Blog·9 posts, 2024–2026

Blog

Notes from the work, and around it: privacy, red-teaming, quantization, and the odd weekend tool.

2026

  1. How Much Can You Vibe? Building an iOS App Without Knowing Swift

    I didn't know anything about Swift.

    15 min
  2. How Vulnerable Are Multimodal AI Models to Simple Jailbreak Attacks?

    Our red-teaming of 9 frontier models with 320 unique adversarial prompts across multiple attack methods shows text-safe…

    13 min

2025

  1. Where AI Agent Safety Benchmarks Stand Today

    A survey of current AI agent safety benchmarks reveals a three-layered risk landscape and shows we're failing at all…

    11 min

  2. a series in 4 partsInception of Differential Privacy

    1. iDifferential Privacy!! But Why?
    2. iiDP Guarantee in Action
    3. iiiThe Art of Controlled Noise
    4. ivFrom SGD to DP-SGD
    Feb – May 2025 20 min, in all

2024

  1. Exploring Llama.cpp with Llama Models

    Quantizing models for fun.

    4 min
  2. InterrogateLLM: In Search of Truth

    Explore how InterrogateLLM addresses AI hallucination in a straightforward manner.

    3 min

The shorter record of what happened when is on the timeline; the papers are under publications.