banner
robclaxton.bsky.social
@robclaxton.bsky.social
An engineer at heart | works in data and AI | enjoys spending time under water
why have an AI oven when you could just relax and be gently roasted in your smart bed?

www.pcworld.com/article/2948...
These smart beds began roasting their owners during AWS outage
Eight Sleep smart beds got stuck in position and refused to cool down during Monday’s massive Amazon cloud outage.
www.pcworld.com
November 18, 2025 at 4:27 PM
I've been reflecting on this, having been making extensive use of Amazon Q CLI recently. It's changing how I explore and develop ideas - for the better. So, I agree, next token prediction isn't sufficient to describe what it's doing.
September 23, 2025 at 7:59 PM
and I'm sure we can expect the weirdness to show up in EUSPRIG's spreadsheet Horror Stories real soon...
eusprig.org/research-inf...
Horror Stories | European Spreadsheet Risk Interest Group
Public reports of spreadsheet errors have been sought out on behalf of EuSPRIG by Patrick O’Beirne of Systems Modelling for many years. There are very many reports of spreadsheet related errors and they seem to appear in the global media at a fairly consistent rate.
eusprig.org
August 20, 2025 at 3:06 PM
I think "stochastic parrot" is a useful reminder that no matter whether we judge the output of a model to great or garbage, it was the same process that produced it - next token prediction.
August 12, 2025 at 8:49 PM
The value of marshmallows could easily collapse in the meantime. Probably best to take them now while they are still worth something...
July 23, 2025 at 9:40 AM
And in a rivetting conclusion to this story, they are now back together again in close harmony.
July 14, 2025 at 10:23 AM
To be fair they had been moving apart for some time.
July 14, 2025 at 9:35 AM
#SkyAtNight seems to stand alone as a programme that regularly brings cutting edge science to a wide audience whilst treating them with respect and not oversimplifying. Long may it continue.
July 13, 2025 at 11:44 AM
I've yet to find anyone that's actually read the Dunning-Kruger paper or any of the subsequent work on the topic.
May 28, 2025 at 12:29 PM
The Voyager probes were, and continue to be, engineering marvels. The work that has gone into managing issues (in some cases from the very early days) is extraordinary. Section 3.7.2.6 of this document gives a feel for what is required.
voyager.gsfc.nasa.gov/Library/Deep...
voyager.gsfc.nasa.gov
May 17, 2025 at 10:26 AM
this is a starter kit for a DuckDB cluster, right?
February 26, 2025 at 8:02 PM
another excellent episode! The RTLS abort procedure sounded...courageous. Bet everyone was glad they never had to try it! There's some good writeup on wikipedia (en.wikipedia.org/wiki/Space_S...) which emphasises just how many factors would have to be juggled to make it successful.
Space Shuttle abort modes - Wikipedia
en.wikipedia.org
February 22, 2025 at 11:57 AM
very much here for this content (reaches to the bookshelf for Bringhurst's 'The Elements of Typographic Style').
February 7, 2025 at 3:52 PM
Another great episode. I loved the description of the fuelled shuttle waiting on the pad, alive with moaning, creaking and valves firing. Brilliant stuff!
February 5, 2025 at 12:31 PM