Tag: llms
A Proposal for Voluntary AI Disclosure in OCaml Code
Proposing a voluntary, machine-readable AI content disclosure scheme for OCaml spanning opam packages, dune, and per-module attributes, aligned with the W3C AI Content Disclosure vocabulary.
machines will never understand language
I thought language was too complicated for machines to understand, until I got sick. a lifelong, meandering journey of parsing, learning, and fixing my broken body.
Of Monsters, Men, and Lawyers
Generalist LLMs are not lawyers, and evaluating them that way is a waste of time. Evaluating LLMs with useful specialized prompts (and eventually, with specialized legal harnesses) is where the work must happen.
.plan-26-02: Back from India and straight into the conservation conference
Hosting the Conservation Evidence conference at Pembroke, recovering from the India trip, and keeping up with LLM developments.
AoAH Day 25: Claude OCaml Marketplace for all your festive coding needs
Wrapping up 25 days of agentic coding with a Claude Code OCaml plugin marketplace to share the skills and tools developed throughout the series.
AoAH Day 24: Tuatara, an evolving Atom aggregator that mutates
Tuatara is a feed aggregator that integrates Claude to evolve and patch its own code when encountering parsing errors, embodying the concept of self-healing software.
AoAH Day 23: Unpac unifies git branching with package management
Introducing unpac, a tool that unifies git and package management into a single workflow where all code dependencies live in one repository as trackable branches.
AoAH Day 22: Assembling monorepos for agentic OCaml development
Materialising opam metadata into git submodules and monorepos, enabling cross-cutting fixes and unified odoc3 documentation across dozens of OCaml libraries.
AoAH Day 21: Complete dynamic HTML5 validation in OCaml and the browser
Porting the W3C's Nu HTML Validator from Java to OCaml and running in the browser dynamically
AoAH Day 20: Human language detection in native code, JS and wasm
Porting the Nu HTML Validator's language detection to OCaml, then optimizing from 115MB to 28MB and fixing WASM array limits for browser deployment.
AoAH Day 19: Zulip bot framework to bring Vicuna the friendly camel back
Building an OCaml Zulip bot framework with functional handlers, and pivoting from TOML to INI codecs for Python configparser compatibility
AoAH Day 18: TOML 1.1 codecs directly from the spec and paper
Building tomlt, a pure OCaml TOML 1.1 parser with bidirectional codecs following the jsont design patterns
AoAH Day 17: OCaml JMAP to plaster my painful email papercuts
Building an OCaml JMAP client that runs in browsers and CLI, then using it to build personalised email workflows for taming notification overload.
AoAH Day 16: Vibesplaining JSON Pointers using OCaml/Javascript
Building interactive OCaml tutorials that compile to JavaScript, using agents to generate executable documentation that teaches protocols like JSON Pointer while you code review.
AoAH Day 15: Porting a complete HTML5 parser and browser test suite
Vibespiling JustHTML from Python to pure OCaml, achieving 100% pass rate on the browser html5lib test suite using agentic workflows.
AoAH Day 14: Debugging a Karakeep CLI against the live service
Vibe coding an OCaml library for the Karakeep bookmarking service by giving an agent a live API key and letting it debug jsont codecs against the real service.
AoAH Day 13: Heckling an OCaml HTTP client from 50 implementations in 10 languages
Agentically synthesising a batteries-included OCaml HTTP client by gathering recommendations from fifty open-source implementations across JavaScript, Python, Java, Rust, Swift, Haskell, Go, C++, PHP and shell.
AoAH Day 12: Eio Connection pooling and event tracing
Building a TCP/TLS connection pooling library for Eio with DNS-based load balancing, stacked error handling, and self-contained HTML visualisations for stress test results.
AoAH Day 10: Building a TUI for Sortal using Mosaic
Building a simpler single-process terminal UI for Sortal using Mosaic's effects-based direct-style API, with Eio integration and discovering multimodal image debugging for terminal layouts.
AoAH Day 11: HTTP Cookies and vibing RFCs for breakfast
Synthesizing three RFC-compliant libraries (punycode, public-suffix, and cookeio) directly from Internet RFC specifications, establishing a workflow for automating standards implementation with proper cross-referencing to spec sections.
AoAH Day 9: Adding a Bonsai terminal UI to Sortal
Experimenting with OxCaml's bonsai_term framework for Sortal's terminal UI, navigating Eio-Async interoperability challenges through JSON-RPC while discovering image-based debugging techniques for terminal applications.
AoAH Day 8: Building a contacts CLI manager with Sortal
Creating Sortal, a CLI contacts management application using Yaml storage, XDG directories, Git-based synchronization, and integrating all previously built libraries into a cohesive CLI tool.
AoAH Day 7: Converting between JSON and Yaml with yamlt
Building yamlt to enable jsont codec definitions to work with both JSON and Yaml, providing data manipulation with location tracking and good error messages for both formats.
AoAH Day 6: Getting a Yaml 1.2 implementation in pure OCaml
Implementing a pure OCaml Yaml 1.2 parser using bytesrw by synthesizing from the specification and existing C library behavior, passing thousands of test suite cases while being 20% faster than the C-based implementation.
AoAH Day 5: Bytesrw Eio adapters and automating opam metadata
Building Bytesrw-Eio adapters for composable byte stream I/O while discovering Claude Skills as a powerful way to automate opam package metadata management through reusable workflow templates.
AoAH Day 4: Going recursive with Claudeio for Claude
Creating OCaml bindings for the Claude API using Eio and jsont codecs by reverse-engineering the JSON-RPC protocol from Python and Go SDKs, enabling Claude to write more Claude-powered OCaml code.
AoAH Day 3: XDG filesystem paths using Eio capabilities
Building an XDG Base Directory Specification library with Eio capabilities and Cmdliner integration, providing sandboxed filesystem access patterns with full environment variable and CLI override support.
AoAH Day 2: Building an OCaml JSONFeed library
Implementing a JSONFeed specification library using jsont codecs, discovering how Claude can automate the construction of complex combinators from prose specifications with excellent error messages.
2025 Advent of Agentic Humps: Building a useful O(x)Caml library every day
An exploration of agentic programming through building useful OCaml libraries daily using Claude Code while establishing groundrules for responsible development.
The AI French Connection to the Practice of Science
Reflections on the Franco-British AI collaboration workshops exploring how AI is transforming scientific practice, plus follow-up funding for the Conservation Copilot project.
AoAH Day 1: Building a Base32 Crockford library in OCaml
Building a Base32 Crockford encoding library in OCaml using Claude Code, establishing the development workflow with sandboxed Docker containers and local development environments.
A Triple-Negative Pretzel
Even LLMs struggle with triple-negatives.
Grokked & Pilfered
Elon Musk’s new Wikipedia clone has been criticized for nicking Wikipedia. Something else it also does: Aggressively cite the Tedium archive.
Presenting our Ecology of the Internet ideas at Aarhus 2025
Presentation at Aarhus 2025 on Internet ecology, proposing AI-driven software diversity to fight protocol ossification and create more resilient networks.
Tracking locations with OwnTracks, Life Cycle and Home Assistant
Setting up self-hosted location tracking using OwnTracks and reverse engineering Life Cycle app data with Claude Code for field work in Botswana.
The Game Genie Generation
For an unlicensed game accessory, the Game Genie sure casts a long shadow. It reshaped the games we already owned—and had a profound effect on copyright law.
Cresting the OCaml AI humps
Community efforts to improve agentic coding experience for OCaml including MCP libraries, opam embeddings, and tooling improvements.
Is AI poisoning the scientific literature? Our comment in Nature
Nature comment on AI-generated paper threats to evidence synthesis proposing federated living evidence databases with human-in-loop review.
Explaining The Vibes
AI gets a lot of hate these days, and it often frustrates me too, but let’s be clear about what it can realistically do. Here’s my attempt to explain by example.
New preprint survey on energy-aware deep learning on embedded hardware
Survey paper on energy-aware approaches for optimizing deep learning training and inference on embedded devices.
Out-of-the-box LLMs are not ready for conservation decision making
PLOS One publication showing pretrained LLMs perform poorly on conservation questions but improve dramatically with Conservation Evidence database training.
The ChatGPT Cheating Crisis Was Inevitable
Once again I find myself writing about the latest AI outrage cycle, this one sparked by an NYMag article about the rampant abuse of ChatGPT in higher education. It should surprise no one that kids are using AI to cheat constantly and we are about to see a generation of functionally illiterate college graduates hit the workforce. Seems bad! I feel obligated to point out that the article is a more than a little sensationalist and references a study claiming that 90% of students are using ChatGPT
Just Because It's Useful Doesn't Mean It's Good
Independent of yesterday's dive into a "useful" application of AI, discourse erupted on Bluesky over Hank Green saying he didn't think the "useless" critique holds water anymore. For the most part people are very angry at the perception of ceding any rhetorical ground to Big AI, which is a fair position to take. Green also put it in a fairly condescending "just asking questions" kind of way, which doesn't necessarily help his case. I don't love the phrasing but there is something to his point: p
Research Models and the Future of Search
In my self-appointed role as moral and practical judge of artificial intelligence and its uses I have rarely come across a product that I didn't end up finding boring or uninteresting after its novelty wore off. Most base or chatbot models are unreliable. Code models can be nice for tabbed autocomplete or boilerplate but I usually don't find them very helpful in the context of the large, complex, and often legacy codebases I work with professionally. I believe people who say they've found ways t
The Death of Learning
I've been thinking more about vibe coding lately and getting pissed about the whole concept. I think there's something deeply wrong with the mindset behind "prompt, don't read the code, don't think about it." I have written in the past about how artificial intelligence is a tool that accelerates the worst impulses of capitalism, and vibe coding is the latest iteration on what is possibly the one I hate the most: sacrificing understanding for efficiency. Here's the thing. I like learning. I know