←
Teddy Svoronos9,402 posts
Teddy Svoronos avatar

Teddy Svoronos

@tedsvo

Senior Lecturer in Public Policy at @Kennedy_School. Tweets my own. Statistics, pedagogy, and other things. Tech-related tweets at @tedsvo_tech

◷ Joined November 2010▧ 493 Photos and videos
Teddy Svoronos avatar
⟳ Teddy Svoronos retweeted
Teddy Svoronos@tedsvo·Oct 25

Amazing addendum - the iPhone Duo behaves correctly in split screen! Can interact with a secondary app while preserving presenter mode.

0:11
Alex Tamkin avatar
⟳ Teddy Svoronos retweeted
Alex Tamkin@AlexTamkin·Oct 27

New Anthropic research: How AI agents shape experts' judgment calls When experts delegate work to an agent, do they still make the key decisions? We tested this in a randomized trial with 50 professional data analysts.

0:15
Teddy Svoronos avatar
Teddy Svoronos@tedsvo·Sep 23

I have a piece in the Globe today about AI and higher ed. TL;DR we spend a lot of time thinking about *how* to teach differently (assessments, cheating) and not enough about *what* to teach differently (reevaluating our learning objectives). bostonglobe.com/2026/09/23/o…

Teddy Svoronos avatar
Teddy Svoronos@tedsvo·Sep 14

PSA for teachers: If you use an iPad to present (eg PowerPoint or GoodNotes) while multitasking with other apps, don’t update to iPadOS 27! Tapping on a different window during a presentation will switch the projection to your iPad screen instead of the presentation.

Teddy Svoronos avatar
Teddy Svoronos@tedsvo·Sep 9

What I hate the most about Apple Watches ambiently recording all your conversations is that Apple has been (correctly) lauded for making features to help people with hearing or vision loss, and they're just taking that technology and dissolving the concept of communal privacy

Cas (Stephen Casper) avatar
⟳ Teddy Svoronos retweeted
Cas (Stephen Casper)@StephenLCasper·Aug 27

I concur with everyone else that the new reports from OpenAI and METR are bonkers. But I think the most important takeaway from these reports is that OpenAI seems to be so obviously backburning. Predictably but conveniently, these reports were exclusively about the factual sequence of events. Neither of these reports had information about the likely negligent decision-making going on inside of OpenAI at the time. The information in these reports is needed, but it's not coming close to everything that is needed right now. The torch now passes to Attorneys General's offices to use their investigative powers. Glad that Alabama took the lead on this a few days ago. alabamaag.gov/wp-content/upl…

Teddy Svoronos avatar
Teddy Svoronos@tedsvo·Aug 21

The problem with AI detectors in my view is not false positives/negatives. It’s that they measure something that we think we care about but actually don’t. Textbook streetlight effect.

Post attachment
Teddy Svoronos avatar
Teddy Svoronos@tedsvo·Aug 16

Worth noting the agentic loop part. We did NOT just write a nice prompt - there was a loop assessing the appropriateness and difficulty of candidate questions. Newer models might be able to one shot this, but that’s not what’s in the paper.

o3-mini (very obsolete) in an agentic loop made good exam questions: "In one of the largest field studies of AI-generated questions to date, we found that this approach achieves psychometric properties on par with those of questions appearing on high-stakes standardized tests."