Log inSign up
Log inSign up
Steven Murdoch
468 posts
Steven Murdoch profile banner
@sjmurdoch

Steven Murdoch

@sjmurdoch
Professor of Security Engineering; Head of @uclisec; Director @OpenRightsGroup. 🐘 mastodon.social/@sjmurdoch 🦋 steven.murdoch.is
London, UK
murdoch.is
Joined November 2008
554 Following
5,435 Followers
RepliesRepliesRepostsRepostsMediaMedia

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • @sjmurdoch
    Steven Murdoch
    @sjmurdoch
    Sep 18
    Many misconceptions circulate about thermal communication. These are covert channels, which allow small secrets to be sent between colluding parties. They can also (sometimes) be turned into side channels where small secrets are leaked to a malicious observer. 🧵
    1
  • @sjmurdoch
    Steven Murdoch
    @sjmurdoch
    Sep 8
    It’s #NationalCodingWeek next week, so I've published a resource on using the BBC @microbit_edu to play a game that explores how radio waves travel, get blocked by/reflected off obstacles, and the challenges of measuring strong signals that can overload sensors.
    BBC micro:bit
    1
  • @sjmurdoch
    Steven Murdoch
    @sjmurdoch
    Aug 18
    I spoke to The Guardian about Claude’s content watermarking. As long as it only modifies how the existing random number generator works, it should have no measurable effect on quality. theguardian.com/technology/202…
    However, LLMs already do not make the best choices. “There’s already randomness involved in any of these large language models,” said Murdoch. “It’s pretty essential to how they work. If it wasn’t for this randomness, then they’d get stuck in loops and start repeating the same thing over and over again.”

In other words, a chatbot does not necessarily decide to call running water a “stream” as opposed to a “brook” because the latter might have a more old-fashioned register. The model does not contemplate these choices: it makes them by chance.
    1
  • @sjmurdoch
    Steven Murdoch
    @sjmurdoch
    Aug 11
    A prompt that’s been very helpful at getting Claude to improve a user-facing app with minimal interaction from me has been to ask it to do a cognitive walkthrough with a given persona. It’s not great at coming up with good personas (IME) but if you give it one it can run with it.
    Take on the persona of [INSERT PERSONA DEFINITION HERE].

Perform a cognitive walkthough. Update the app so that at each stage:

The user will try to achieve the right result
The user will notice that the correct action is available
The user will associate the correct action with the result they're trying to achieve
After the action is performed, the user will see that progress is made toward the goal
This persona is only one of several that will use the app, so take care not to compromise existing qualities of the app without good reason.

Add a summary of each cognitive walkthrough performed to COGNITIVE-WALKTHROUGHS.md, including the persona, issues identified, and changes made.
    Claude and ClaudeDevs
    @sjmurdoch
    Steven Murdoch
    @sjmurdoch
    Aug 7
    Replying to @sjmurdoch
    This is 100% Claude Code generated; somewhat tested but not reviewed. Your mileage may vary. Confirm with the official app before planning your journey (in particular, the route planner doesn’t know opening hours). scottiesinthecity.co.uk/app
  • @sjmurdoch
    Steven Murdoch
    @sjmurdoch
    Aug 7
    Are you trying to complete the @MaggiesCentres Scotties in the City art trail in Glasgow? I made a route planner to find an optimised path to get as many as you can, given a starting point and distance/time limit. sjmurdoch.github.io/glasgow-scotti…
    Map showing a route around Glasgow to collect Scottie Dogs.
    1