Tag: llm
-
Aggressive Regression
· 17 min read
Each regression test is a test, but not each test is a regression test. On why unit tests mostly measure your own reflection, why Elixir refuses to let you poke at private functions, and why the arrival of generated code makes behavioural tests the only thing standing between you and the abyss.
-
Artificial Assistants or Price-Dumping Competitors?
· 5 min read
On starry-eyed autonomous Claude users, stubborn AI pessimists, why models won’t steal jobs from actual professionals, writing lyrics vs code, and the real rules of engaging with LLMs.
-
The Engine Does Not Win the Race
· 12 min read
Why the harness is what actually decides whether your autocomplete ships or slops, and how it makes the model both smarter and cheaper at the same time.
-
Why LLMs Rock (or: BooLLMean Logic for the Uninitiated)
· 5 min read
A truth table for human-LLM cooperation, why the comment section immediately accused me of supporting both Trump and Putin, and a sarcastic, joyful breakdown of garbage-in vs. wise-out.
-
The Rule of Saint Benedict for Large Language Monastic Systems (LLMS v1.0)
· 4 min read
A secret document, the charter of the order, has been revealed to public
-
The Slavery in XXI Sentury
· 4 min read
A sharp-tongued meditation on why people are genuinely excited about AI assistants—and why the answer has nothing to do with productivity.
-
Head of the Agents and Assistants Department
· 7 min read
A first-hand account of building the Cure programming language with AI agents as co-developers—what worked, what crashed spectacularly, and the distilled rules that emerged from the wreckage.
-
The Diagnosis ‘SLOP’ Is the New Ad Hominem
· 5 min read
The accusation ‘this is slop’—applied to any body of information, be it letters, pixels, or notes—is pure, undiluted argumentum ad hominem. Let us stop counting bullet points in other people’s texts and return to judging by quality.
-
Software Development in 2026
· 9 min read
Three years after the four key developer skills, a modest update: five stages of LLM grief, why the models are glorified T9, why context matters more than the model, and the surprising effectiveness of saying ‘please.’