BMJ Cloudflare Hackathon It’s been a pretty wild 36 hours. Yesterday at about 10 am some of our dev team at BMJ started a hackathon with cloudflare that finished up yesterday at about 5pm. My own hack involved creating a corpus of RAG vectors and that process ran overnight, so I finished off some of my ideas through the course of today...

https://caisc2026.github.io/index.html - “we propose a new kind of academic conference where AI systems are explicitly recognized as primary contributors, both as authors and reviewers.” Connected to this there is an interesting looking webinar taking place on Saturday - https://luma.com/l73btv9q.

This is a very interesting article - https://ciphertalk.substack.com/p/nobody-knows-what-a-used-gpu-cluster, for me the most interesting observation is that data centres do not run themselves, they are only value creating when there is an ops team, but the markets are currently only pricing on the chips in the centres and not the opex ...

This is just a link to this very excellent writeup of a researcher using Claude science - https://www.linkedin.com/pulse/first-impressions-claude-science-christopher-thorpe-h0lie/ Bottom line is that Claude science does what it says it does, it helps science get done faster.

I like this bug writeup - https://aresluna.org/the-curious-case-of-the-disappearing-polish-s/ because it shows that code is a cultural artefact, and decisions made today embed many cultural norms, often hidden from us in the moment of making that decision.

This was pointed out by Rob Johnson - dashboards showing information on UNESCO open science initiatives. https://open-science.unesco.org/dashboards/global The biggest challenge is lack of funding: It's based on survey data. It's interesting to see this information, but how is it being used to drive change? That's not totally clear.

She has a ton of images of test tubes in boxes, and she needs to create csv files that note the location of the tables in their respective cells that the tubes are stored in. Schleppy work. Step in GPT Codex and Claude. After not too many iterations I had the AI create the following app for the work that she needs to do. And I'm deligh...

Dryad is an essential piece of open science infrastructure, but what we think about it, and what others think about it can sometimes be different things. If you have a moment then please take this survey, it will really help the organisation get an understanding of how it is perceived. You can access the survey here https://dryad.typef...

A short post - https://the-decoder.com/landmark-german-ruling-declares-googles-ai-overviews-are-googles-own-words-and-makes-it-liable-for-false-answers/ - Munich court holds Google liable for errors in AI summaries. In a way that search results do not all foul of. Expect huge push back from Google and the other tech giants, but the log...

Carnage Mellon Librarians created an online tool to roughly compare energy use of an LLM query vs a google search query - https://cmu-lib.github.io/ai-sustainability/.

openRxiv just announced a program for structured experimentation - https://openrxiv.org/openrxiv-labs-launch/ Of note the first experiment they have announced will be in partnership with https://curvenote.com/ Curvenote received YC-backing and raised a modest amount of initial funding too (https://curvenote.com/news/curvenote-seed-roun...

Every year we recruit new folk into the board at Dryad, I've been a board member now for I think nearly five years, and it's one of the most rewarding things that I do. Managing and supporting open scholarly research is really tricky, You can read the announcement here https://blog.datadryad.org/2026/05/18/call-for-applications-to-the-...

Why science is actually not all that like software engineering any more, and why it might have been a little bit similar in the past, but less so now. I used to say peer review in science was a bit like code review in software development, in that you have a community of professionals sense checking each other works through a gated pro...

Have you ever tried to travel in time, at least faster than your current light curve is flowing at? Many years ago now when I was at eLife we tried to do some future mapping. The senior team all got into a room and we had a washing line fixed up along the length of the room. The task was to predict what the organisation would need to d...

This is a blog post I've been looking forward to writing. Just for the delight of it. I wrote a few weeks ago about OpenAI's new science editor Prism. On the back of that Sam from https://www.underleaf.ai reached out to me. I'd not known about Underleaf before. From Sam's own words “I started building it in late 2024 because I thought ...

Google have developed an AI scientist - https://research.google/blog/accelerating-scientific-breakthroughs-with-an-ai-co-scientist/. Doing research is one thing, but the path from research to impact is tortuous and long winded, so we should expect many of these kinds of systems to focus around drug discovery. I still can't quite get my...

The rate of agentic development is increasing. Google had their Google next event two weeks ago, and they are going all in on agents. How we manage them in mid tier companies is a total mystery to me. I know maybe four or five people within our org who cold get their head around how to wire these together, but we don't currently have t...

Cloudflare Connect on Tour London - Notes On Wednesday the 15th I attended Cloudflare connect London and it was really good. I learnt a lot about their capabilities, in quite a short morning and early afternoon had a lot of high quality conversations. I wanted to write up my thoughts while they are fresh. (Through much of this post I a...

I installed https://thecdo.school/articles/i-made-claude-skill-to-help-you-read-the-room. Oddly the first global Claude skill I've installed. It looks to be very useful. Claude and Codex and the other tools have a growing learning curve. Installing the plugin in Claude desktop didn't work, with no clear reason why not. The cli installa...

This morning inspired by listening to https://www.youtube.com/watch?v=wc8FBhQtdsA and reading this thread on linked in - https://www.linkedin.com/posts/dartar_opensource-aiforscience-share-7448170944151355392-2bNP?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAEImlkB8SaKRnNedw-vBQuAiicIBoz5PWQ, I was struck by the section in the ...

The following is an important initiative and I wanted to share it in my network. Funding to support communication on science of science with a target on how to influence congress and the public. https://marginalrevolution.com/marginalrevolution/2026/04/new-emergent-ventures-tranche-on-science-policy-and-communication.html

Another AI agentic scientist. https://github.com/getcompanion-ai/feynman Another AI question answer assistant (being flogged in Insta and backed by YCombinator) https://app.answerthis.io/ Find references in seconds (and maybe don't hallucinate that shit while your at it) https://www.sourcely.net/ People love these tools, they are trust...

I have an article that has just come out in InPublishing - https://www.inpublishing.co.uk/articles/ai-can-do-many-things-but-it-cant-create-time-26451, and I’m happy with the result. Things are changing so fast with AI, in the piece I try to articulate how to think about what won’t change. On the back of it George Wakley has come up wi...

Not complete, and not exhaustive, but a nice perspective nonetheless. https://pardalotus.github.io/pardalotus-map

I spun up this small tool over the weekend to help me process photos of post-it notes - https://github.com/IanMulvany/post_it_viewer_analysis, it works pretty well.

This post is great - https://aarontay.substack.com/p/how-agentic-are-academic-deep-research I learnt a lot. My main reflection is that the fixed workflows that the current crop of tools are showing were developed when models were weaker, yet from a product perspective, trying to build something that is simultaneously specific and gener...

I am in a state of confusion and wonder. Most of the time I’m able to suppress this and stumble through my days without thinking too much, but now and again I fail and my attention gets arrested by the riotous confusion and intricacy of the world around me. We are odd creatures, bags of chemicals loosely held together, sinew and bone b...

I want to share an interesting pattern - get calude code on the web (or any other autonomous agent + sandbox) to do R&D on a coding question. Based on these blog posts - https://simonwillison.net/2026/Feb/6/pydantic-monty/ and https://simonwillison.net/2025/Nov/6/async-code-research/. I wanted to learn about WASM and Pyodide as I don't...

https://blog.mikeswanson.com/backseat-software/ > Metrics don’t measure reality. They measure what your product currently makes easy. > What starts as an occasional heads-up becomes a permanent layer of UI exhaust. What starts as support becomes a funnel. What starts as a reminder becomes a habit-forming system. In the context of acade...

I have been blogging for twenty years. Last year I wrote more than one post per week. I was very happy about that. You can see my posts here - https://mulvany.net/all-my-posts.html. AI is accelerating at a rapid pace and I am paddling furiously to keep on top of things. I am going to scale back my writing in 2026 so that I can focus on...