Another example of the bitter lesson?
03.10.2025 13:27 β π 3 π 0 π¬ 0 π 0@ed-sherman.bsky.social
Deputy Director ONS Data Science Campus https://datasciencecampus.ons.gov.uk
Another example of the bitter lesson?
03.10.2025 13:27 β π 3 π 0 π¬ 0 π 0I'm afraid we've been at it again
17.09.2025 21:05 β π 79 π 10 π¬ 3 π 0βSuper sewerβ chief bows out to the sweet smell of success on.ft.com/4gdIXWU
10.09.2025 06:27 β π 25 π 3 π¬ 1 π 4π
05.09.2025 12:52 β π 0 π 0 π¬ 0 π 0Tempting fate? No, the second scoop is an earl grey sorbet!
21.08.2025 12:43 β π 2 π 0 π¬ 0 π 0See also: the phenomena of second scoop regret with gelato h/t @scientificdiscovery.dev
15.08.2025 13:47 β π 1 π 0 π¬ 1 π 0New method detects inflation turning points using disaggregated price data, revealing early signals in Argentina's 2024 disinflation and US tariffs in 2025, from Alberto Cavallo and Gaston Garcia Zavaleta https://www.nber.org/papers/w34102
12.08.2025 15:00 β π 28 π 3 π¬ 1 π 0I think this way of doing is exactly the right one for figuring out AI tool applications and trying to improve internal processes (albeit itβs not my kind of language). The discovery/alpha/beta cycle or framework can become a straight jacket. Or trap. Or whatever the right metaphor is.
11.07.2025 12:41 β π 1 π 0 π¬ 0 π 0I've started a new blog called 'ways of doing' which is about ways of doing digital. Thanks to @matthewsheret.com for suggesting the name.
09.07.2025 08:54 β π 19 π 7 π¬ 2 π 2Whats happening to birth rates around the world? Iβll be chairing a panel today at 1.30pm on how population change will change our world. With @rebeccasear.bsky.social, Emily Jackson & @jburnmurdoch.ft.com - all welcome in our cool central London building!
www.thebritishacademy.ac.uk/events/summe...
Another prompt injection paper review! This time it's "An Introduction to Googleβs Approach to AI Agent Security" by Santiago DΓaz, Christoph Kern, and Kara Olive
Some interesting ideas in here, particularly around Google's three core principles for agent security simonwillison.net/2025/Jun/15/...
Something something, innovation is not linear
14.06.2025 06:52 β π 1 π 0 π¬ 0 π 0How the world has changedβ¦ but this did make me look up what it was made of, and it has its own wiki page
en.wikipedia.org/wiki/Gunge
For those interested in mirror life, yesterday the Government Office for Science published an expert roundtable summary on it
www.gov.uk/government/p...
Really interesting to see this topic getting attention. Did you see that yesterday the Government Office for Science just published an expert roundtable summary on mirror life?
www.gov.uk/government/p...
My experience of taking SPL was great. I took 10 weeks off at the start (more later) and wasnβt remotely ready to come back till 5. I strongly encourage all the fathers to be in my teams to take a substantial chunk off at the start too
10.06.2025 15:40 β π 2 π 0 π¬ 0 π 0Increasingly believe that this should be required reading before anyone is allowed to do science.
I spend an *inordinate* amount of time these days citing this piece and explaining that RCTs are not instrinsically and exclusively causal.
www.sciencedirect.com/science/arti...
Here's a great little tutorial on getting started with my LLM tool for running local models!
One of the best ways you can contribute to any open source project is to publish unofficial guides like this one
Developing a quantitative, multi-sector, semi-endogenous growth model of structural change in innovation and production, incorporating demand-pull and technology-push forces, from Diego A. Comin, Danial Lashkari, and MartΓ Mestieri https://www.nber.org/papers/w33855
01.06.2025 13:00 β π 9 π 1 π¬ 0 π 1Population aging affects firm entry and profits via demand: older households' higher consumer inertia acts as a barrier to entry and lets incumbents raise markups, from Gideon Bornstein https://www.nber.org/papers/w33820
27.05.2025 17:00 β π 1 π 4 π¬ 0 π 0What will the new arts and crafts movement be?
25.05.2025 19:05 β π 0 π 0 π¬ 1 π 0Management practices account for about 20 percent of cross-country differences in productivity, with the US having the world's best managed firms, from Raffaella Sadun, Rachel J. Schuh, Jonathan S. Hartley, John Van Reenen, and @nickbloom.bsky.social https://www.nber.org/papers/w33765
12.05.2025 19:00 β π 14 π 1 π¬ 0 π 0Just out: our new NBER working paper explores what happens when management is treated like a technologyβsomething that spreads across firms, boosts productivity, and lowers the cost of change.
12.05.2025 14:59 β π 11 π 4 π¬ 2 π 2Thereβs a time that comes for all of us, a threshold of second scoop regret and when sorbet becomes an increasingly attractive optionβ¦
05.05.2025 19:01 β π 1 π 0 π¬ 0 π 0I love that place (and others near by) - but is very extra. Last time I just couldnβt get through it, my body just shut down
05.05.2025 17:51 β π 1 π 0 π¬ 1 π 0Dave's approach to evals is really interesting. It encouraged me to take a look at promptfoo, which appears to be the most mature of the current stack of open source eval tools simonwillison.net/2025/Apr/24/...
24.04.2025 18:59 β π 42 π 7 π¬ 3 π 1Distilling key facts about housing affordability and its decline in the US and reviewing the literature on housing supply and the causes of declining affordability, from Nathaniel Baum-Snow and Gilles Duranton https://www.nber.org/papers/w33694
20.04.2025 17:00 β π 22 π 5 π¬ 0 π 0I understand what youβre saying, and I can see itβs a nice hobby for those that like it, I just think theyβd be better in London as actual parks or woods, or homes. Also, the more we build in London the less we need to build outside, and people in London have a lower environmental impact
17.04.2025 17:59 β π 0 π 0 π¬ 0 π 0I think outside of London the tradeoffs for land use are not really such a big deal. Itβs not my thing, but happy for others! :)
17.04.2025 15:47 β π 0 π 0 π¬ 0 π 0