Cascadity Audio
← All streams

#table

Everything tagged table, across every stream.

0

Derek Thompson on the decline of Americans hosting each other for dinner

The charts.

Derek Thompson charts how the share of Americans who host or attend a monthly dinner party halved between 1975 and 1998 and asks what hosting looks like now.

Rabbit Holes 1
  • Bowling Alone

    After exploring You Are No Longer Invited to Dinner, the book it builds on is Bowling Alone.

    Robert Putnam’s 2000 book on the decline of American social capital since 1950, the study this piece updates.

    Wikipedia · Wade · 5 min

    0
0

"Does Georgism work? Five years later": Lars Doucet revisits the case for a land value tax

An advocate’s update, including what he now thinks he got wrong.

Lars Doucet, whose review of Henry George’s Progress and Poverty won the first ACX book review contest, revisits his case for land value taxes, noting LVT-enabling laws passed this year in two U.S. states.

Rabbit Holes 2
0

Vermont’s home-battery network has become the state’s largest power source

A working model other utilities are studying, with real household numbers.

Green Mountain Power leases two home batteries to customers for $55 a month over 10 years and pools them into a virtual power plant; more than 5,500 Vermonters now have them, and the BBC reports the network is the state’s largest power source, keeping participants powered through storms.

Rabbit Holes 1
0

The U.S. had its warmest, most humid September on record, the Washington Post reports

The Post has the national picture; the NWS summary shows what the heat looked like on the ground.

The Washington Post reports that the United States had its warmest and most humid September on record. We couldn’t open the Post’s article to check its national figures. The National Weather Service office in Paducah, Ky., reports temperatures 7–8°F above normal across its region, the warmest September on record at Cape Girardeau and Poplar Bluff, Mo., and humidity "more on par with … July".

Rabbit Holes 1
0

California grape growers can’t sell their harvest as wine sales fall

KQED talks to growers in the vineyards; the numbers show how fast drinking habits are reshaping farmland.

Wine sales have fallen more than 20% over five years, and growers have taken roughly a quarter of California’s vineyards out of production. KQED reports Lodi-area growers leaving grapes unpicked and tearing out vines for almonds, pistachios and olives.

Rabbit Holes 1
0

Kids turned the comment section of a quiet NPR podcast on Spotify into a secret group chat

The transcript is free; the audio is better.

This American Life tells how the host of NPR’s Wild Card noticed gibberish comments on a Spotify episode page that turned out to be teenagers using it as a hidden chat room.

0

A professor on redesigning a course after AI agents could do every assignment

Concrete assessment changes other educators can copy or argue with.

The instructor of Machine Learning in Production (100–170 students) describes moving assessment away from take-home work toward TA interactions, exams and video demos once AI agents could complete all assignments, including changes that go against evidence-based teaching practice.

Cascadic Analysis 2
UndertowRole underspecification / the Doorman Problem

What hidden risk could pull against this, even if the news is good?

Take-home assignments did more than measure learning; they forced practice. Redesigning around what AI can complete may remove the part of a course that built the skill, not just the part that graded it.

0
UndertowDelayed Consequence / False Success Problem

What hidden risk could pull against this, even if the news is good?

Students leaning on AI can look competent on assignments for a long time. The gap shows up later, in exams, interviews or jobs, after the feedback that could have corrected it has passed.

0
Rabbit Holes 1
0

Jeff: an open 0.8B "System 1" model that makes fast multiple-choice decisions

The README’s per-task tables show where a tiny model in front of a big one pays off and where it doesn’t.

Jeff is an open 0.8B-parameter model with swappable adapters that picks between options with calibrated probabilities. Its authors report 94.6% mean accuracy across eight tasks at about 0.25 s per decision, versus 86.6% and 8.1 s for a 27B model making every decision itself.

Cascadic Analysis 2
UndertowDelayed Consequence / False Success Problem

What hidden risk could pull against this, even if the news is good?

A small, fast model deciding in front of a big one looks great on accuracy averages. The costly errors are the confident wrong calls it never escalates; they look like successes until their downstream effects appear.

0
UndertowBenchmark scores can overstate real-world reliability

What hidden risk could pull against this, even if the news is good?

94.6% mean accuracy is the authors’ measurement on their own test rows. Real decisions in your domain may be distributed differently, and calibration measured on one dataset can drift on another.

0
Rabbit Holes 1
  • Thinking, Fast and Slow

    After exploring Jeff: millisecond decisions, any domain, the "System 1" idea comes from Thinking, Fast and Slow.

    Daniel Kahneman’s 2011 book distinguishing fast, instinctive "System 1" thinking from slower, deliberate "System 2", the split Jeff’s authors borrow.

    Wikipedia · Wade · 5 min

    0