BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: AgentBreaker: A blind spot detector for your coding 
 agents\n   Tags: AI Village | Creator Talk/Panel\n   When: Friday\, Aug 7\
 , 16:30 - 16:59 PDT\n   Where: LVCCW Level 1 Hall 3 1104 (Creator Stage 4)
  - [1]Map\n\n   Description:\n\n   Coding agents have the capability to wr
 ite and ship production code\n   with little human review\, which can lead
  to large-scale security\n   threats. At Corridor\, we believe in securing
  code generation at the\n   source: by augmenting coding agents with the n
 ecessary security\n   tooling.\n\n   While evaluating our product\, we qui
 ckly realized that there was no\n   principled way to measure whether a gi
 ven agent harness actually\n   performs well on the axes we care about (li
 ke security). Static\n   benchmarks go stale\, get contaminated\, and rare
 ly capture the\n   multi-step behavior of an autonomous agent.\n\n   To ad
 dress this\, we introduce AgentBreaker\, an open-source framework\n   that
  extends the automated benchmark construction tool AutoBencher\,\n   takin
 g it from evaluating single LLM calls to full coding-agent\n   harnesses. 
 We instantiated the framework on our task of secure code\n   generation an
 d show that it reliably surfaces actionable\, agent\n   harness-specific w
 eaknesses.\n\n   We release AgentBreaker and its methodology so you\, too\
 , can evaluate\n   agent harnesses on the axes that matter to you.\n\n   S
 peakers:Aditi Narasimhan\,Farzaan Kaiyom\n\n   SpeakerBio:  Aditi Narasimh
 an\n\n   Aditi Narasimhan is the founding AI research engineer at Corridor
 \,\n   where she works to improve the security of agentic code generation\
 n   tools. Previously\, she was at Carnegie Mellon\, where her research\n 
   focused on AI ethics\, pragmatic theory\, and tech ethics pedagogy.\n\n 
   SpeakerBio:  Farzaan Kaiyom\n\n   Farzaan has built AI products at start
 ups from the pre-seed stage\n   through Series C\, wearing hats as a found
 er and ML Engineer. He holds\n   a BS/MS in Computer Science from Stanford
  where he focused on AI\n   safety. While at Stanford CRFM\, he worked on 
 HELM\, which won the\n   Stanford Open Source Software Prize\, and present
 ed research at ICLR.\n\n   '\n\n   1. #LVCCW_Level1_Hall3\n\n\n
DTEND:20260807T235900Z
DTSTART:20260807T233000Z
LOCATION:AI Village - LVCCW Level 1 Hall 3 1104 (Creator Stage 4)
SUMMARY:AgentBreaker: A blind spot detector for your coding agents
END:VEVENT
END:VCALENDAR