April 2026

A practical guide to creating persistent memory for your AI programming assistant that remains intact across sessions. In April 2026, Andrej Karpathy introduced a deceptively simple concept: instead of relying on RAG or endless context windows, you should let an LLM maintain a structured Markdown wiki as a long-term knowledge base. You simply provide the raw data, and the model compiles, organizes, and updates the wiki on its own. You do not need to handle manual editing—the model […]

Ver mais

Fifty times global compute capacity, and a dystopian manual published in 1984. There’s a version of this article where I try to be balanced. Where I lay out “the pros and the cons” and offer some measured take that makes me sound thoughtful and considered. That version of this article can get fucked. Because the more time I spend with AI — building with it, watching what it’s doing to the world, understanding the trajectory we’re on — […]

Ver mais

Note: This article is written with assistance of the free version of Google Gemini AI quick mode.     TL;DR: Scroll to the bottom for the keyword “Summary”     As we know, COMEFROM is a sarcastic joke mocking the evilness of GOTO, and almost no seasoned professional software engineer will even consider to use it in production of any commercial codebase, so it’s only natural that nearly nobody will take this programming construct seriously. But what if I tell […]

Ver mais

The evolutionary origin of useless thoughts and the creative cost of thinking them A phone call can destroy your nervous system in under three seconds. You don’t decide to be afraid. You don’t reason your way into dread. You hear the news and cortisol is already flooding your bloodstream, heart rate spiking, the fine hairs on your forearm standing at attention. State is upstream of thought. Not the other way around. The first thought a human ever had […]

Ver mais

Zero-shot reasoning on text-rich networks (TRNs) remains a challenging frontier, as models must integrate textual semantics with relational structure without task-specific supervision. While graph neural networks rely on fixed label spaces and supervised objectives, recent large language model (LLM)-based approaches often overlook graph context or depend on distillation from larger models, limiting generalisation. We propose TRN-R1-Zero, a post-training framework for TRN reasoning trained solely via reinforcement learning. TRN-R1-Zero directly optimises base LLMs using a Neighbour-aware Group Relative Policy […]

Ver mais

A deeper look at the missing layer in modern systems The Problem You Don’t Notice – Until You Do Most systems don’t fail loudly. They don’t crash or burn or take down everything at once. Instead, they drift. Something small breaks. A constraint is violated. A user notices before the system does. You open the logs expecting answers. And the logs are there. Plenty of them. Requests came in, retries happened, downstream services responded, and eventually things succeeded. […]

Ver mais

Physical activity (PA) is widely recognized as protective against mental distress, yet whether this benefit varies systematically across population subgroups remains poorly understood. Using pooled data from ten consecutive annual waves of the U.S. Behavioral Risk Factor Surveillance System (2015-2024; n = 3,242,218), we investigate heterogeneity in the association between leisure-time PA and frequent mental distress (FMD, >=14 days/month) across age groups. Survey-weighted logistic regression reveals a striking age gradient: the adjusted odds ratio for PA ranges from […]

Ver mais