AI by Hand
Professor Tom Yeh's 'AI by Hand' initiative aims to demystify artificial intelligence by meticulously breaking down its math, algorithms, and architectures from first principles. This deep technical dive into AI's foundational elements resonates strongly with the Hacker News community, who are often keen to understand complex systems without abstraction. The project champions the idea that true comprehension comes from building and analyzing these systems 'by hand'.
The Lowdown
Professor Tom Yeh's 'AI by Hand' is a Substack publication and research initiative focused on teaching the foundational elements of artificial intelligence. It emphasizes understanding AI's underlying mathematics, algorithms, and architectures through a hands-on, 'by hand' approach, moving beyond high-level abstractions to foster true comprehension.
- The core philosophy centers on interpretability and explainability, dissecting AI models at their most fundamental levels.
- Content includes articles, live seminars, and a research library for subscribers, detailing topics like the internal workings of specific models (e.g., Qwen 3.6).
- The project aims to equip learners with the ability to truly understand AI by engaging with its creation from the ground up, echoing the sentiment that 'what I cannot create, I do not understand'.
Ultimately, 'AI by Hand' offers a rigorous path for those seeking to move beyond superficial understanding and delve into the core mechanics that drive modern AI systems.
The Gossip
Subscription Scrutiny
Many initial visitors to the site encountered a subscription wall, leading to confusion and frustration about accessing content. Commenters questioned the user experience design, with some offering workarounds like clicking 'Later->' or navigating directly to the library page. This highlighted a common HN complaint about websites that immediately gate content.
Foundational Frameworks
The 'by hand' educational philosophy struck a chord with the community, leading to shared resources and similar projects. Users pointed to other initiatives, books, and personal projects (like 'ml-by-hand' inspired by 'micrograd') that advocate for building AI systems from scratch to gain a deeper, more intrinsic understanding of their mechanics.
Pedagogical Praise
Despite initial UX hiccups, there was significant appreciation for the 'by hand' methodology itself. Commenters praised the approach, noting its effectiveness in teaching complex topics like quantum computation algorithms and expressing enthusiasm for diving into such a detailed exploration of AI fundamentals.