London
51.51° N, 0.13° W
kept since 2017
40 entries
Hi, I’m Husain.
I collect questions, take systems apart, and write about the pieces I find.
This is my public notebook: research seeds, reading notes, experiments, and the occasional idea that refuses to leave me alone.
I do not wait for every thought to become a finished argument. Some entries explain what I know; others show me where my understanding still gives way.
since Jan 2026
still open
Question 1 (traceability). How do we build AI systems that can explain where their answers came from?
Remarks.A citation pasted underneath is not enough. I want a real trail of reasoning, evidence, uncertainty and context. The question sits between trustworthy AI, knowledge graphs, research methods and human judgement.
all 40 →
From the notebook
- N° 40
Garbage In: The Unglamorous Half of Machine Learning
I gave a model random numbers and meaningless labels, then watched it score 95 per cent. The result led me backwards through the quieter decisions that shape a model before training begins.
notemachine learning, data, statistics
- N° 39
Writing Knowledge Down
A rule about birds works until the bird is a penguin. From that small exception, I follow the larger problem of representing knowledge without making reasoning impossible.
noteai, knowledge representation, history of ai
- N° 38
Finding the Needle: How Retrieval Ranks Documents
Search begins by reading in advance, but finding a word is only the easy part. I work through indexing, ranking and the surprisingly fragile business of deciding whether retrieval actually worked.
notenlp, search, python
- N° 37
How a Convolutional Network Sees
Two networks can inspect the same pixels while making very different assumptions about them. I follow the nine shared numbers that give convolution its advantage, then test what happens when visual features are borrowed, stretched and generated.
noteai, computer vision, neural networks
& elsewhere
Research, and the garden next door
I am learning my way deeper into AI, machine learning and deep learning, with a particular interest in the structures that make intelligent systems possible to inspect, question and trust. I am less interested in treating AI as magic than in understanding what happens between evidence and answer.
Keywords: trustworthy AI; traceability; human–AI collaboration; knowledge graphs; deep learning.
Research →Some thoughts stay here as notes. Others grow into lessons on Curio Synapse, where I work through difficult ideas in science and technology in public.
This site is the notebook. Curio Synapse is the garden.
Visit Curio Synapse ↗1 active
Sometimes the best way to understand an idea is to build a small version of it. The workbench keeps prototypes, tools, experiments and systems that started as questions.
See what I’m building →