Does financial AI developed primarily on US/EU data systematically misprice African credit risk? Evidence from SA lending markets
The deeper problem is that credit AI models were never really designed to price African credit risk at all. They were built on the assumption that creditworthiness leaves a formal paper trail. In South Africa, that assumption fails for a large majority of the adult population, and the consequences range from exclusion to predatory targeting.
SOUL.md: the file that makes your AI agent actually yours
The NousResearch Hermes agent has a file called SOUL.md. It doesn't live in the repository; it lives in ~/.hermes/. It's a user file, not a developer file. And it's the most interesting thing about how the agent is designed.
96.6% recall with no API keys: MemPalace stores AI context locally
Every LLM session you've ever had is probably gone. MemPalace is a local-first AI memory system that stores content verbatim, indexes it in a structured hierarchy, and retrieves it with 96.6% recall at R@5 on LongMemEval, running entirely on your local machine with no API key required.
The self-healing skill: AI that learns between sessions
The self-healing skill treats AI as a metacognitive system: one that observes its own behavior, curates what it learns, and turns repeated workflows into reusable knowledge. Here is how the loop actually works.
Your AI agent is winging it. Skills are how you fix that.
AI models are good at improvising. That's both the feature and the problem. Agent skills are reusable definitions that teach an AI agent how to handle a specific kind of task, consistently, according to rules you set.
Systems Engineering: Building Agentic Software That Works
Most agentic apps fail not because of bad models, but bad system design. Here's the five-layer framework that actually works.
Green Light for Hate: Leadership Failure and Anti-LGBTQ Violence in Namibia
Namibia presents a constitutional contradiction that has become a crisis. Courts have progressively extended rights to LGBTQ persons, while the legislature has enacted laws criminalising queer visibility. In this space, six queer Namibians were killed in 2023–2024. Not one death could be prosecuted as a hate crime.
South Africa's National AI Policy Framework: A Blueprint for Responsible Innovation
In October 2024, the Department of Communications and Digital Technologies published South Africa's first formal national AI policy. Not legislation, not a finished product: a strategic foundation and an opening position in a longer, more contested conversation about how this country governs one of the most consequential technologies of our time.
Method
The standard behind the sentences.
The editorial rules I use to keep professional writing useful, specific, and clear.
Clarity over cleverness
Good professional writing strips away jargon and says precisely what it means.
Every piece starts with the question: what does the reader need to understand, and what is the clearest path to that understanding? Style is earned after substance.
Interdisciplinary framing
The most useful analysis draws from more than one domain.
Connecting finance to legal structure, or legal reasoning to AI capability, produces insights that single-domain specialists cannot generate. Breadth is a deliberate choice.
Specificity as credibility
Vague claims erode trust. Specific claims build it.
Citing an exact rate decision, a named regulatory instrument, or a concrete data point signals that the author has done the work. Generalities are a form of intellectual laziness.
Opinion, not performance
Taking a position is more useful than surveying all positions.
Readers benefit from a reasoned view, not a comprehensive list of perspectives. The goal is to sharpen thinking, not to demonstrate exhaustiveness.