge difference in performance of various LLMs depending on how detailed my ask is. This directly correlates with how useful these models are to me in a real world setting.
January 27, 2025 at 4:44 AM
Are there any LLM evals with increasingly more vague instructions to solve the proposed question?
Because I have noticed a huge difference in performance of various LLMs depending on how detailed my ask is. This directly correlates with how useful these models are to me in a real world setting.
Because I have noticed a huge difference in performance of various LLMs depending on how detailed my ask is. This directly correlates with how useful these models are to me in a real world setting.
When training LLMs does the training process also include incremental improvements in quality based on edits?
Something for the model to know what an improvement from the source material looks like. This could be git/change logs for any code/text.
Something for the model to know what an improvement from the source material looks like. This could be git/change logs for any code/text.
January 12, 2025 at 10:32 PM
When training LLMs does the training process also include incremental improvements in quality based on edits?
Something for the model to know what an improvement from the source material looks like. This could be git/change logs for any code/text.
Something for the model to know what an improvement from the source material looks like. This could be git/change logs for any code/text.
January 12, 2025 at 7:14 PM
Want to know if you can REALLY afford that shiny new thing? 🤔 Here's the REAL talk:
#moneytips #affordability #financialhealth #budgeting
#moneytips #affordability #financialhealth #budgeting
December 17, 2024 at 9:01 PM
Want to know if you can REALLY afford that shiny new thing? 🤔 Here's the REAL talk:
#moneytips #affordability #financialhealth #budgeting
#moneytips #affordability #financialhealth #budgeting
The S&P 500 and Nasdaq closed at record highs today, bolstered by a robust jobs report indicating strong economic momentum.
The S&P 500 edged up 0.2%, while the Nasdaq climbed 0.8%. Conversely, the Dow Jones dipped 0.3%.
Meanwhile, Bitcoin continues to hover near its all-time high.
The S&P 500 edged up 0.2%, while the Nasdaq climbed 0.8%. Conversely, the Dow Jones dipped 0.3%.
Meanwhile, Bitcoin continues to hover near its all-time high.
December 6, 2024 at 10:23 PM
The S&P 500 and Nasdaq closed at record highs today, bolstered by a robust jobs report indicating strong economic momentum.
The S&P 500 edged up 0.2%, while the Nasdaq climbed 0.8%. Conversely, the Dow Jones dipped 0.3%.
Meanwhile, Bitcoin continues to hover near its all-time high.
The S&P 500 edged up 0.2%, while the Nasdaq climbed 0.8%. Conversely, the Dow Jones dipped 0.3%.
Meanwhile, Bitcoin continues to hover near its all-time high.
Is there anything more satisfying than editing a LLM prompt well enough for it to respond to a complicated task accurately within the first reply?
December 2, 2024 at 1:00 AM
Is there anything more satisfying than editing a LLM prompt well enough for it to respond to a complicated task accurately within the first reply?
Top three tips for investing:
1. Save three months of expenses
2. Buy VOO
3. Buy VGT
1. Save three months of expenses
2. Buy VOO
3. Buy VGT
December 2, 2024 at 12:02 AM
Top three tips for investing:
1. Save three months of expenses
2. Buy VOO
3. Buy VGT
1. Save three months of expenses
2. Buy VOO
3. Buy VGT
I am thankful for family, friends, and for improved training/validation loss plots 🦃🍁🍽
November 28, 2024 at 5:50 PM
I am thankful for family, friends, and for improved training/validation loss plots 🦃🍁🍽
I feel really dumb when I realize that PyTorch and Pytoch lightning need the layers to be within the __init__() method to initiate backprop
November 17, 2024 at 12:36 AM
I feel really dumb when I realize that PyTorch and Pytoch lightning need the layers to be within the __init__() method to initiate backprop
Python and git really need a package version number increase requirements similar to when uploading to PyPi.
November 7, 2024 at 4:40 PM
Python and git really need a package version number increase requirements similar to when uploading to PyPi.
Keep your Docker images clean: use ‘RUN pip cache purge’ in your Python app builds.
November 5, 2024 at 6:20 AM
Keep your Docker images clean: use ‘RUN pip cache purge’ in your Python app builds.
Just a reminder that SMOTE does not work for strong classifiers. If it did you would see it being used in research/production/kaggle.
arxiv.org/pdf/2201.08528
arxiv.org/pdf/2201.08528
arxiv.org
November 4, 2024 at 4:22 PM
Just a reminder that SMOTE does not work for strong classifiers. If it did you would see it being used in research/production/kaggle.
arxiv.org/pdf/2201.08528
arxiv.org/pdf/2201.08528
Naive Me: *AWS free tier is great to see if their tech stack is applicable to me at no cost!*
AWS: here is a bill for $287. Thank you for choosing the free tier.
AWS: here is a bill for $287. Thank you for choosing the free tier.
November 3, 2024 at 1:26 PM
Naive Me: *AWS free tier is great to see if their tech stack is applicable to me at no cost!*
AWS: here is a bill for $287. Thank you for choosing the free tier.
AWS: here is a bill for $287. Thank you for choosing the free tier.
Is everyone moving to BlueSky now??
November 3, 2024 at 12:08 PM
Is everyone moving to BlueSky now??
9pm is never too late for a cup, right?
November 2, 2024 at 2:25 AM
9pm is never too late for a cup, right?
What are your top LLM searching apps? Mine are:
1. Perplexity
2. ChatGPT Browser
3. Arc Search
#AI #search #replacingGoogle
1. Perplexity
2. ChatGPT Browser
3. Arc Search
#AI #search #replacingGoogle
November 1, 2024 at 8:59 PM
What are your top LLM searching apps? Mine are:
1. Perplexity
2. ChatGPT Browser
3. Arc Search
#AI #search #replacingGoogle
1. Perplexity
2. ChatGPT Browser
3. Arc Search
#AI #search #replacingGoogle
I have noticed while training transformer models that if the context window is too small the performance is terrible. Can one speed up training convergence by setting a minimum context window length?
#AI #machinearning #gpt
#AI #machinearning #gpt
November 1, 2024 at 2:41 PM
I have noticed while training transformer models that if the context window is too small the performance is terrible. Can one speed up training convergence by setting a minimum context window length?
#AI #machinearning #gpt
#AI #machinearning #gpt
Has anyone tried to plot distribution of the weights/biases of an LLM?
October 21, 2024 at 2:35 PM
Has anyone tried to plot distribution of the weights/biases of an LLM?