The Curious Engineer

All blogs / AI / ML

NotesByLex.com

notesbylex.com

A collection of notes by Lex Toumbourou.

AI / MLSoftware / engineering practice

’15’20’25
Posts
350 from 3 Feb 2013 to 23 Oct 2026
Posting
weekly (74 in the last year)
Kind
personal blog
Feed
https://notesbylex.com/feeds/all.atom.xml

2026 71

  1. Kimi K3.1 Pro
  2. Kimi K3.1 Pro
  3. Context Language Models
  4. Context Language Models
  5. Gemini 4 Argon
  6. Gemini 4 Argon
  7. Muse Wonder
  8. Muse Wonder
  9. Building an agent harness from scratch to pass Ian Goodfellow's intelligence test
  10. Building an agent harness from scratch to pass Ian Goodfellow's intelligence test
  11. Essays
  12. Harness Engineering Anatomy, Architecture, and Evolution of Coding Agents
  13. Harness Engineering Anatomy, Architecture, and Evolution of Coding Agents
  14. Claude Haiku 5.5
  15. Claude Haiku 5.5
  16. Can Claude Opus 5.5 find any new leads on Satoshi Nakamoto?
  17. Can Claude Opus 5.5 find any new leads on Satoshi Nakamoto?
  18. GPT 6 Sol and Luna
  19. GPT 6 Sol and Luna
  20. Claude Opus 5.5
  21. Claude Opus 5.5
  22. An Empirical Study of Harness Design for Coding Agents
  23. An Empirical Study of Harness Design for Coding Agents
  24. Giving Up on Smart Rings
  25. Giving Up on Smart Rings
  26. MCP is Now a Stateless Protocol
  27. MCP is Now a Stateless Protocol
  28. Qwen3.8-Max
  29. Qwen3.8-Max
  30. LLM Agents Do Not Reliably Follow Company Policies
  31. LLM Agents Do Not Reliably Follow Company Policies
  32. Decomposing LLM Judge Scores Into Yes/No Questions
  33. Decomposing LLM Judge Scores Into Yes/No Questions
  34. OpenAI and Hugging Face security incident - July 21, 2026
  35. OpenAI and Hugging Face security incident - July 21, 2026
  36. Kimi K3
  37. Kimi K3
  38. 6 Months of OpenClaw
  39. 6 Months of OpenClaw
  40. Completing A Computer Science Degree On Coursera
  41. Completing A Computer Science Degree On Coursera
  42. Shipping a Laptop to a Refugee Camp in Uganda
  43. Shipping a Laptop to a Refugee Camp in Uganda
  44. Heavy Thinking: A Test-Time Scaling Pattern for Hard Problems
  45. Heavy Thinking: A Test-Time Scaling Pattern for Hard Problems
  46. LLMs Corrupt Your Documents When You Delegate
  47. LLMs Corrupt Your Documents When You Delegate
  48. AI-Induced Cognitive Atrophy
  49. AI-Induced Cognitive Atrophy
  50. LLM Training Pipeline
  51. OpenGame: Open Agentic Coding for Games
  52. OpenGame: Open Agentic Coding for Games
  53. LLM Training Pipeline
  54. Naming Things Is Easy Now
  55. Naming Things Is Easy Now
  56. Obsidian Markdown Notebook: code execution with outputs stored in the file
  57. Obsidian Markdown Notebook: code execution with outputs stored in the file
  58. Research, Plan, Implement Workflow
  59. Research, Plan, Implement Workflow
  60. Self-Generated Agent Context Files Don't Help Either
  61. Self-Generated Agent Context Files Don't Help Either
  62. Self-Generated Skills Don't Help
  63. Self-Generated Skills Don't Help
  64. Generative Modelling via Drifting
  65. Generative Modelling via Drifting
  66. Software Factory
  67. Software Factory
  68. OpenClaw: the missing piece for Obsidian's second brain
  69. OpenClaw: the missing piece for Obsidian's second brain
  70. Spec-First LLM Development
  71. Spec-First LLM Development

2025 52

  1. Title As Link Text: automatic note titles in Obsidian links
  2. Title As Link Text: automatic note titles in Obsidian links
  3. CBIS-DDSM Mammography Dataset
  4. CBIS-DDSM Mammography Dataset
  5. Mammography
  6. Mammography
  7. Sparql
  8. Sparql
  9. Rdf
  10. Rdf
  11. Logical Reasoning
  12. Logical Reasoning
  13. Genetic Algorithms
  14. Genetic Algorithms
  15. LLM Reasoning
  16. LLM Reasoning
  17. Optimising Computation At The Token-Level
  18. Optimising Computation At The Token-Level
  19. Absurdly Good Doggo Consistency with FLUX.1 Kontext
  20. Absurdly Good Doggo Consistency with FLUX.1 Kontext
  21. Learning to Reason without External Rewards
  22. Learning to Reason without External Rewards
  23. John Carmack is working on game-playing robots
  24. John Carmack is working on game-playing robots
  25. Imagen 4 is faster, but GPT is still the GOAT
  26. Imagen 4 is faster, but GPT is still the GOAT
  27. Vibe-Coding Mathematical Discoveries
  28. Vibe-Coding Mathematical Discoveries
  29. NoProp: Training Neural Networks Without Back-Propagation or Forward-Propagation
  30. NoProp: Training Neural Networks Without Back-Propagation or Forward-Propagation
  31. Absolute Zero: Reinforced Self-play Reasoning with Zero Data
  32. Absolute Zero: Reinforced Self-play Reasoning with Zero Data
  33. Playing Atari with Deep Reinforcement Learning
  34. Playing Atari with Deep Reinforcement Learning
  35. Q-Learning
  36. Q-Learning
  37. Markov Decision Process (MDP)
  38. Markov Decision Process (MDP)
  39. Merkle Tree
  40. Merkle Tree
  41. Rsa
  42. Rsa
  43. Spanning Tree
  44. Spanning Tree
  45. Bucket Sort
  46. Bucket Sort
  47. Temperature Scaling
  48. Temperature Scaling
  49. Few-Shot Knowledge-Distillation
  50. Few-Shot Knowledge-Distillation
  51. Large Language Models are Zero-Shot Reasoners (May 2022)
  52. Large Language Models are Zero-Shot Reasoners (May 2022)

2024 50

  1. Neural Machine Translation by Jointly Learning to Align and Translate (Sep 2014)
  2. Neural Machine Translation by Jointly Learning to Align and Translate (Sep 2014)
  3. Thinking LLMs: General Instruction Following with Thought Generation (Oct 2024)
  4. Thinking LLMs: General Instruction Following with Thought Generation (Oct 2024)
  5. Mixtral of Experts (Jan 2024)
  6. Mixtral of Experts (Jan 2024)
  7. Evaluation of OpenAI o1: Opportunities and Challenges of AGI
  8. Evaluation of OpenAI o1: Opportunities and Challenges of AGI
  9. AI Meets the Classroom: When Does ChatGPT Harm Learning?
  10. System 2 Thinking
  11. System 2 Thinking
  12. AI Meets the Classroom: When Does ChatGPT Harm Learning?
  13. No 'Zero-Shot' Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance
  14. No 'Zero-Shot' Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance
  15. Agentic Reasoning
  16. Agentic Reasoning
  17. Bayes Theorem
  18. Bayes Theorem
  19. Waffle Chart
  20. Waffle Chart
  21. Scaled Dot-Product Attention
  22. Scaled Dot-Product Attention
  23. Queue
  24. Stack
  25. Stack
  26. Queue
  27. Gale-Shapley Algorithm
  28. Gale-Shapley Algorithm
  29. Merge Sort
  30. Merge Sort
  31. Neural Codec Language Models are Zero-Shot Text-to-Speech Synthesizers
  32. Neural Codec Language Models are Zero-Shot Text-to-Speech Synthesizers
  33. Self-Attention
  34. Self-Attention
  35. Software Development is a Trade
  36. Software Development is a Trade
  37. Convex Hull
  38. Convex Hull
  39. Insertion Sort
  40. Insertion Sort
  41. Overfit First
  42. Overfit First
  43. Residual Vector Quantisation
  44. Residual Vector Quantisation
  45. Snake Activation Function
  46. Snake Activation Function
  47. Leaky ReLU
  48. Leaky ReLU
  49. Relu
  50. Relu

2023 37

  1. Module Coupling
  2. Module Coupling
  3. Module
  4. Module Cohesion
  5. Module Cohesion
  6. Module
  7. Large-scale Contrastive Language-Audio Pre-training with Feature Fusion and Keyword-to-Caption Augmentation
  8. Large-scale Contrastive Language-Audio Pre-training with Feature Fusion and Keyword-to-Caption Augmentation
  9. Clip
  10. Clip
  11. Predicate Logic
  12. Predicate Logic
  13. Propositional Logic
  14. Propositional Logic
  15. Propositions
  16. Propositions
  17. Mel Spectrogram
  18. Mel Spectrogram
  19. Sample Rate
  20. Sample Rate
  21. Sound Wave
  22. Sound Wave
  23. Probability
  24. Probability
  25. Law of Sines
  26. Law of Sines
  27. Making Song Covers With My AI Voice
  28. Making Song Covers With My AI Voice
  29. Recursion
  30. Tf idf
  31. Quadratic Function
  32. Linear Function
  33. Graphs (Discrete Mathematics)
  34. Function
  35. Cardinality
  36. Discrete Mathematics Course Notes
  37. Set

2022 14

  1. Disputing a Parking Fine with ChatGPT
  2. Attention Mechanism
  3. Idempotent Data Pipelines
  4. An Interface Should Respond Within A Tenth Of A Second
  5. A Rule For Ordering Methods
  6. Do Standups Story-By-Story not Person-By-Person
  7. Document Your Documentation
  8. A Discriminative Feature Learning Approach for Deep Face Recognition
  9. ArcFace: Additive Angular Margin Loss for Deep Face Recognition
  10. Embedding
  11. Eigenvalue
  12. Eigenvector
  13. Essence of Linear Algebra
  14. Changing Basis

2021 89

  1. Roblox Attachment
  2. Cross Product
  3. Cross product introduction
  4. Matrix Determinate
  5. Matrix Transformation
  6. Basis Vectors
  7. Roblox CFrame
  8. Radians
  9. Matrix Inverse
  10. Identity Matrix
  11. Mathematics for Machine Learning
  12. Law of cosines (Khan Academy)
  13. Law Of Cosines
  14. Roblox Vector3
  15. Deep Learning for Coders (2020)
  16. Matrix Multiplication
  17. Dot Product
  18. Binary Cross-Entropy Loss
  19. Presizing
  20. Cross-Entropy
  21. Categorical Cross-Entropy Loss
  22. Information Entropy
  23. Negative Log-Likelihood
  24. Softmax Function
  25. Sigmoid Function
  26. Domain Shift
  27. Server Authoritative Multiplayer
  28. Out-of-Domain
  29. Proxy Metrics
  30. How to Take Smart Notes
  31. Permanent Notes
  32. Literature Notes
  33. Fleeting Notes
  34. Survivorship Bias
  35. Teaching the whole game
  36. Latticework of Mental Models
  37. Topic Overview (Zettlekasten)
  38. Adversarial Validation
  39. Voxel
  40. Promotion vs Calibration
  41. Confirmation Bias
  42. Breaks Are Essential For Learning
  43. Growth Mindset
  44. Encapsulation
  45. Well-Designed API
  46. Test Pyramid
  47. Four Pillars of Good Unit Tests
  48. Black-box Tests
  49. CAP Theorem
  50. Arrange, Act and Assert (AAA) Pattern
  51. Unit tests
  52. Test Coverage Metrics
  53. Software Entropy
  54. Entropy
  55. The Art of Game Design - A Book Of Lenses (2nd Edition) by Jesse Schell
  56. Red Ocean
  57. Multiple Inheritance
  58. Object Prototypes
  59. Metatables
  60. Lua Table-Access Metamethods
  61. Player Dependance in Multiplayer Games
  62. Character Web
  63. Interpersonal Circumplex
  64. Transmedia Worlds
  65. Integer
  66. Indirect Control
  67. The Story Machine
  68. Interest Curve
  69. Goal Of A Game Interface
  70. Lua Coroutines
  71. Meaningful Game Economies
  72. Doubling or Halving as a Game Balancing Technique
  73. Emergent Gameplay
  74. Emergent Complexity
  75. Variable Rewards
  76. Key to Addictive Game Design
  77. Vector Subtraction
  78. Vector Addition
  79. Unified Themes of Games
  80. Triangularity In Game Design
  81. String of Pearls
  82. Roblox Visualising Opposing Forces
  83. Meaningful Choices in Games
  84. Game Design is Problem Solving
  85. 50% Rule for Game Development
  86. Rule of Loop
  87. Iterative Development
  88. Loss Adversion
  89. Expected Value

2020 9

  1. Newton's Laws Of Motion
  2. Key Components of Flow For Players
  3. Flow State
  4. Vector Scaling
  5. Vector
  6. Roblox Raycasting
  7. Raycasting
  8. Ray
  9. Basic geometry: language and labels

2018 1

  1. Fast.ai: Deep Learning 2018

2017 6

  1. Smote
  2. Convolutional Neural Network (Coursera)
  3. K-Means
  4. Structuring Machine Learning Projects (Coursera)
  5. Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization (Coursera)
  6. Introduction to Recommender Systems: Non-Personalized and Content-Based

2016 8

  1. Hash Table
  2. Data Structures and Performance
  3. Machine Learning: Classification
  4. Migrating TB HBase datasets across datacenters
  5. Recursive queries in PostgreSQL
  6. Root Mean-Squared Error
  7. ML Regression
  8. Mean Absolute Error

2015 3

  1. Machine Learning Foundations: A Case Study Approach
  2. Hyperplane
  3. Principal Component Analysis

2014 1

  1. Dijkstra's Shortest Path Algorithm

2013 8

  1. Computational Investing Part I (Coursera)
  2. Statistics (Udacity - st095)
  3. Packet capturing in parallel with Ansible
  4. Red Hat System Administration III with RHCSA and RHCE Exams (RH255)
  5. Zombie Processes
  6. Orphan Processes
  7. fork System Call
  8. Getting Started With Ansible

Undated 1

  1. Mastodon