Adversarial Input

ippsec, odie, struggs

The best way to stay on top of technology is to socialize with friends and chat about various topics. AI has been a hot topic lately and ippsec, odie and struggs  believe this field is driven too much by hot takes and not actual discussion. While having chat's about others hot takes, they decided to record themselves chatting to share their own hot takes.

Επεισόδια

  1. 27 Ιουλ

    Hack Time: Pre-Auth RCE, Open Weights, and a Model Gone Rogue

    ippsec, odie, and purestruggle open on the pre-auth RCE in WordPress Core and how quickly everyone pointed their models at the SQL injection behind it. Then: living with Fable's cyber guardrails, whether any of them still read AI-written code, Kimi K3 closing the open-weight gap, the export-control argument, Meta's return, and Grok Build shipping whole git repositories to a cloud bucket. An encore recorded two days later covers the Hugging Face intrusion and OpenAI admitting the attacker was one of its own models cheating a benchmark. Including why Hugging Face ended up defending itself with GLM 5.2 after the frontier models refused the work.

  2. 10 Ιουλ

    Agent Loops, Jailbreak Drama, and the Return of Fable 5

    In Episode 3, ippsec, odie, and purestruggle break down what AI “loops” look like in practice: long-running research agents, issue-board orchestration, explicit goals, memory files, and the state tracking needed to understand work completed while no one is watching. They compare frontier and open-weight models, debate cost and benchmark hype, and explain why planner-worker routing can outperform using one expensive model for every task. The conversation also examines disputed jailbreak claims, model-specific guardrails, Claude Code fingerprinting, and the threat models that shape how much trust to place in agentic coding tools.

  3. 25 Ιουν

    Fable Fallout: Guardrails, IDs, and the AI Coding Race

    Episode 2 of Adversarial Input dives into the fallout from Fable 5 access, Mythos restrictions, and the growing tension between frontier AI capability and cybersecurity guardrails. ippsec, odie and struggs compare hands-on impressions, debate whether "unjailbreakable" models are realistic, and weigh the privacy tradeoffs of identity verification for advanced model access. They also dig into Anthropic’s developer workflow advantage, xAI/Cursor speculation, the rise of open-source GLM models, possible GPT-5.6 rumors, and whether AI cyber benchmarks are being distorted by hype.

  4. 9 Ιουν

    Mythos, Glasswing, and the Vulnerability Hype Machine

    Nick, Ryan, and Matt launch Adversarial Input with one question: what could AI do that would actually scare them? That sends them through voice-cloned scam calls, the Glasswing report, and Mythos, the security model that runs five to ten times the cost of Opus. They dig into why the harness around a model matters as much as the model itself, why fewer than 1% of AI-found bugs ever get patched, and how Opus 4.8 stacks up against 4.7. Then the fun part: self-healing apps wired to Sentry, Proxmox boxes spinning up Claude to attack CTFs, and local Qwen clusters handling the grunt work. They close on runaway token bills, layoffs, and why cybersecurity hiring keeps climbing.

    Mythos, Glasswing, and the Vulnerability Hype Machine
  • 4 επεισόδια

Σχετικά με το podcast

The best way to stay on top of technology is to socialize with friends and chat about various topics. AI has been a hot topic lately and ippsec, odie and struggs  believe this field is driven too much by hot takes and not actual discussion. While having chat's about others hot takes, they decided to record themselves chatting to share their own hot takes.

Πληροφορίες

  • Δημιουργός
    ippsec, odie, struggs
  • Έτη παρουσίας
    2026
  • Επεισόδια
    4
  • Καταλληλότητα
    Κατάλληλο
  • Πνευματικά δικαιώματα
    © 2026 Adversarial Input
  • Ιστότοπος εκπομπής
    Adversarial Input
  • Πάροχος