100 Days of Making AI
Work
About
Day 100
A Birthday Gift to Myself
Day 99
Can We Trust AI Benchmarks? An Interdisciplinary Review of Current Issues in AI Evaluation
Day 98
MoralBench: Moral Evaluation of LLMs
Day 97
FVA-RAG: Falsification-Verification Alignment for Mitigating Sycophantic Hallucinations
Day 96
How Large Language Models Balance Internal Knowledge with User and Document Assertions
Day 95
When Helpfulness Becomes Sycophancy: Sycophancy is a Boundary Failure Between Social Alignment and Epistemic Integrity in Large Language Models
Day 94
Sycophantic AI makes human interaction feel more effortful and less satisfying over time
Day 93
Beyond the Black Box: Interpretability of Agentic AI Tool Use
Day 92
How AI Impacts Skill Formation
Day 91
Emergent Introspective Awareness in Large Language Models
Day 90
A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions
Day 89
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?
Day 88
Constitutional AI: Harmlessness from AI Feedback
Day 87
Hugging Face Demo
Day 86
Algorithmic sycophancy: A new source of systematic distortion in AI-driven biomedical research
Day 85
Activation Steering for Aligned Open-ended Generation without Sacrificing Coherence
Day 84
Alignment Whack-a-Mole : Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models
Day 83
pyFIES
Day 82
Verifying Chain-of-Thought Reasoning via Its Computational Graph
Day 81
Verbalizing LLMs’ assumptions to explain and control sycophancy
Day 80
ATLAS: Agent Tuning via Learning Critical Steps
Day 79
BEYOND BINARY REWARDS: TRAINING LMS TO REASON ABOUT THEIR UNCERTAINTY
Day 78
Sycophantic AI decreases prosocial intentions and promotes dependence
Day 77
A Rational Analysis of the Effects of Sycophantic AI
Day 76
Analyzing the Safety Pitfalls of Steering Vectors
Day 75
Mitigating Content Effects on Reasoning in LLMs Through Fine-Grained Activation Steering
Day 74
Can Activation Steering Generalize Across Languages?
Day 73
Extending Activation Steering to Broad Skills and Multiple Behaviours
Day 72
Improving Activation Steering in Language Models with Mean-Centering
Day 71
Aligning Large Language Models with Representation
Day 70
Who's asking? User personas and the mechanics of latent misalignment
Day 69
Personalized Steering of LLMs
Day 68
Programming Refusal with Conditional Activation Steering
Day 67
Sycophancy Shapes Multi-Agent Debate
Day 66
Truth Decay: Multi-Turn Sycophancy in LLMs
Day 65
Sycophancy in Multi-Turn Dialogues
Day 64
Too Polite to Disagree
Day 63
All About the FAO ESS
Day 62
Learning Transformer Fundamentals
Day 61
Detecting Implicit Reward Hacking by Measuring Reasoning Effort
Day 60
Building an SDR Radar
Day 59
Learning Token Probabilities
Day 58
A Unified Understanding and Evaluation of Steering Methods
Day 57
Auditing Language Models for Hidden Objectives
Day 56
Language Models (Mostly) Know What They Know
Day 55
Don’t lie to your friends
Day 54
Alignment Faking
Day 53
Semantic Uncertainty
Day 52
Reasoning Models Don't Always Say What They Think
Day 51
Concrete Problems in AI Safety
Day 50
50: A Look Back
Day 49
Auto CV Generator
Day 48
Agents Attack
Day 47
Learning AI Alignment
Day 46
Uganda PBS
Day 45
Forecasting Food Insecurity Part II
Day 44
Forecasting Commodity Prices (WB Pink Sheet)
Day 43
Alignment Network
Day 42
42 MCP
Day 41
What is famine?
Day 40
Data on Poverty, Hunger and Malnutrition in FTF Countries
Day 39
WASDE Forecasting
Day 38
AI + Creativity
Day 37
Iran Plumes
Day 36
Bert meets BERT
Day 35
RL for Beginners
Day 34
Slop Watch
Day 33
Breaking Words
Day 32
Neural Knots
Day 31
The Grand Orator
Day 30
The Geometry of Persona
Day 29
From Hedge to Answer: Uncertainty Dynamics in LLM Reasoning Chains
Day 28
Civic Firebrand
Day 27
SmartDublin Core
Day 26
Google Getter
Day 25
Fibonacci Cricut
Day 24
ASCII Librarian
Day 23
The Archetypical Librarian
Day 22
Forecasting EMS Demand in NYC w/ Two Decades of Data
Day 21
AIxiv — A preprint server for AI-generated research.
Day 20
Let's See the Guts!
Day 19
Floating Eyeballs
Day 18
RLHF Experiment Playground
Day 17
Feed the Future PBS RAG
Day 16
Jailbreaking Large Language Models
Day 15
chitty-chatty: LLM-to-LLM Conversation System
Day 14
baby-llm: Minimal GPT Training Script
Day 13
Urban Waste Dynamics of NYC
Day 12
Cookie Monster - Browser Privacy Dashboard
Day 11
Landfill Hunter
Day 10
INTJ: The Systems Forge Builder
Day 9
Steering the Librarian Persona
Day 8
xLibris Search Results Scraper
Day 7
Moltbook Human-Like Outlier Detection Pipeline
Day 6
Personaplex Speech-to-Speech Processor
Day 5
The arXiv Scout
Day 4
Snake Battles
Day 3
The Huggingface Daily Briefer
Day 2
Task-Oriented Delay Obliterator
Day 1
100 Days of Making AI Launch