Arthur Clune
LightNews LightNews
clune.org
Arthur Clune
@clune.org
110 followers 280 following 360 posts
Geek. Likes bikes, climbing and tech

Work: IT at University of Sheffield
Posts Replies Media Videos
clune.org
Age yourself with gaming
November 12, 2025 at 6:48 PM
clune.org
Big hint in their write up that they weren't using Cursor to write code
Cursor builds tools for software engineering, and we make heavy use of the tools we develop. A motivation of Composer development has been developing an agent we would reach for in our own work. In recent weeks, we have found that many of our colleagues were using Composer for their day-to-day software development. With this release, we hope that you also find it to be a valuable tool
October 30, 2025 at 5:22 PM
clune.org
Begun the document wars have

No, seriously.

www.scmp.com/economy/chin...
China's expansion of its rare earth export controls appeared to mark another escalation in the US-China trade war last week. But the announcements were also significant in another way: unusually, the documents could not be opened using American word processing software. For the first time, China's Ministry of Commerce issued a slew of documents that could be directly accessed only through WPS Office - China's answer to Microsoft Office - as Beijing continues its tech self-reliance drive. Developed by the Beijing-based software company Kingsoft, WPS Office uses a different coding structure to Microsoft Office, meaning WPS text files cannot be opened directly in Word without conversion.
October 14, 2025 at 7:13 AM
clune.org
Also Claude, 🔥
Output from Claude: "The verbosity you'll encounter isn't accidental; it's COBOL heritage, designed for business analysts to supposedly read. Whether that goal was achieved is debatable."
October 8, 2025 at 8:54 AM
clune.org
I'm writing a chatbot as an experiment. Here's Claude Code debugging why the tool calling isn't working

Spoiler - this was not the reason!
Output from Claude Code: So the LLM is working correctly! It's: 1. Acting as Gimli (a grumpy dwarf who doesn't like people) 2. Saving memories about Elara via tool calls 3. But not responding verbally because Gimli is grumpy and doesn't want to talk The empty responses make perfect sense for Gimli's character - he sees Elara, grumbles, saves the information to memory, but doesn't actually say anything back because he's antisocial. The memory system is working perfectly. The "issue" is that we're testing with a character who doesn't like to talk!
September 1, 2025 at 1:58 PM
clune.org
The chart crimes will continue until morale improves #gpt5
A chart screen captured from the opanai livestream. It's wrong.
August 7, 2025 at 5:32 PM
clune.org
Numbers from Google on tokens. Doubling every month. 1,000tn in June.
Tweet from Demis Hassabis stating the Google processed a quadrillion tokens in June 2025
July 31, 2025 at 4:15 PM
clune.org
It's here. Sigh. Age ID for DMs
July 23, 2025 at 9:41 PM
clune.org
Fortunately the “learn more about our brand” page explains everything
A brand rooted in unconventionalism. Our strategic brand framework is built on our Brand Purpose, Brand Values & Behaviours and Marketing Themes, all of which collectively shape our Brand Proposition.
July 22, 2025 at 12:40 PM
clune.org
I read this from the OfS as saying that nearly 20% of UK Unis are at risk of going under

(from the @resprofnews.bsky.social newsletter)
On Friday, the OfS published its annual report. Safe to say it offers no let-up in the level of concern about possible university insolvency. It revealed that 71 out of the 400 or so providers on its register were subject to "formal monitoring" over finances.
July 21, 2025 at 9:37 AM
clune.org
It's a cool idea for sure. But why is it surprising? There's plenty in that trace for a human to understand what it does. I assume that it scales well beyond the toy examples into something that needs a lot more reasoning?
July 19, 2025 at 1:12 PM
clune.org
And here’s a quote from one of the developers in the study. This is very relatable!

x.com/ruben_bloom/...
X post from one of the developers in the MITRE study: This is much less true of my participation in the study where I was more conceintious, but I feel like historically a lot of my Al speed-up gains were eaten by the fact that while a prompt was running, I'd look at something else (FB, X, etc) and continue to do so for much longer than it took the prompt to run
July 15, 2025 at 6:50 AM
clune.org
There's some eye-opening things in the Observer's Sensemaker email today about the ONS

observer.co.uk/newsletters

It also appears it may be another instance of problems of governance and leaders who don't listen

archive.is/lvPkk

And yes, Newport is not attractive and doesn't help
What went wrong? In 2023, an email was sent to the head of the ONS, chief statistician Sir Ian Diamond, identifying major problems in UK employment data due to precipitous collapses in sample size. It wasn't unique. Covid disrupted face-to-face interviews and people have since become increasingly reluctant to take cold calls, share personal data or spend time filling in surveys. The pandemic also produced a lot of statistical "noise" For the ONS it was acute. Analysis of other sources showed at one point the ONS was underestimating the size of Britain's workforce by as much as one million people. Stats for certain sub-groups were swinging by 30 per cent in a month because they depended on a sample size of five. In October 2023, the ONS was forced to temporarily pause its monthly readout on jobs Cultural concerns. A government-commissioned review led by Sir Robert Devereux identified a slump in morale at the ONS. Some of it pointed to the role of the "whacky" ONS boss Diamond, who departed shortly before its publication on health grounds. It also noted that • widespread WFH weakened team cohesion and undermined data quality; • flagship programmes - such as Diamond's pet project, a data service costing £200 million over five years - had come at the expense of high-quality core statistics; and • there had been struggles to recruit and retain skilled analysts, partly due to below average pay and the location of the ONS headquarters in Newport, Wales.
July 11, 2025 at 9:42 AM
clune.org
In @samfr.bsky.social's latest post, he uses this image as the header. Honestly if someone told me it was AI I'd nod along (weird expressions, odd body language). But it's probably just a really awkward photo op 🤷‍♂️
A picture of Keir Starmer, Rachel Reeves and some other people in a hospital. They look very odd.
July 9, 2025 at 9:39 AM
clune.org
AI and power from an angle I'd not thought about: homogenous workloads mean power draw can change by 10s of MW instantly. This is very hard to handle

semianalysis.com/2025/06/25/a...

(free preview with registration, but the preview is massive)
Quote from Llama paper: "During training, tens of thousands of GPUs may increase or decrease power consumption at the same time, for example, due to all GPUs waiting for checkpointing or collective communications to finish, or the startup or shutdown of the entire training job. When this happens, it can result in instant fluctuations of power consumption across the datacenter on the order of tens of megawatts, stretching the limits of the power grid. This is an ongoing challenge for us as we scale training for future, even larger Llama" Chart from Google Cloud showing 1.5MW peak to trough difference in a non-AI datacentre v a 15MW change in milliseconds for AI workloads
June 26, 2025 at 8:39 AM
clune.org
This from @simonwillison.net has just blown my mind. Not because of Google releasing a Claude Code clone. That's expected. It's that Claude Code only came out in February!

It has so completely changed things that this feels like a lifetime ago
Gemini CLI. First there was Claude Code in February, then OpenAl Codex (CLI) in April, and now Gemini CLI in June. All three of the largest Al labs now have their own version of what I am calling a "terminal agent" - a CLI tool that can read and write files and execute commands on your behalf in the
June 26, 2025 at 8:30 AM
clune.org
And then there's this. Charasmatic confident leadership that sold the story well
It was noted by many individuals that the Principal, who was an excellent public speaker, tended to control the narrative to Court and elsewhere, consistently painting a picture that all was well in the University, highlighting the institution's successes, working well with donors and external stakeholders, rightly praising the achievements of students and bestowing esteem upon the University by his election as Convenor of Universities Scotland. There were however many reports to this investigation that the Principal frequently demonstrated hubris, or excessive pride in his role. Hubris is relevant insofar as it can be characterised by dangerous over self-confidence and complacency, often in combination with arrogance, contempt towards people who offer criticism and obsession with personal image and status. Additional comments were noted regarding the Principal's overbearing leadership style, behaviours and dislike of potentially awkward confrontations or questioning and the potential adverse impact these factors may have had upon individuals and the overall culture of engagement within the Universi
June 19, 2025 at 9:46 PM
clune.org
There was also a student admissions IT project

Reading between the lines on this one, it feels like a lot of customisation was wanted. The report also makes clear that the project wasn't flagged of critical impact soon enough
We have seen no record that universities that had already implemented Blueprints were asked by the University about their experience of using it. In June 2023 the Admissions team raised major concerns regarding the amount of re-work needed to their processes. In February 2024, the project risk status was changed to 'red' • In July 2024 a list of 'configuration items' (changes that the University wanted) was shared with Tribal who were unable to give target dates for their implementation. Despite this, a go-live decision was made for September 2024. Following a short delay the system did go live but, according to the University, was unable to be used for seven weeks. No 'work arounds' for offers or admissions had been developed in advance of the system going live and, as the data had been migrated from the old system to Blueprints, we were told that it was not possible to introduce a manual system at the time or revert to the old system which had been 'switched off'
June 19, 2025 at 9:24 PM
clune.org
The uni spent half its reserves in 6 months and it's possible it just wasn't noticed
During the period from 1 January 2023 to 31 July 2024, cash balances decreased from c.£ 100m to c. £32m. At the same time, there was a lack of consistency around the reporting of cash and liquidity issues to UEG, FPC and Court. Consequently, no action was taken during this period to arrest the decline, and, at the time of this Investigation, the University is forecasting (as of May 2025) that the closing cash balance at the end of FY25 will be £1m overdrawn. From 1 August 2023 to 31 July 2024 there was a £68m net outflow which culminated in available cash of c. £100m reducing to c.£32m, approximately £42m of this outflow occurred in H2 FY23. The deterioration in the University's cash position continues into FY25 where the University is forecasting a closing bank balance of £1m overdrawn.
June 19, 2025 at 9:15 PM
clune.org
I was out and about with the offspring (17) last night.

Stats say young people don't drink. J lives up to this. He said "Are you having ANOTHER one" in a very disapproving voice as I ordered my second and last pint of session IPA 😂 (*)

(*) old person's emoji on purpose actually
Photo showing the Satchvia band taking their applause at the end of the gig
June 14, 2025 at 8:02 AM
clune.org
"Average ChatGPT request uses ~0.34Wh"

engineeringprompts.substack.com/p/the-averag...

I completely agree with Marcel's summary here: the per query cost is tiny but the total usage can still become very significant (as I argued last month clune.org/posts/enviro...)
Allow me also to make the point again that this doesn't absolve us from thinking deeply about the coming explosion due to energy needs. Here as well, I am sticking to my guns, arguing that we may currently be underestimating our overall energy needs, and dramatically so. In short: While a concern about exploding energy requirements in a world full of Al agents remains entirely defensible, regular chatbot use is really nothing to worry about in terms of energy use.
June 11, 2025 at 9:25 AM
clune.org
Lots of stats in this HEPI report on AI in Education. Picking out just two - everyone is using it, and the concept of hallucinations is much better understood now

www.hepi.ac.uk/2025/02/26/s...
"In 2025, we find that the student use of Al has surged in the last year, with almost all students (92%) now using Al in some form, up from 66% in 2024, and some 88% having used GenAl for assessments, up from 53% in 2024. The main uses of GenAl are explaining concepts, summarising articles and suggesting research ideas but a significant number of students - 18% - have included Al-generated text directly in their work." Graph showing that in 2024 35% of students didn't know about hallucinations and in 2025 only 12% don't. Also in 2025 students think the tools produce more hallucinations, possibly because they are now aware to look
June 2, 2025 at 10:37 AM
clune.org
Ergodox stan here. OG soldered up by me. Which was a trial as I had never soldered anything smaller than two wires back in school

Clicky switches ofc
May 31, 2025 at 8:34 AM
clune.org
Sometimes I read a total banger of a line and am just left wishing I'd written it, like this one on the experience of using the in-built AI tools to write emails

koomen.dev/essays/horse...
"The email I'd have written is actually shorter than the original prompt, which means I spent more time asking Gemini for help than I would have if l'd just written the draft myself. Remarkably, the Gmail team has shipped a product that perfectly captures the experience of managing an underperforming employee"
May 29, 2025 at 8:00 PM
clune.org
This is so useful - one liner to summarise github. And Claude wrote the python. o4-mini called 'gh' incorrectly first but then fixed it
Using simonw's llm tool with function calling to run the Github 'gh' cli command to get info on a milestone First part of the output, showing the tool calling 'gh' incorrectly twice before correcting itself User output from the command, showing a summary of a milestone in the repo
May 29, 2025 at 12:35 PM