MyTh · My Thoughts

MyTh

Short for My Thoughts. Numbered, dated, and argued in public — research, universities, technology, policy, and anything else that will not leave me alone. Where I make a factual claim, I cite where it came from.

What this is

One thought per entry, numbered MyTh 001 onwards and listed newest first. Each has its own page. Nothing is ever renumbered.

House rule

Every number, quote and claim carries a citation to a named source. If I am wrong, the source is right there for you to check.

Whose views

Mine alone, written in a personal capacity. Not the position of the University of Southampton or any body I sit on.

Corrections

Email me and I will correct the text and say what changed. Arguments welcome; I would rather be corrected than quoted approvingly.

0 thoughts published Subscribe
MYTH007

The Question I Never Heard

I have bombed two interviews for jobs I badly wanted, and in both of them my failure was the same: I could not understand a word the interviewer was saying. Not the audio — the speech. This is what technical interviews actually measure, what the evidence says about that, what the good companies do that the bad ones do not, and the one failure mode that has no appeal, no protocol and no place to complain.

21 August 2026 Hiring Interviews Careers Accent For students ~16 min read
MYTH006

The Seat He Already Had

Clacton has just spent something north of a quarter of a million pounds to elect a man who already held the job, in a contest the other parties refused to enter, six weeks after the country acquired its seventh prime minister in ten years without a single member of the public voting for him. I want to be precise about what this cost, what it bought, and why the anger underneath it is aimed at roughly the right thing for roughly the wrong reason.

14 August 2026 Politics UK Democracy Cost of living Immigration ~14 min read
MYTH005

The Cat Saw a Dog

In March 2013, in a converted chocolate factory in Moscow, I put a slide up that said we cannot tell whether the cat saw a dog or a dog saw a cat — and told the room that word order was the future of text mining. I was right about the direction and wrong about the mechanism. The more useful story is what I had already learned, by then, about which ideas actually get published.

12 August 2026 My research NLP Information retrieval Research culture ~12 min read
MYTH004

Independent, and Never Alone

My students own their ideas. My job is to raise those ideas until a top venue cannot refuse them — and to be in the room the whole time, testing, arguing, refusing to be impressed too early. The full record of twenty-two people, what they built, and what this style of supervision actually costs the supervisor.

10 August 2026 Supervision PhD Research culture KTP ~13 min read
MYTH003

One Every Five Days

On good academic practice: how to not plagiarise, how to decide who goes on a paper, why reviewer collusion is the hardest of these to catch, and what a very large publication count does, and does not, tell you about the person holding it. Written deliberately about nobody.

9 August 2026 Research integrity Higher education Peer review Publishing ~15 min read
MYTH002

The Default Human

A model trained mostly on one slice of the world does not learn that slice as a culture. It learns it as neutrality, and everything else as deviation. But “make it less Western” turns out to be a worse answer than it sounds — and the field cannot yet measure whether it worked.

7 August 2026 AI Culture My research ~9 min read
MYTH001

The Four Squeezes: what is actually happening to academic life

A wage that has lost close to a third of its value, a student market that grew and shrank at the same time, a machine that now writes and marks, and a generation quietly recalculating whether any of it is worth it. Three of those four are worse than the headlines say. One is better.

7 August 2026 Higher education AI Policy ~10 min read
Subscribe

New entries, delivered without an email address.

MyTh publishes a web feed. Paste the address below into any feed reader and new entries arrive automatically — no account, no email address, and no way for me to know who is reading. There is no mailing list and there are no comments; this is the whole subscription mechanism.

Readers that take it: NetNewsWire (free), Feedly, Inoreader, Reeder. You can also open the feed itself — browsers no longer read feeds natively, so that page explains what it is rather than showing you markup.

Next

MyTh 008 is being argued with itself.

These arrive when a thought is finished rather than on a schedule. If you want to disagree with one — particularly if you have better data — that is the most useful thing you can send me.

Argue with me Subscribe The peer-reviewed version