iiinova — where your digital self keeps evolving

Richard G Clegg @richardclegg.bsky.social · 14/10/2026

I was genuinely surprised by it. I would have bet quite heavily on the actual lower bound being n log n and been wrong. I think a lot of people would be in that position. The lack of practical utility is not really the point I agree.

020 0

Richard G Clegg @richardclegg.bsky.social · 12/10/2026

Hope you don't think it shameless self promotion, I made a video on libxz. youtu.be/LY8_35a_DEA?...

youtu.be

XZ Exploit - Computerphile

YouTube video by Computerphile

110 0

Richard G Clegg @richardclegg.bsky.social · 12/10/2026

Interviewer: can you explain this gap in your CV? Internet researcher: the air gap ensures that if a hacker breaches my employment history or education record the personal details section remains secure.

010 0

Richard G Clegg @richardclegg.bsky.social · 10/10/2026

Honestly hooray for reasonable sized fonts. I see so many talks where I cannot even read them - I am usually sitting near the front.

020 0

Richard G Clegg @richardclegg.bsky.social · 10/10/2026

I presume they work with a circulated list of emails subjects and addresses to do the spam email and just messed up their automation. So more idiotic than creepy.

010 0

Richard G Clegg @richardclegg.bsky.social · 10/10/2026

For two decades I have received a lot of low effort spam email about PhD places but "I am reaching out specifically because of your work on Engineering, Eng 204, which is closely related to the kind of doctoral research I want to do." (That is my room number.) This is a new low.

141 0

Richard G Clegg @richardclegg.bsky.social · 09/10/2026

Amazing! Koko in Camden was unbelievably stinky in the weeks afterwards. Turns out stale cigarette masks quite a lot of vomit!

010 0

Richard G Clegg @richardclegg.bsky.social · 09/10/2026

There was a short period after the smoking ban where for a set of pubs and clubs we discovered the horrible stale cigarette smell was covering a really really horrible puke and god knows what smell. (I think they all discovered they needed cleaners soon after but oh my god difficult times.)

120 0

Reposted by Richard G Clegg

Doug Clow @dougclow.bsky.social · 09/10/2026

I thought lt adversarial poetry would be eg a jailbreak targeting common chatbots but this one is trad malware using a slop-like poem on GitHub to obfuscate the signalling of the current IP address of its command & control server.

011 0

Richard G Clegg @richardclegg.bsky.social · 09/10/2026

It would be a shame if a pushback reduced usage. Large scale dockless ebike can replace huge numbers of journeys. We see it to some extent in london and to a larger extent in China. It would be a shame if ecological benefits were lost because of the downsides.

020 0

Richard G Clegg @richardclegg.bsky.social · 08/10/2026

That is fair. If it got to ramming and boarding action then it wouldn't last long.

000 0

Richard G Clegg @richardclegg.bsky.social · 08/10/2026

If not surrounded the ship of the line can outpace the trireme and out range. At which point you play yakkety sax and speed up the tape to watch an extremely slow battle with the stern chasers firing.

110 0

Richard G Clegg @richardclegg.bsky.social · 08/10/2026

Didn't this turn out to be something very minor like the text implied three parameters when there were four in the lean model.

100 0

Richard G Clegg @richardclegg.bsky.social · 05/10/2026

Little's theorem: given certain properties (true of real world queues) the average queue length is the average arrival rate multiplied by the average service time.

010 0

Richard G Clegg @richardclegg.bsky.social · 05/10/2026

I prefer the non celeb version but it is very compelling.

010 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

Ah... this is pretty cool yes. I wasn't previously aware.

010 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

*BUT* there is a limit and I agree there's a point where it's not "synthetic data" it's just "stuff you made up".

010 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

The data cannot be anonymised by any known technique. The research task is a "close as possible" synthetic analogue that has strong guarantees it cannot reveal details of the original data. The research community optimises algorithms on the synthetic data that you hope work on the real data.

110 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

Systems like differential privacy do this and provide robust theoretical guarantees but don't work in all circumstances. Real world scenario: org has highly confidential data. They want the research community to provide algorithms that tell them things about the data but without revealing the data.

110 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

I'm more sympathetic here. I don't myself understand why they do it but enough serious people put time in (and before it was a "buzz" thing) to make me think there's something. I've worked on synthetic analogues representing properties of real data where the real data is highly secure.

200 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

I'm neutral on it. It does nobody harm to read the warning but maybe it does almost no good for it to be there. The incredible passion in the debate though is comedy gold.

010 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

I have never quite understood the synthetic data thing. Enough people have poured research effort into it (well before LLMs) that I end up feeling it must be me missing something.

110 0

Richard G Clegg @richardclegg.bsky.social · 01/10/2026

This is great. I traced this and you are correct and it derived from the discussion I mentioned. It was the three little pigs (I misremembered) and the legendarily bitter spoiler debate that led to the general policy being formed in 2007.

110 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

The Wikipedia discussion page on whether The Three Bears needed a spoiler warning was a thing of beauty back in the day. Genuinely people putting their emotions on the line. I suspect it is all deleted now.

110 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

Genuinely why I am a frequetist. My heart rebels against a form of science that fundamentally begins with "well I reckon..."

010 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

OK mate, for a pony I can only do you a point estimate but if you can spare a ton I can put a nice confidence interval around it. You want something Bayesian, I'll be needing a monkey and I'm going to see a prior up front before I put the lads on site.

110 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

I'm also biting back the urge to teach confident intervals despite my distaste for Bayesians. Honestly I'm having a ball doing the slides.

010 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

Heheh... to my shame I just produced a slide saying "There is a really cool thing here. The CENTRAL LIMIT THEOREM but we don't have time to learn it." I am honestly sad that we don't but it's a practical course on data/ML and I already spent 10 slides showing coin tosses and different normals.

110 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

Yeah. That is sensible if you have the class size that supports it. Scales badly but works for small/mid classes.

110 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

What was the assessment response?

100 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

Preparing some teaching material on data analysis -- some basics. I found it a really exciting moment when I first understood why there was an (n-1) in the formula for sample variance. It's a small thing but I can remember how satisfying I found that ugly weirdness turning into an "aha!"

120 0

Richard G Clegg @richardclegg.bsky.social · 30/09/2026

This week one PhD student handed in corrections, another handed in final thesis and another is finalising their first major conference paper. It's the circle of life. (I've never even seen this film.)

Circle of life as illustrated by the lion king.

020 0

Reposted by Richard G Clegg

Zachary Neal @zpneal.bsky.social · 30/09/2026

The special issue on 𝗖𝗹𝗮𝘀𝘀𝗿𝗼𝗼𝗺 𝗔𝗰𝘁𝗶𝘃𝗶𝘁𝗶𝗲𝘀 𝗳𝗼𝗿 #𝗧𝗲𝗮𝗰𝗵𝗶𝗻𝗴 #𝗡𝗲𝘁𝘄𝗼𝗿𝗸𝘀 is now available in @insna.bsky.social 𝘾𝙤𝙣𝙣𝙚𝙘𝙩𝙞𝙤𝙣𝙨. The editorial introduction contains short descriptions of each activity, with direct links to the open-access papers providing detailed lesson plans. doi.org/10.21307/con...

doi.org

142 0

Richard G Clegg @richardclegg.bsky.social · 29/09/2026

Managed to sign "I like big boats" to my buddy during a wreck dive. She was not impressed. Still trying to figure out a way to sign "and I cannot lie".

020 0

Richard G Clegg @richardclegg.bsky.social · 29/09/2026

A quick Google image search for UDC posters was a really queasy comedy horror show.

010 0

Richard G Clegg @richardclegg.bsky.social · 28/09/2026

Also the infuriating way that it is pushed as the default bugs me. I have my own back up systems so don't need shonky stuff like One Drive but MS is so eager to push it it makes it seem like you messed up if you do not use it.

020 0

Richard G Clegg @richardclegg.bsky.social · 28/09/2026

Also, you know, watching a series of votes that are in the 50%s against quite far right propositions makes reading how well it's going seem honestly worrying. Like looking at an imminent accident while someone cheerfully explains how well they're driving.

110 0

Richard G Clegg @richardclegg.bsky.social · 28/09/2026

"Uneducated Brits voted for Brexit by 52% but the wise and educated Swiss voted against a population cap (FFS a population cap) by 55%" feels a tenuous narrative. Switzerland does not feel like the poster child for "welcome people not from round here".

110 0

Richard G Clegg @richardclegg.bsky.social · 28/09/2026

Looking forward to it. I only read the first one. I just reread the published version last week. I genuinely love the weirdos.

100 0

Richard G Clegg @richardclegg.bsky.social · 26/09/2026

Fair. Being annoyed by the antagonist is defensible, being annoyed by the antagonist and protagonist, now it is just winding you up. I wonder if the different V origin stories are more or less irritating?

000 0

Richard G Clegg @richardclegg.bsky.social · 26/09/2026

I presumed Johnny being irritating (less so as the game continues but still very irritating) was authorial intent. You are supposed to want rid of him very badly.

100 0

Richard G Clegg @richardclegg.bsky.social · 26/09/2026

Yeah. It brings the issue more into focus. I had heard the connectome model somehow did not replicate real behaviour when wired to a best approximation environment but I cannot find a good reference. (That is, is there some element not captured by that model?)

100 0

Richard G Clegg @richardclegg.bsky.social · 26/09/2026

I share the queasiness about sentience but I am there already for large neutral networks. I don't think mirroring a real system makes it more likely to be conscious

110 0

Richard G Clegg @richardclegg.bsky.social · 25/09/2026

Perpendicular bisector is a wrestling move RJ. I am sure of it. :)

110 0

Richard G Clegg @richardclegg.bsky.social · 18/09/2026

The deaths were my fault (my choices). Those choices were made dozens of hours of game play ago (not simply recoverable). They were not obvious "save A or save B" choices. I still think about it. It happened because I was willing to miss bits (maybe large bits) of the game.

001 0

Richard G Clegg @richardclegg.bsky.social · 18/09/2026

So many gamers are obsessive about a "maximise your experience" play through (with walkthrough). The joy of the catastrophic failure eludes them. One of my most memorable gaming experiences was in mass effect 2. My choices got characters I liked killed. That is powerful.

100 0

Richard G Clegg @richardclegg.bsky.social · 18/09/2026

My name is Ozymandias king of kings. Look upon my works ye mighty and despair. For example these vast and *trunked* legs of stone, got my full body and face too. Lots of complete statues of me exist. I am frustrated by this trunkless legs narrative.

Statue of Ramses II in Cairo museum. His legs are vast but not trunkless.

071 0

Richard G Clegg @richardclegg.bsky.social · 18/09/2026

I think you saw this before but worth reviewing when so many people are convinced they can pick out AI. (I hand write because it is easy for me then get an AI to grammar check and critique. I have Claude set to be extremely sarcastic so it can be cutting.) www.marklawrence.buzz/2025/08/so-i...

marklawrence.buzz

So ... is AI writing any good? .... PART 2!

So ... is AI writing any good? Two years ago I asked the same question and pitted 4 good authors against ChatGPT 4 in a flash fiction head-t...

010 0

Richard G Clegg @richardclegg.bsky.social · 18/09/2026

Honestly I will be so glad when people realise they cannot tell. My academic writing is formal with correct grammar so some think it is AI. I find myself being asked by co authors to remove AI style phrases that are not AI and we all know are not. (E.g. a phrase with a statement and inversion).

010 0