December 2025: What you missed

Happy New Year!

Friends have been asking how they can stay up to date with what I'm working on and keep track of the things I read, write and share. RSS feeds don't seem to be state of the art anymore, apparently. So I'm starting this mailing list.

Recently I've been spending some time on an autonomous AI-powered academic research system. The pipeline handles literature review, structured evidence extraction, optional computation, and paper assembly using a multi-agent architecture with 25 specialized agents. It's built as a filesystem-first system where most components write durable artifacts into project folders, making outputs inspectable and gates re-runnable (link see below). I've also been finishing up a paper on Attention Dynamics in Online Communities, looking at power laws, preferential attachment, and early success prediction on Hacker News.

To my surprise, the site quietly crossed 10,000 unique monthly visitors in November. I didn't expect that, but it’s a nice motivator to keep going.

What I've been writing

Book Review: Why Machines Learn

A review of Ananthaswamy's Why Machines Learn: A clear tour of the math behind modern machine learning, with real derivations and geometry.

How AI is Shaping My Investment Portfolio for 2026

My 2026 investment strategy: portfolio preparation for AI advancements, dollar risks, and Europe's growth with allocation rationale.

Not Logan Roy: Netflix vs. Paramount's Bidding War

Netflix's $72B Warner Bros. deal vs Paramount's hostile $30/share tender: streaming economics, deal mechanics, and antitrust uncertainty.

Nike's Crisis and the Economics of Brand Decay

Economic analysis of Nike's crisis: how complementary assets, path dependency, and strategic misalignment destroyed competitive advantage.

What I've been working on

GitHub - giatenica/gia-agentic-short: Autonomous AI-powered academic research system.

Autonomous AI-powered academic research system. Contribute to giatenica/gia-agentic-short development by creating an account on GitHub.

GitHub - philippdubach/hn-archiver: A HackerNews archiving system built on Cloudflare Workers. Captures posts and comments in real-time, tracks how they change over time, classifies content with AI, and supports semantic similarity search using vector embeddings.

A HackerNews archiving system built on Cloudflare Workers. Captures posts and comments in real-time, tracks how they change over time, classifies c...

What I've been reading

■ American Option Pricing Under Time-Varying Rough Volatility: A Signature-Based Hybrid Framework via arXiv

■ The Thinking Game: A journey into the heart of DeepMind via Youtube

■ Choosing the Right Python Docker Image for Finance Workloads: Surprising differences in container startup times and memory usage via Jiripik

■ If You Don't Design Your Career, Someone Else Will: Intentional career planning versus reactive career management via Gregmckeown

■ I sell onions on the Internet: Insane way to start a business via Deepsouthventures

■ Marginalia Search: A search engine that favors text-heavy sites and punishes modern web design via Marginalia-Search

■ Marketron through the Looking Glass: From Equity Dynamics to Option Pricing in Incomplete Markets via arXiv

■ AI trading in real markets: Exploring the limits of LLMs as quant traders via Nof1

■ If AI replaces workers, should it also pay taxes?: Bill Gates says yes! via English

■ Time might not exist – and we're starting to understand why: How modern physics challenges our fundamental understanding of time via Sciencefocus

■ Global Memory Shortage Crisis: Analysis of the semiconductor memory shortage and its potential impact on smartphone and PC markets in 2026 via Idc

■ The Universal Weight Subspace Hypothesis: Neural networks trained on diverse tasks converge to similar low-dimensional weight subspaces via arXiv

Have a great start to the new year! Looking forward to sharing more insights with you in 2026 - Phil

Have feedback, comments, or ideas? I'd love to hear from you: [email protected]

Blog | Projects | Research | GitHub

to unsubscribe please reply to this email