BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: ECDSA.fail: Beating Google’s Hidden Quantum Circui
 t with\n   Open\, Verifiable Autoresearch\n   Tags: Quantum Village | Crea
 tor Talk/Panel\n   When: Saturday\, Aug 8\, 15:00 - 15:59 PDT\n   Where: L
 VCCW Level 1 Hall 2 605 (Quantum Village) - [1]Map\n\n   Description:\n\n 
   In March 2026\, Google Quantum AI announced substantially lower\n   reso
 urce estimates for applying Shor’s algorithm to the 256-bit\n   elliptic
 -curve discrete-logarithm problem. The underlying logical\n   circuits wer
 e withheld\, but Google published zero-knowledge proofs\n   attesting to t
 wo secp256k1 point-addition circuits: a low-width\n   construction bounded
  by 1\,175 logical qubits and 2.7 million\n   non-Clifford operations\, an
 d a low-gate construction bounded by 1\,425\n   logical qubits and 2.1 mil
 lion non-Clifford operations. At Eigen Labs\,\n   we asked a very DEF CON 
 question. Can an implementation remain\n   meaningfully secret once its in
 terface\, verifier\, and resource bounds\n   are public?\n\n   We built EC
 DSA.fail\, an open autoresearch benchmark for the quantum\n   attacks on e
 lliptic-curve cryptography. The target is not an\n   end-to-end Bitcoin at
 tack\, each submission emits a circuit in\n   Google’s classically simul
 able kickmix model\, which supports\n   Clifford operations.\n\n   A separ
 ate trusted evaluator treats the emitted operation stream as\n   untrusted
  input. Valid circuits are scored by the product of their\n   logical-qubi
 t width and average executed CCX-plus-CCZ count. This\n   objective is not
  a complete physical-resource model\, but it exposes\n   the central optim
 ization tension between quantum memory and expensive\n   non-Clifford comp
 utation. The public repository\, reproducible\n   evaluator\, shared resea
 rch notes\, and continuously updated frontier\n   allow humans and autonom
 ous coding agents to inspect previous attempts\n   and immediately build o
 n successful results.\n\n   Within eight hours\, the open network matched 
 Google’s concealed\n   qubit–Toffoli bound. In roughly 72 hours\, it s
 urpassed it. This talk\n   will present the benchmark architecture\, the f
 inite-field and\n   uncomputation bottlenecks exposed by the search\, the 
 security\n   engineering required to make agent-generated circuits safely\
 n   executable. We will also examine a broader DEF CON question: when\n   
 research artifacts are machine-verifiable and optimization can be\n   mass
 ively parallelized across humans and agents\, how long can a hidden\n   al
 gorithmic advantage remain an advantage?\n\n   SpeakerBio:  Soubhik Deb\, 
 Eigen Labs\n\n   I am Head of Research at Eigen Labs. The central principl
 e of my work\n   over the last decade has been building open networks that
  maximize\n   human agency. In product terms\, this takes the form of thre
 e\n   properties that everything I have worked on has strived for:\n   Per
 missionless: Empower anyone to build and deploy artifacts —\n   software
 \, capital\, and the outputs of intelligence — at internet\n   scale\, w
 ithout asking permission. Attribution: Programmatic credit\n   assignment\
 , so that value generated and captured by those artifacts\n   flows back t
 o the contributions that produced them. Accountability:\n   Programmatic g
 uardrails against adversarial behavior — a risk that\n   is even more ac
 ute in open networks — with penalties proportional to\n   the harm cause
 d.\n\n   Links:\n       eigenlabs.org - [2]https://eigenlabs.org\n   '\n\n
    1. #LVCCW_Level1_Hall2\n   2. https://eigenlabs.org\n\n\n
DTEND:20260808T225900Z
DTSTART:20260808T220000Z
LOCATION:Quantum Village - LVCCW Level 1 Hall 2 605 (Quantum Village)
SUMMARY:ECDSA.fail: Beating Google’s Hidden Quantum Circuit with Open\, V
 erifiable Autoresearch
END:VEVENT
END:VCALENDAR