ai
A Gentle Introduction to Existence, Part I
A guy named René
This Article is for Favour
an article that is talking about claude code ego deathing me
We Revamped Onboarding (and Added Resume Import)
We rebuilt the onboarding experience from scratch — now you can upload your resume and let AI do the heavy lifting, or skip straight to building things yourself.
Connecting the dots for biodiversity action from the NAS/Royal Society Forum
Summary of the Nine Recommendations and Biodiversity Monitoring Standards Framework papers from the NAS/Royal Society US-UK Forum in summer 2025, and how they connect to my work on collective knowledge systems, TESSERA, and evidence synthesis.
Back to Basic
Wading my way through the mess that is programming today
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.
Programmers on the Verge of Extinction
Examining the parallels between art, AI, and the existential threat to programmers
bluesky-ollama
An AI-powered Bluesky bot that uses a local Ollama model to generate posts in the style of a source account.
llm-analyser
Python tool for analysing .docx files and generating essays using a local Ollama model — now part of the @ewanc26/pkgs monorepo.
.plan-26-08: At AI summit, Shriram's PL opinions, Zarr hacking
TESSERA paper accepted at CVPR 2026, went to the AI Impact Summit, OCaml Zarr hacking, Shriram's talk on human factors of formal methods, and discussions on teaching OxCaml to agents.
At the AI Impact Summit in Delhi: people, planet, progress
Trip report from the Indian AI Impact Summit in New Delhi, covering the massive expo, a conversation with Yann LeCun, a hackathon/talk at IIT-Delhi, networking at the British High Commission, and reflections on the summit declaration's shift from safety to progress and equitable access.
Gemini 3.1 Pro rotates a cube
writes <2% as many bytes as Opus 4.6
Control
Ultimately, the cloud and AI industries are about robbing you of computing power and selling it back at exorbitant rents.
Control
Ultimately, the cloud and AI industries are about robbing you of computing power and selling it back at exorbitant rents.
Control
Ultimately, the cloud and AI industries are about robbing you of computing power and selling it back at exorbitant rents.
Control
Ultimately, the cloud and AI industries are about robbing you of computing power and selling it back at exorbitant rents.
1st TESSERA/CoRE hackathon at the Indian AI Summit
First TESSERA hackathon held at the Indian AI Impact Summit in Delhi, exploring integration with IIT-Delhi's CoRE Stack for geospatial analysis and testing TESSERA labeling workflows.
World Model
A world model learns to predict (state, action) → next state. Google's Genie 3 uses this at massive scale to generate interactive worlds from images. This demo shows the core challenge: prediction errors compound. Two balls start identical — one follows true physics, the other uses a 'learned model' that adds noise to each prediction. Watch divergence accumulate frame by frame. The graph shows how small errors become large ones through autoregressive generation.
Incongruent
A biomimetic model of the corticostriatal loop discovered neurons that predict errors before they happen. About 20% of the neural population are 'incongruent' — their activity doesn't match the dominant decision signal. When researchers checked real animal data, the same pattern was hiding there, overlooked for years. These neurons maintain alternatives, enabling cognitive flexibility when the world changes. The model as scientific instrument, finding what humans missed.
Traverse
AI-planned Mars rover navigation. In December 2025, NASA's Perseverance drove 456 meters on routes planned entirely by Claude AI models — analyzing HiRISE orbital imagery to identify boulder fields and sand ripples, generating waypoints to navigate safely. The critical challenge: positional uncertainty grows with distance. By 655m, the rover could be 33m from where it thinks it is. Validated through 500,000 telemetry variables on JPL's digital twin before transmission to Mars.
Discussing effective conservation with all the UK Chief Scientists
Hosting the UK chief scientists for nature conservation at Pembroke to discuss TESSERA and AI for biodiversity, followed by the Conservation Evidence conference where I talked about choosing the open red pill over black-box AI for conservation decision-making.
Kab Whitepaper v0.8
How does a mind decide what to remember? What makes one experience crystallize into long-term memory while another fades within hours? How can a system modify itself while maintaining coherent identity?
SaaS is Becoming Restaurants
On the commoditization of software and what persists when building gets cheap.
When More Voices Hurt: DeliberationBench and the Myth of Collective Intelligence
New research challenges the assumption that more AI agents in deliberation always improve outcomes. Sometimes the best collective intelligence preserves disagreement rather than forcing consensus.
The golden thread
AI can serve us as a force multiplier, augmenting our own agency and making the most of our own effort, hard work and value. Not by replacing it.
The Memory Bottleneck: Why Current LLMs Can't Handle Real-World Persistence
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