RSS

jcksanderson

Karma: 224

We’re talk­ing past our mod­els; or, How a model defined its “evil” vec­tor as dread

jcksanderson20 Jul 2026 7:27 UTC
45 points
5 comments9 min readLW link

In­ter­pretabil­ity is be­com­ing in­creas­ingly uninterpretable

jcksanderson9 Jul 2026 7:04 UTC
34 points
1 comment4 min readLW link
(jcksanderson.com)

Don’t ig­nore the car crashes, and re­mem­ber your fresh­man CS

jcksanderson26 Jun 2026 7:06 UTC
38 points
0 comments2 min readLW link
(jcksanderson.com)

jck­sander­son’s Shortform

jcksanderson24 Jun 2026 10:31 UTC
2 points
6 comments1 min readLW link

Iter­a­tive Fine­tun­ing is Mostly Idempotent

Zephaniah Roe, jcksanderson and Julian H
11 May 2026 6:41 UTC
24 points
0 comments5 min readLW link

In­tro­duc­ing the XLab AI Se­cu­rity Guide

Zephaniah Roe, jcksanderson and Julian H
27 Dec 2025 16:50 UTC
20 points
1 comment5 min readLW link

In­trigu­ing Prop­er­ties of gpt-oss Jailbreaks

Zephaniah Roe and jcksanderson
13 Aug 2025 19:42 UTC
20 points
0 comments10 min readLW link
(xlabaisecurity.com)