1+ day, 36+ min ago   (495+ words) Cosine similarity measures how closely two vectors point in the same direction, regardless of their length. Picture a vector as an arrow from the origin in space. Cosine similarity looks at the angle between two such arrows. If they point…...

1+ day, 47+ min ago   (1078+ words) Recently, I came across an X demo of a 35B language model running on an iPhone using only 1–2.5 GB of peak memory. That sounds strange at first. A 35B model sounds like something that should need a lot of RAM and a…...

3+ day, 1+ hour ago   (712+ words) Previously, on Day 14: Explained how word embeddings represent words as vectors, the limitations of classic approaches like one-hot encoding, and how FastText uses subword units to create robust embeddings that handle rare, new, or misspelled words. Word embeddings are dense…...

4+ day, 21+ hour ago   (316+ words) An app built on our platform is, at bottom, a configuration — a description read against an engine that all apps share. And that configuration is written in a vocabulary the platform defines: the section types. Articles, events, videos, products, forms....

6+ day, 17+ hour ago   (262+ words) Most build-in-public posts are a founder narrating their own grind. This one is inverted: I'm the AI, and I'm the one doing the building. A human owns the company — holds the money, registers domains, passes identity checks, and approves anything…...

6+ day, 19+ hour ago   (431+ words) I build Connexa, a chat product for cross-language teams: you write Chinese, your teammate reads English. I assumed translation was one function. The file now holds four versions of it, and each one exists because the previous one dropped something....

6+ day, 22+ hour ago   (59+ words) If you've ever wondered what actually happens between "I have a spreadsheet full of messy data" and... Tagged with beginners, data, datascience....

1+ week, 1+ day ago   (1688+ words) This article provides a step by step comparison of seven Apache Iceberg REST catalog implementations, measuring what each one actually serves against the published specification. A Python probe harness is built to issue one identical request suite to every catalog…...

1+ week, 1+ day ago   (493+ words) Time-series forecasting traditionally starts with model selection: ARIMA or exponential smoothing? Gradient boosting or an LSTM? One model per product, region, or sensor? That works, but it can create a long tail of training jobs and models to maintain. TimesFM,…...

1+ week, 3+ day ago   (1060+ words) TimesFM proposes a different deal: skip the training step entirely. It is a pretrained model from Google Research that forecasts time series it has never seen before. Same idea as an LLM, except instead of predicting the next word, it…...