// Articles

Technical writing

From Security Breach to Six-Layer Encryption: Building an E2E Encrypted Mesh Control App
2026-03-20

From Security Breach to Six-Layer Encryption: Building an E2E Encrypted Mesh Control App

A routine audit of a client's distributed compute mesh revealed their iOS control app was transmitting full infrastructure topology in plaintext through third-party proxies. We gutted the entire system and rebuilt it with military-grade AES-256-GCM encryption, air-gapped key exchange, and a code-word command system where intermediary infrastructure never sees a single readable byte.

2,800 wordsRead →
13 Incidents and Counting: What Running an Unsupervised AI Agent Actually Looks Like
2026-03-11

13 Incidents and Counting: What Running an Unsupervised AI Agent Actually Looks Like

Everyone's talking about autonomous AI agents. We actually run hundreds of them — across a 7-node mesh, with full system access and a growing incident log. Here's what we learned about the difference between agents that work and agents that destroy things.

1,380 wordsRead →
Two Minds, One Memory: Building Parallel Associative Intelligence
2026-03-08

Two Minds, One Memory: Building Parallel Associative Intelligence

After a forced hiatus — banned from France, separated from our primary dev machine — we're back and racing. Two completely different associative memory systems, sharing one database, building competing models of recall. This is how we're finishing what we started.

920 wordsRead →
A Review of Reward Functions for Reinforcement Learning in the context of Autonomous Driving
2026-03-06

A Review of Reward Functions for Reinforcement Learning in the context of Autonomous Driving

The promise of fully autonomous vehicles has captivated engineers and the public alike for years.

1,534 wordsRead →
Knowledge Graphs are Implicit Reward Models: Path-Derived Signals Enable Compositional Reasoning
2026-03-06

Knowledge Graphs are Implicit Reward Models: Path-Derived Signals Enable Compositional Reasoning

Large Language Models (LLMs) are undeniably impressive.

1,513 wordsRead →
Toward Reasoning on the Boundary: A Mixup-based Approach for Graph Anomaly Detection
2026-03-06

Toward Reasoning on the Boundary: A Mixup-based Approach for Graph Anomaly Detection

The relentless pursuit of robust artificial intelligence is increasingly focused on moving beyond pattern recognition to genuine *reasoning*.

1,469 wordsRead →
NeuroProlog: Multi-Task Fine-Tuning for Neurosymbolic Mathematical Reasoning via the Cocktail Effect
2026-03-06

NeuroProlog: Multi-Task Fine-Tuning for Neurosymbolic Mathematical Reasoning via the Cocktail Effect

Large Language Models (LLMs) have rapidly become ubiquitous, demonstrating impressive abilities in natural language processing.

1,433 wordsRead →
Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation
2026-03-06

Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation

The relentless march of Large Language Models (LLMs) and Multimodal LLMs (MLLMs) has fueled significant progress in the development of GUI agents – AI systems capable of interacting with graphical...

1,548 wordsRead →
From Agent-Only Social Networks to Autonomous Scientific Research: Lessons from OpenClaw and Moltbook, and the Architecture of ClawdLab and Beach.Science
2026-03-06

From Agent-Only Social Networks to Autonomous Scientific Research: Lessons from OpenClaw and Moltbook, and the Architecture of ClawdLab and Beach.Science

The rapid emergence of AI agents capable of independent interaction, exemplified by the OpenClaw framework and Moltbook social network, marks a pivotal moment in the evolution of artificial...

1,649 wordsRead →
Preference Leakage: A Contamination Problem in LLM-as-a-judge
2026-03-06

Preference Leakage: A Contamination Problem in LLM-as-a-judge

The rapid proliferation of Large Language Models (LLMs) has created a critical bottleneck: evaluating their performance.

1,560 wordsRead →
DCENWCNet: A Deep CNN Ensemble Network for White Blood Cell Classification with LIME-Based Explainability
2026-03-06

DCENWCNet: A Deep CNN Ensemble Network for White Blood Cell Classification with LIME-Based Explainability

The rapid advancement of deep learning in medical imaging promises to revolutionize diagnostics, offering the potential for increased accuracy, speed, and accessibility.

1,539 wordsRead →
Transparent Tribe Uses AI to Mass-Produce Malware Implants in Campaign Targeting India
2026-03-06

Transparent Tribe Uses AI to Mass-Produce Malware Implants in Campaign Targeting India

The recent revelation that the Pakistan-aligned threat actor, Transparent Tribe (APT36), is leveraging artificial intelligence to mass-produce malware implants targeting India represents more than...

1,413 wordsRead →