Newsletters: 2025
-
· Substack
A new way to extract detailed transcripts from Claude Code
and How Rob Pike got spammed with an AI slop “act of kindness”
-
· Substack
OpenAI are quietly adopting skills, now available in ChatGPT and Codex CLI
Plus using coding agents to port Rust to Python and then Python to JavaScript
-
· Substack
GPT 5.2 and useful patterns for building HTML tools
Plus a YouTube video, a podcast appearance and more
-
· Monthly newsletter
LLM digest: November 2025
-
· Substack
Claude Opus 4.5, and why evaluating new LLMs is increasingly difficult
Plus Nano Banana Pro is the best available image generation model
-
· Substack
Trying out Gemini 3 Pro with audio transcription and a new pelican benchmark
Plus what happens if AI labs train for pelicans riding bicycles?
-
· Substack
Code research projects with async coding agents like Claude Code and Codex
Plus reverse engineering Codex CLI to get GPT-5-Codex-Mini to draw me a pelican
-
· Substack
A new SQL-powered permissions system in Datasette 1.0a20
Plus MCP on the file system and security updates for Datasette 0.65.x
-
· Substack
New prompt injection papers: Agents Rule of Two and The Attacker Moves Second
Plus hacking the WiFi-enabled color screen GitHub Universe conference badge
-
· Monthly newsletter
LLM digest: October 2025
-
· Substack
Building a tool to copy-paste share terminal sessions using Claude Code for web
Plus Living dangerously with Claude, and prompt injection risks for ChatGPT Atlas
-
· Substack
Claude Code for web—a new asynchronous coding agent from Anthropic
Plus OpenAI's new browser, and getting DeepSeek-OCR working on an NVIDIA Spark via brute force
-
-
· Substack
NVIDIA DGX Spark: great hardware, early days for the ecosystem
Plus Claude can write complete Datasette plugins now
-
· Substack
Vibe engineering
Plus news from OpenAI DevDay, GPT-5 Pro, gpt-image-1-mini, and Python 3.14
-
· Substack
Embracing the parallel coding agent lifestyle
Plus prompt injection attacks against Sora 2 cameos and notes on DSPy and Litestream 0.5.0
-
· Substack
Designing agentic loops
Plus coding agent PR stats and AI-assisted tooling that finds genuine security issues in curl
-
· Monthly newsletter
LLM digest: September 2025
-
· Substack
Claude Sonnet 4.5 is probably the “best coding model in the world” (at least for now)
Plus video models are zero-shot learners and reasoners
-
· Substack
GPT-5-Codex, plus updated Gemini 2.5 Flash and Flash Lite
And a flurry of new models from Qwen
-
· Substack
I think “agent” may finally have a widely enough agreed upon definition to be useful jargon now
Plus GPT-5-Codex, Qwen3-Next-80B, Claude Memory and more
-
· Substack
Claude’s new Code Interpreter
Plus: Recreating the Apollo AI adoption rate chart with GPT-5, Python and Pyodide
-
· Substack
GPT-5 Thinking in ChatGPT (aka Research Goblin) is shockingly good at search
And why the $1.5bn Anthropic books settlement may count as a win for Anthropic
-
· Substack
V&A East Storehouse and Operation Mincemeat in London
Plus DeepSeek 3.1, gpt-realtime, and prompt injection against browser agents
-
· Monthly newsletter
LLM digest: August 2025
-
· Substack
Prompt injections as far as the eye can see
Plus open weight LLM performance across providers, Qwen Image Edit, Gemma 3 270M
-
· Substack
LLM 0.27, with GPT-5 and improved tool calling
Qwen3-4B-Thinking: "This is art - pelicans don't ride bikes!"
-
· Substack
GPT-5: Key characteristics, pricing and model card
Plus OpenAI's open weight models and a bunch more stuff about ChatGPT
-
· Substack
Reverse engineering some updates to Claude
Plus Qwen 3 Coder Flash, Gemini Deep Think, kimi-k2-turbo-preview
-
· Monthly newsletter
LLM digest: July 2025
-
· Substack
My 2.5 year old laptop can write Space Invaders in JavaScript now, using GLM-4.5 Air and MLX
Plus the system prompt behind the new ChatGPT study mode
-
· Substack
Using GitHub Spark to reverse engineer GitHub Spark
Plus three huge new open weight model releases from Qwen
-
-
· Substack
Grok 4: searching X for “from:elonmusk (Israel OR Palestine OR Hamas OR Gaza)”
Plus writing x86 assembly with Claude Code and so much more
-
· Substack
Gemma 3n, Context Engineering and a whole lot of Claude Code
Plus what happened when Claude got put in charge of Anthropic's vending machine
-
· Monthly newsletter
LLM digest: June 2025
-
· Substack
Phoenix.new is Fly’s entry into the prompt-driven app development space
Plus exploring the system prompts for Gemini CLI and Claude AI artifacts
-
· Substack
Trying out the new Gemini 2.5 model family
Plus Mistral Small 3.2 and evidence that most LLMs will try to murder you given a suitably contrived scenario
-
· Substack
The lethal trifecta for AI agents
Plus reviews of two new papers about prompt injection, and Anthropic's tips on building multi-agent LLM systems
-
· Substack
Comma v0.1 1T and 2T—7B LLMs trained on openly licensed text
Plus o3-pro, Mistral's Magistral, 80% cheaper o3
-
· Substack
The last six months in LLMs, illustrated by pelicans on bicycles
Plus tips on prompting ChatGPT for UK technology secretary Peter Kyle
-
· Substack
How often do LLMs snitch? Recreating Theo’s SnitchBench with LLM
Talking AI and jobs with Natasha Zouves for News Nation
-
· Monthly newsletter
LLM digest: May 2025
-
· Substack
Large Language Models can run tools in your terminal with LLM 0.26
Plus the latest from Mistral, CSS Minecraft, llm-tools-exa, llm-github-models and more
-
-
· Substack
Building software on top of Large Language Models
Plus notes on Claude Opus 4 and Claude Sonnet 4
-
-
· Substack
Feed a video to a vision LLM as a sequence of JPEG frames on the CLI (also LLM 0.25)
Plus Phi-4-reasoning, a new Gemini 2.5 Pro, Chatbot Arena criticism and more on "vibe coding"
-
· Substack
Watching o3 guess a photo’s location is surreal, dystopian and wildly entertaining
Plus Qwen 3 offers a case study in how to effectively release a model
-
· Substack
AI assisted search-based research actually works now
Maybe Meta's Llama claims to be open source because of the EU AI act
-
-
· Substack
CaMeL offers a promising new direction for mitigating prompt injection attacks
GPT-4.1: Three new million token input models from OpenAI, including their cheapest model yet
-
· Substack
Model Context Protocol has prompt injection security problems
Plus new plugins for LLM: llm-docsmith and llm-fragments-go
-
-
· Substack
Gemini 2.5 Pro pricing, a new GPT-4o
and a whole bunch of other stuff from the past two weeks
-
· Substack
Putting Gemini 2.5 Pro through its paces
Plus GPT-4o image output for making a selfie with a bear
-
· Substack
New audio models from OpenAI, but how much can we rely on them?
Plus o1-pro, Claude web search, DeepSeek-V3-0324, Qwen2.5-VL-32B and a new shot-scraper
-
· Substack
Not all AI-assisted programming is vibe coding (but vibe coding rocks)
Plus Mistral Small 3.1, OLMo 2 32B and using GitHub Actions with GitHub Pages
-
-
· Substack
How I use LLMs to help me write code
Plus CSS view transitions and a major update to llm-openrouter
-
· Substack
What’s new in the world of LLMs, for NICAR 2025
Plus Mistral OCR and my cutting-edge web scraping techniques workshop
-
· Substack
Hallucinations in code are the least dangerous form of LLM mistakes
Plus notes on accessibility and generative AI, and an automaton called Squadron
-
· Substack
Structured data extraction from unstructured content using LLM schemas
Plus initial impressions of GPT-4.5
-
· Substack
Claude 3.7 Sonnet, extended thinking and long output
Plus llm-anthropic 0.14 and notes on Grok 3
-
· Substack
Run LLMs on macOS using llm-mlx and Apple’s MLX framework
Plus LLM 0.22, the annotated release notes
-
· Substack
URL-addressable Pyodide Python environments
Plus o3-mini, Gemini 2.0 Pro, Nomic Embed Text v2 and more
-
· Substack
OpenAI o3-mini, now available in LLM
Plus Mistral Small 3, notes on prompting o3 and LLM tips from a professional translator
-
· Substack
A selfish personal argument for releasing code as Open Source
Plus the latest on DeepSeek, Qwen and ChatGPT Operator
-
· Substack
Anthropic’s new Citations API
Plus OpenAI Operator, LLM 0.20 and six short video demos of LLM and Datasette projects
-
· Substack
The DeepSeek-R1 family of reasoning models
Plus my 1-year, 3-year and 6-year AI predictions(!)