MCP is Now a Stateless Protocol
A new MCP specification that removes handshakes and mandatory sessions.
NotesByLex.com
Latest posts
Qwen3.8-Max
A new frontier model from the Qwen team
LLM Agents Do Not Reliably Follow Company Policies
A new benchmark that tests whether AI agents can follow long company policies during realistic work
Decomposing LLM Judge Scores Into Yes/No Questions
An LLM-judge approach that brings interpretability and actionability to your scores.
OpenAI and Hugging Face security incident - July 21, 2026
Ultimate benchmaxxing: hacking Hugging Face for test answers.
Kimi K3
Moonshot AI's 2.8T open-weight frontier model for coding and knowledge work.
6 Months of OpenClaw
Macros, workouts and life admin.
Completing A Computer Science Degree On Coursera
A short write-up on my experience completing a Computer Science BSc from my bedroom.
Shipping a Laptop to a Refugee Camp in Uganda
The surprisingly difficult ordeal of sending a decade-old MacBook to a friend in a Ugandan refugee camp.
Heavy Thinking: A Test-Time Scaling Pattern for Hard Problems
Now we have GPT Pro at home
LLMs Corrupt Your Documents When You Delegate
A large-scale study on long-horizon document tasks.
AI-Induced Cognitive Atrophy
We probably need to do deliberate daily mental exercise
OpenGame: Open Agentic Coding for Games
An agentic framework for end-to-end game creation
Naming Things Is Easy Now
I guess there is only one hard thing left in Computer Science?
Obsidian Markdown Notebook: code execution with outputs stored in the file
A Jupyter Notebook-style Obsidian plugin that runs code in your notes and stores the outputs directly in the Markdown file.
Research, Plan, Implement Workflow
An approach to agentic software development that I use
Self-Generated Agent Context Files Don't Help Either
Self-generated agent context files don't help.
Self-Generated Skills Don't Help
Curated skills boost agent performance by 16 points; self-generated ones don't help at all.
Generative Modelling via Drifting
A new paradigm for single-step generative modelling
Software Factory
is verification the future of software engineering?
OpenClaw: the missing piece for Obsidian's second brain
how I integrate OpenClaw into my Obsidian vault
Spec-First LLM Development
in which the LLM maintains a spec file alongside the project
Title As Link Text: automatic note titles in Obsidian links
An Obsidian plugin that automatically replaces filenames in your links with the actual note title, supporting both Wikilinks and Markdown links.
CBIS-DDSM Mammography Dataset
ImageNet for Mammography
Mammography
A breast cancer screening method
SPARQL
a query language for RDF databases
RDF
A graph-based data model for the Semantic Web
Genetic Algorithms
an optimisation technique inspired by natural selection
Optimising Computation At The Token-Level
Optimising computation at the token-level
Absurdly Good Doggo Consistency with FLUX.1 Kontext
Experiments with multi-turn character consistent editing
Learning to Reason without External Rewards
aka Self-Confidence is All You Need
John Carmack is working on game-playing robots
on John Carmack's Upperbound 25 Talk Notes
Imagen 4 is faster, but GPT is still the GOAT
a few comparisons of Google's Imagen 4 vs OpenAI's gpt-image-1
Vibe-Coding Mathematical Discoveries
Using evolutionary algorithms with LLM-coding agents
NoProp: Training Neural Networks Without Back-Propagation or Forward-Propagation
an alternative training method to backprop that does local layer learning
Absolute Zero: Reinforced Self-play Reasoning with Zero Data
learn to reason without any human-annotated data.
Playing Atari with Deep Reinforcement Learning
a classic paper applying neural networks to RL for game playing
Q-Learning
a reinforcement learning algorithm for finding optimal policies
Markov Decision Process (MDP)
A mathematical framework for modelling decision-making under uncertainty
Merkle Tree
a data structure where each node contains the hash of its child nodes
RSA
a public-key encryption system reliant on the practical difficulty of factorising large numbers
Spanning Tree
a sub graph of a connected graph that contains all vertices, but no cycles
Bucket Sort
a distribution-based sorting algorithm that works by dividing elements into buckets
Temperature Scaling
a parameter that controls how confident Softmax predictions are
Few-Shot Knowledge-Distillation
Routes LLM tasks to cheaper or more powerful models based on task novelty.
Large Language Models are Zero-Shot Reasoners (May 2022)
improve zero-shot prompt performance of LLMs by adding “Let’s think step by step” before each answer
Neural Machine Translation by Jointly Learning to Align and Translate (Sep 2014)
improve the Encoder/Decoder alignment with an Attention Mechanism
Thinking LLMs: General Instruction Following with Thought Generation (Oct 2024)
a prompting and fine-tuning method that enables LLMs to engage in a "thinking" process before generating responses
Mixtral of Experts (Jan 2024)
a Sparse Mixture of Experts (SMoE) language model
Evaluation of OpenAI o1: Opportunities and Challenges of AGI
a comprehensive evaluation of o1-preview across many tasks and domains.