Silicon in Irish (Part 4): Benchmark v2
Gemini 3 Flash wins the expanded 1,341-test Irish grammar benchmark at 72.4% accuracy
Posts tagged “ai”
Clear filterFilter by tag:
abstractionsaiantifragileautomationbenchmarkbrandingbrowser-extensionbruce-leecatalinachatgptcodecognitive-sciencecreativitycritical-thinkingdiglotdiglot-weave-methodduolingoeducationexpofrictionfuture-of-workgaeilgegeminihistoryirishirish-languageknowledge-managementlanguagelanguage-learninglearninglearningsllmllmsluxurymacosmental-modelsnlpnode.jsopenaiorganizational-designpaul-grahamplasmoproductivityprompt-engineeringreactreadingreflectionrobotsself-improvementsoftware-developmentstartupstastetechtechnologythoughtstiktoktranslationui/uxweb-developmentweevewriting
Silicon in Irish (Part 3): An Interlude
Comparing three independent studies on LLM Irish grammar competence - all hitting the same ~70% ceiling
SOPs and the Art of Showing Other People
Anthropic recently released Claude Skills—a framework for packaging reusable workflows that Claude can execute on-demand. Think of them as Standard Operating Procedures for AI agents.
The DAVID Framework for AI Evaluations
A Framework for AI Evals That Actually Works
Mines and Rabbit Holes #2
Welcome to Mines and Rabbit Holes #2. It is great to say that there is indeed a follow up and I hope you enjoy!
AI and Mental Masturbation
AI feels like a real conversation. It isn’t.
Why Structured Training Will Triumph Over AI Edutainment
You can't be both the student and the teacher
Gaeilge
Cén Ghaeilge ar Clanker?
The word of the year. Focail na mbliana.
Mines and Rabbit Holes
#1?
Luxury as the Anti-Algorithm
Harnessing Human Imperfection and Bold Vision in an AI-Driven World
The Soon-to-be Obsolete Skill of Prompt Engineering
300k -> 0k
Train the Model, Train the Team
Why Documentation is the New Superpower
The Blinking Cursor
Friend or Foe? Neither: Frenemy.
ChatGPT Is the New MS-DOS
We Haven’t had the Windows 95 Moment Yet
The Cognitive Collapse
How the Memory Paradox and AI Slide Are Undermining Human Intelligence
Good Taste Isn't Autocorrect
The Great Cognitive Outsourcing
The New Terrain of Critical Thinking
Silicon in Irish (Part 2)
Un-reason-LLM-bly good?
The Pyramid is Dead. Long Live the Diamond.
The Traditional Org Chart is Being Toppled by AI
Curiosity and The Compression Imperative
Why We Hunt for Better Metaphors
Silicon in Irish (Part 1)
AI's Unexpected Fluency in Irish
Leaky Abstractions Redux
Your AI Coding Assistant Keeps You Dry—Until It Doesn't
Stairs, Slides, and Ladders
Navigating the AI Training Gap
Re-Eval Your Life
A Holiday Guide to Better Questions
an ode to code
programming languages >> natural languages
The Kaleidoscope Hypothesis for Education
Abstracting and synthesising our way in and out of novel problems
Thinking More, Faster
Approaching the Speed of Thought for Content Creation
Humans Only; AI Need Not Apply
AI aids development; but, without the human touch, it will never be accessible to all.
Honey, The AI Shrunk the Language Divide
AI can be the great variance amplifier for minority languages
Does Language Learning have to be like Tiktok?
You can escape the attention-mining rush and learn faster simply by reading what you like.
Anyone in Need of Assistants?
DevDay for OpenAI brings AI agents to the masses
Functional Filtering with ChatGPT
LLMs still going brrr thanks to timesaving with JSON outputs