About:

Vicki Boykis is the author of the Normcore Tech newsletter, where she shares posts from her website, updates on her work on Viberary, and general news. She writes about topics related to machine learning, recommender systems and information retrieval, and the intersection of tech and society.

Website:

Specializations:

Outgoing Links:

Subscribe to RSS:
The blog post revisits the evolution of embedding sizes in machine learning, particularly in the context of text embeddings. It discusses the historical context of embedding dimensions, noting that while 200-300 dimensions were on...
The author discusses the benefits of Full Line Code Completion in PyCharm, highlighting its effectiveness in improving developer productivity while maintaining control over code. They emphasize the importance of logging in debuggi...
Machine learning engineers face the daunting task of navigating production challenges, balancing coding responsibilities with the need for collaboration and error management.
The author reflects on the importance of quality in software development and the dedication required to achieve mastery in coding. They emphasize the value of caring about the code and the users, advocating for a community of deve...
The article discusses the importance of continuous testing and evaluation of LLM-based applications, emphasizing a proactive approach called 'walking around the app.' This involves regularly checking various aspects of the applica...
Embrace the creative process in engineering by exploring, building from scratch, and engaging with the community rather than solely focusing on efficiency.
The author reflects on their 20 years of reading Hacker News and its impact on their career in tech, from being a non-technical major to confidently deploying code. They credit Hacker News for exposing them to technical topics, he...
The author reflects on their experiences in 2024, focusing on their work with LLMs in the tech industry, the importance of engineering discipline in machine learning, and their personal development in learning Italian and music mi...
The text discusses the debate in the software community about whether LLMs can replace developers. It emphasizes the importance of the collective knowledge and experience of developers in writing code, and how this differs from th...
The text discusses the use of large language models (LLMs) for mathematical calculations and reasoning. It explores the historical context of human-computer relationships, the challenges of using LLMs for calculations, and the eff...

0Dead Internet Souls

2024-09-20

The text discusses the concept of 'dead souls' in the context of the internet, drawing parallels to the historical practice of landowners paying taxes for deceased serfs in Russia. It highlights the prevalence of automated and art...
...
The text discusses the concept of user mental models and affordances, and how users navigate different types of online content. It explores the collision of user expectations with the introduction of open-ended language models and...
The text discusses the current state of information retrieval for consumers, highlighting the decline of old companies like Google and the rise of new indie search engines. The author shares their experience using new search engin...

0Redis is forked

2024-04-17

The author discusses the recent forking of Redis and the impact of the licensing changes on the community and future use of Redis. The post reflects on the author's personal experiences with Redis and the implications of the chang...
The text discusses the evolution of machine learning in production, focusing on large language models (LLMs) and their impact on traditional machine learning models. It explains the concept of machine learning as compression and h...
The author shares their favorite books of 2023, including Demon Copperhead by Barbara Kingsolver, The Outlaws by Javier Cercas, Becoming a Technical Leader by Gerald Weinberg, The Last Interview by Anthony Bourdain, Tomorrow and t...
The blog post discusses the use of conditional imports in Python and the reasons behind it. It explains the differences in typechecking enforced by mypy and typechecking at runtime, particularly when dealing with large sets of cus...
The author discusses the transition of Normcore Tech from a ZIRP Substack to a new platform, expressing a desire to consolidate their online presence and write in Markdown for portability. They also mention recent career developme...
The text discusses the pain points and challenges of developing deep learning applications, focusing on issues such as Docker images, build times, CPU/GPU knowledge, model artifact storage, versioning, and reproducibility. It also...

0Naming things

2023-07-11

The author discusses the challenges of naming things in writing and programming, and how it is a crucial aspect of both. They share their personal experiences with naming characters in stories and naming classes in code. The autho...
The author has been working on a project to understand and explain large language models and embeddings. They have written a 70+ page paper about embeddings and have shared it for free. The process was grueling but rewarding, and ...
The text discusses the changing landscape of social media platforms and the shift in user and content creator behavior. It highlights the decline of Twitter and Facebook, the rise of new platforms like TikTok, and the challenges f...
The text discusses the concept of terroir and its relation to the workplace, the importance of rituals in processing transitions, and the lack of physical parting ceremonies in the remote work era. It also highlights the impact of...