April
I have been catching up with all the recent development in the world of LLMs and Agentic Workflows. Deepseek made a splash beginning of the year but apart from bigger models the only recent interesting development is the sky-rocketing adoptions of MCP about which I wrote more in SumUp MCP atom.
Reading
Modern Day Oracles or Bullshit Machines? is a collection of lessons for the world of ChatGPT where anything can be bullshit. It’s an exceptionally well-done multi-chapter curriculum suitable for class rooms.
One afternoon around 4:50 p.m. John von Neumann came by and saw what Fermi had on the blackboard and asked what he was doing. So Enrico told him and John von Neumann said “That’s very interesting.” He came back about 15 minutes later and gave him the answer. Fermi leaned against his doorpost and told me, “You know that man makes me feel I know no mathematics at all.”
Working With Fermi at Chicago and Post-war Los Alamos by Richard L. Garwin hints at the greatness of John von Neumann. It was cited in a random Reddit thread alongside other quotes, stories, and achievements of which I was embarrassingly mostly familiar with von Neumann architecture and his contributions to the computer science. As it turns out, von Neumann’s contributions span mathematics, physics, economics, and statistics in just as much depth in what amounts to a career with field-coverage and significance unmatched during his life.
The core narrative control is straightforward: 1) everything’s great, and 2) if it’s not great, it’s going to be great. Whatever’s broken is going to get fixed, AI is wunnerful, and so on.
I Quit! The Tsunami of Burnout Few See hits close to home. Every company all-hands with talk about EBITDA, active users, product adoption but when the time comes for the tough discussion regarding last eNPS survey, we get a few empty promises and then hear about the declining score until the next survey. I came to terms with the company being purely for-profit (albeit nobody is able to state it clearly) and I know that assuming anything else was foolish. Yet, I still naively hope that one day, we put people first.
Watching
About the magic behind the speed and responsiveness of Linear. Linear uses mobx for state management and relies heavily on offline-first approach for responsiveness.
This month I learned
The Hilbert curve is a continuous fractal space-filling curve first described by the German mathematician David Hilbert in 1891, as a variant of the space-filling Peano curves discovered by Giuseppe Peano in 1890.