Favorites

72 bookmarks · page 1 of 2
★ 2026-10-06 · statology.org

5 Statsmodels Functions That Give You the "Why" Behind the "What"

Statsmodels answers the questions sklearn was never built to answer: is this coefficient real or noise, does this added complexity earn its keep, can I trust these p-values in the first place, and which specific data points are quietly running the whole show.

Note: Run a regression in scikit-learn, and you get exactly two things back: a set of coefficients and a .predict() method. Ask it for a p-value on any of those coefficients, and there’s no built-in way to get one. This isn’t a missing feature scikit-learn’s maintainers overlooked; it’s a deliberate design choice they’ve stated directly: scikit-learn is built for prediction, not statistical inference…

★ 2026-09-22 · open.substack.com

The Quest for Universal Flu Vaccines

Modern flu vaccines have an average efficacy of just 40 percent, and they must be revamped each year. How can we make vaccines that are “universal” — both broadly-protective and highly potent?

influenza evolves so quickly that a novel virus emerges each flu season
inactivate viruses with formaldehyde
★ 2026-07-26 · newyorker.com

Rope-a-Dope

Are the Enhanced Games an existential threat to fairness or a cure for death? Zach Helfand reports.

The next morning, I caught a training session at the pool. It was difficult to discern the outward effects of enhancement on the athletes. The swimmers looked like swimmers—sinewy and long, with huge hands and feet. An exception was James Magnussen, an Australian who’d been the first person to sign…
★ 2026-07-12 · open.substack.com

The Eight Load-Bearing Humans Every Civilization Needs

Every civilization runs on eight irreducible roles. AI is hollowing out the derivative layer on top of them. Which layer does your life actually rest on?

Every society runs on the same small set of irreducible jobs, the people who grow food, heal bodies, build and fix things, move goods, transmit knowledge, defend surplus, settle disputes, and hold communities together.
★ 2026-06-23 · openculture.com

1,700 Free Online Courses from Top Universities

Take online courses from the world’s top universities for free. Below, you will find 1,700 free online courses from universities like Yale, MIT, Harvard, Oxford and more. Note: This page includes a lot of Massive Open Online Courses (MOOCs).

★ 2026-05-06 · nytimes.com

Josh Hart is more than a glue guy. He’s an example of ‘the street sweeper’ mentality

Hart remains an example of the street sweeper mindset. He does not need the ball to make an impact, and he understands who he is.

“If it falls to your lot to be a street sweeper, sweep streets like Michelangelo painted pictures. Sweep streets like Beethoven composed music. Sweep streets like Leontyne Price sings before the Metropolitan Opera. Sweep streets like Shakespeare wrote poetry.”
★ 2026-04-14 · thebreakthrough.org

The Aluminum Tech Stack

few analysts have sought to dissect how China’s market share of unglamorous bulk raw material inputs and intermediates across its minerals, metallurgy, and chemicals sectors actually directly enables their market share in these finished, shiny consumer products.
In reality, metals and chemicals production require muscular industrial policy of their own to succeed
★ 2026-02-24 · rubyllm.com

Getting Started

Start building AI apps in Ruby in 5 minutes. Chat, generate images, create embeddings - all with one gem.

★ 2026-01-16 · timharford.com

What We Get Wrong About Technology

Blade Runner (1982) is a magnificent film, but there’s something odd about it. The heroine, Rachael, seems to be a beautiful young woman. In reality, she’s a piece of technology — an organic robot …

The toilet-paper principle suggests that we should be paying as much attention to the cheapest technologies as to the most sophisticated.
barcode
★ 2026-01-01 · simonwillison.net

2025: The year in LLMs

This is the third in my annual series reviewing everything that happened in the LLM space over the past 12 months. For previous years see Stuff we figured out about …

It turned out that the real unlock of reasoning was in driving tools.
notable result is that AI assisted search actually works now.
★ 2025-12-31 · sebastianraschka.com

The State Of LLMs 2025: Progress, Problems, and Predictions

A 2025 review of large language models, from DeepSeek R1 and RLVR to inference-time scaling, benchmarks, architectures, and predictions for 2026.

DeepSeek released their R1 paper in January 2025, which showed that reasoning-like behavior can be developed with reinforcement learning, it was a really big deal
DeepSeek R1 was released as an open-weight model
★ 2025-07-19 · gojiberries.io

Advertising Without Signal: The Rise of the Grifter Equilibrium

Economists credit ads with two welfare‑enhancing roles: 1. Informative – trimming search costs (Stigler 1961 (pdf)). 2. Signaling – In classic models, high-quality sellers are more willing to incur large, sunk ad costs because they expect to recoup them through future…

Five Frictions That Flatten the Signal
Because tarnished names can be retired cheaply, long‑run reputation no longer disciplines bad actors.
★ 2025-04-20 · quantamagazine.org

To Make Language Models Work Better, Researchers Sidestep Language | Quanta Magazine

We insist that large language models repeatedly translate their mathematical processes into words. There may be a better way.

This reliance on filtering concepts through the sieve of language can also result in a loss of information
deep neural networks that allow language models to continue thinking in mathematical spaces before producing any text.