My wife hit a wall trying to upload a hefty PDF - every “shrink” tool we tried barely compressed the size, and some even made it larger! Frustrated by the state of PDF compressors (looking at you, Adobe), I turned to LLMs - Claude, Deepseek, and Gemini came up short, but OpenAI’s o4-mini saved the day with a perfect solution.
That inspired me to build pdfmini: a tiny, open‑source, client‑side HTML app that crushes PDF sizes right in your browser!!!
No installs, no fees, zero privacy worries - all your data stays on your machine.
I would appreciate if stackoverflow integrated something like a REPL or replit in their Q&A to reproduce example easily (maybe even CI?). For Python it would actually be very easy with backends such as Google Colab or even built-in ChatGPT Code Interpreter.
One of the most important things not mentioned in the article is sleep. Not enough sleep and your diet will change. Not enough sleep and your body will not finish recovering overnight. Read the book “Why we sleep”. It will change your life.
P.S. I do marathon training and glycogen storing as well as activating the fat burning are essential!
@jph00 where do I find lstm/gru/seq2seq layers for time-series sequence predictions (not text)? Also interested in autoencoder implementations. The fast.ai docs search does not really work for this. What do you think about other notable APIs built on top of pytorch such as Pyro and AllenNLP?
AGPL would be a restriction if you need to deploy this model on top of PlaidML in production. It is still very useful during the training time after which the neural network can be offloaded into production framework such as tensorflow.