Author

Owen Park

AI security researcher

Covers AI application security, model risk, and agentic system abuse.

ai-securityresearchapplication-security

Coverage by Owen Park (28)

Claude Max 100 versus 200 Claude Code plan usage comparison for developers

Claude Max $100 vs $200: Which Plan Gives Developers More Real-World Claude Code Usage?

Claude Max 5x is the better-value plan for many solo developers, while Max 20x gives heavier Claude Code users more room for long sessions, large repositories and agentic workflows.

ChatGPT Pro 100 versus 200 Codex plan usage comparison for developers

AI Coding Tools AI Security

ChatGPT Pro $100 vs $200: Which OpenAI Codex Plan Gives Developers More Real-World Usage?

OpenAI's $100 Pro plan is the best short-term Codex value for many solo developers, while the $200 Pro plan remains the safer choice for continuous large-repo, cloud-task and code-review workflows.

LiteLLM CVE-2026-42208 AI gateway SQL injection incident map

AI Gateway Security AI Security

LiteLLM CVE-2026-42208: Critical Pre-Auth SQL Injection Exposes AI Gateway Secrets

Critical LiteLLM CVE-2026-42208 pre-auth SQL injection exposes AI gateway databases, virtual keys and upstream model-provider credentials.

Ollama Cloud slowdown inference queue and tokens-per-second performance diagram

AI Cloud AI Security

Ollama Cloud Slowdown Frustrates Users as Low Token Rates Raise Questions About Hosted Open-Model Economics

User-reported token rates as low as 8 tokens per second show why Ollama Cloud needs clearer performance metrics, queue visibility and paid-tier expectations.

Desks covered

AI Security Research

Latest

Sep 12, 2026, 12:37 PM EDT · Research · Mlx12 minMLX vs llama.cpp on Apple Silicon: Which Local LLM Runtime to Actually Use

One is a portable inference engine, the other an array framework that also trains — and that single difference settles almost every practical question.

Sep 12, 2026, 12:37 PM EDT · Operations · Llm Evaluation14 minHow to Build an LLM Evaluation Harness That Catches Real Regressions

Public benchmarks only shortlist a model; your own cases decide whether to ship it. Here is the dataset, the scorers and the CI gates that catch real drift.

Sep 12, 2026, 12:37 PM EDT · Research · Prompt Caching11 minPrompt Caching Explained: Cutting Agent Costs Without Touching the Model

Your agent re-sends the same long prefix on every step and pays to recompute it. Here is the KV mechanism, the ordering rule it forces, and the break-even math.

Sep 12, 2026, 12:37 PM EDT · Research · Speculative Decoding11 minSpeculative Decoding and Multi-Token Prediction: How LLMs Generate Faster

Decode is memory-bound, so checking several tokens in one pass is nearly free. Here is the mechanism, the arithmetic behind it, and why quality never moves.