56/And DeepSeek strikes deep
Monday, 2025-01-27 to Sunday, 2025-02-02
Created:
Thoughts
¶ Creating space. When you are trying to attract people to work with you, you have to create a space for them to visualize themselves in. One of the biggest challenges as a founder is to be able to attract the next person to join you. A lot of this comes down to - not just convincing the other person of the potential of the idea – your friend, your mom, the investor, the guy at the meetup, etc might be excited about the idea, but it is not the same as someone deciding to come work with/for you. Where do they see themselves as being able to contribute to the execution?

The idea of beauty, and art is something that has repeatedly come up last week, while reading the art life, david lynch, and rick rubin.
Books
The Art Life by Robert Henri, which I bought on the ‘recommendation’ by David Lynch, who passed away last month (Jan 2025), in his book Catching The Big Fish
Still working through Dead Companies Walking by Scott Fearon. The theme of the book is inline with bet on the contrarian ideas that work idea. Being a correct contrarian is a lot of work, and takes courage, conviction in own’s ability to do the work, and correct risk assessment to stay alive (ie., not go bankrupt) etc.
Param reminded me that I had purchased Rick Rubin’s The Creative Act: A Way of Being, so that’s back on top of the kindle reading stack.
Large Language Models
This week’s flavour is deepseek.
I have purchased Chip Huyen’s AI Engineering book. Looking forward to reading it.
Also found aman.ai to be a good wiki for many AI/ML concepts. More people should aspire to make their own.
containers/ramalama:
The goal of RamaLama is to make working with AI boring. by doing
certain decisions on your behalf and run the LLM with a simple command
like – ramalama run instructlab/merlinite-7b-lab
, and have
it run without having to figure out if your pc has GPU or not, etc.
Mathematics for Machine Learning book.
MLX is a NumPy-like array framework designed for efficient and flexible machine learning on Apple silicon, brought to you by Apple machine learning research.
Chat with a PDF file using Ollama and Langchain - M1lt0n on software – a tutorial to use locally running llm on ollama; in this example llama3.
Qwen2.5-Max: Exploring the Intelligence of Large-scale MoE Model | Qwen
Software
LocalSend: Share files to nearby devices – “The only real downside is that you have to open the app on both sides for the share to complete”
Link Dump
- Signals and Threads | The Future of Programming; podcast with Richard Eisenberg, one of the core maintainers of Haskell. He recently joined Jane Street’s Tools and Compilers team, where he hacks on the OCaml compiler. He and Ron discuss the powerful language feature that got him into PL design in the first place—dependent types—and its role in a world where AIs can (somewhat) competently write your code for you. They also discuss the differences between Haskell and OCaml
- unslop_the_huggingslop.py
- What I learnt this month: Jan 2025, where he mentioned Algorithms and Data Structures for Massive Datasets, and the inspiration for me to bring back the weekly journal format
- Pradyumna Prasad has a new blog - pradyuprasad.com
- Sheldon Axler’s “Linear Algebra Done Right, 4th Edition”, the often cited, but seldom read (by me) linear-algebra book.
- MISO Business Practices Manuals (public info).
- val.town is a website to write and deploy Typescript code – can create APIs, and “cronjobs”. The use-cases page is indicative of what can be done.
Pages created
deepseek, search-ai, cuda-triton, people/david-lynch|David Lynch, clippings/localsend|LocalSend