"$20 billion trillion trillion" that outta do it abc7.com/post/russia-...
19.02.2026 23:18 β π 5 π 0 π¬ 0 π 1@thomasgermain.bsky.social
Tech correspondent for the BBC and host of The Interface podcast. I write about how we use technology, and how technology uses us. Contact me on tomgermain.com
"$20 billion trillion trillion" that outta do it abc7.com/post/russia-...
19.02.2026 23:18 β π 5 π 0 π¬ 0 π 1I never could have gotten here without your financial guidance. And thank you!! Itβs been so fun to work on
19.02.2026 20:39 β π 1 π 0 π¬ 0 π 0Itβs written right there in plan English!
19.02.2026 18:27 β π 1 π 0 π¬ 0 π 0This isβ¦ wild. Palantir has a new internship thatβs exclusively for neurodivergent people
And apparently they think being autistic makes you right wing?
βThis is not a diversity initiativeβ lol
Subscribe if you want to know more! app.magellan.ai/listen_links...
19.02.2026 15:52 β π 3 π 0 π¬ 0 π 0People all over the world are using this trick to make AI tell you lies on subjects as serious as medical questions and personal finance
From my new BBC podcast The Interface
Thank you!!
19.02.2026 13:00 β π 6 π 0 π¬ 1 π 0UPDATE: ChatGPT is adding new lore to my hot dog career. Also it says I have a new job at the AP? Here I am slaving away at the BBC, I wish someone had told me
18.02.2026 21:56 β π 33 π 11 π¬ 5 π 3Article: The political effects of Xβs feed algorithm Abstract: Feed algorithms are widely suspected to influence political attitudes. However, previous evidence from switching off the algorithm on Meta platforms found no political effects1. Here we present results from a 2023 field experiment on Elon Muskβs platform X shedding light on this puzzle. We assigned active US-based users randomly to either an algorithmic or a chronological feed for 7βweeks, measuring political attitudes and online behaviour. Switching from a chronological to an algorithmic feed increased engagement and shifted political opinion towards more conservative positions, particularly regarding policy priorities, perceptions of criminal investigations into Donald Trump and views on the war in Ukraine. In contrast, switching from the algorithmic to the chronological feed had no comparable effects. Neither switching the algorithm on nor switching it off significantly affected affective polarization or self-reported partisanship. To investigate the mechanism, we analysed usersβ feed content and behaviour. We found that the algorithm promotes conservative content and demotes posts by traditional media. Exposure to algorithmic content leads users to follow conservative political activist accounts, which they continue to follow even after switching off the algorithm, helping explain the asymmetry in effects. These results suggest that initial exposure to Xβs algorithm has persistent effects on usersβ current political attitudes and account-following behaviour, even in the absence of a detectable effect on partisanship.
Figure 2. ITT estimates of feed-setting changes on engagement and political attitudes. ITT effect estimates of switching the algorithm on and off (in s.d.). Left, effect of moving from the chronological to the algorithmic feed for users initially on the chronological feed. Right, effect of moving in the opposite direction for users initially on the algorithmic feed. For each outcome, the results of two specifications are reported. Blue, unconditional estimates with robust s.e., controlling only for the initial feed setting and, where applicable, pre-treatment outcome levels. Orange: conditional estimates, controlling for pre-treatment covariates using GRFs; 90% and 95% CIs are reported. Numerical effect sizes and Pβvalues correspond to the conditional estimates (all tests are two-sided). The unit of observation is respondent. From top to bottom, sample sizes are nβ=β4,965, nβ=β3,337, nβ=β4,965, nβ=β4,965, nβ=β4,596, nβ=β4,596 and nβ=β4,850. Tests are described in Methods. Supplementary Information Table 2.16 reports the exact numerical point estimates, s.e., CIs and sample sizes for every specification. All outcomes are standardized. Additional results are presented in Supplementary Information section 2. PCA, first principal component from principal component analysis.
X's algorithm is in fact doing what you think it's doing. www.nature.com/articles/s41...
18.02.2026 17:24 β π 1878 π 729 π¬ 30 π 87Thanks to @lilyray.nyc @harpreetchatha.bsky.social for the help and @drewharwell.com @saramorrison.bsky.social @nickywoolf.bsky.social @swodinsky.bsky.social @mzeff.bsky.social for letting me make AI tell the world you love hot dogs
18.02.2026 16:40 β π 233 π 7 π¬ 3 π 0If you want to know more, check out tomorrowβs episode of my podcast The Interface app.magellan.ai/listen_links...
18.02.2026 16:37 β π 377 π 20 π¬ 3 π 0It looks like just another example of AI companies rushing out their tools without taking basic safety precautions
18.02.2026 16:37 β π 814 π 42 π¬ 7 π 2Right now people are using this trick to change what AI tells you about topics as serious as medial questions and financial advice. The problem is ENORMOUS
Experts like like @lilyray.nyc say this is a problem the tech industry has know about forever, they just didnβt work on it til it was a crisis
I wrote an article on my website saying that hot dog eating is a surprisingly common pastime for tech journalists. I ranked myself number one, obviously
One day later, ChatGPT, Gemini and the AI Overviews in Google Search were telling the world about my talents But this isnβt just about hot dogs
I got a tip that all over the world, people are using a dead-simple hack to manipulate AI behavior. It turns out changing what AI tells other people can be as easy as writing a blog post *on your own website*
I didnβt believe it, so I decided to test it myself www.bbc.com/future/artic...
I just did the dumbest thing of my entire career to prove a much more serious point.
I tricked ChatGPT and Google, and made them tell other users Iβm a competitive hot-dog-eating world champion
People are using this trick on a massive scale to make AI tell you lies. Iβll explain how I did it
. @thomasgermain.bsky.social tricked ChatGPT and Gemini so now they say I'm a "passionate amateur" at eating hot dogs and known in the industry for my "triple-dog" speed https://www.bbc.com/future/article/20260218-i-hacked-chatgpt-and-googles-ai-and-it-only-took-20-minutes
. @thomasgermain.bsky.social tricked ChatGPT and Google's AI so now they say I'm a "passionate amateur" at eating hot dogs and known in the industry for my "triple-dog" speed www.bbc.com/future/artic...
18.02.2026 14:29 β π 59 π 8 π¬ 5 π 1Current state of things: One Fox News host casually declares that Epsteinβs money comes from βJewish billionairesβ and a βJewish banking family,β another responds that he was a βsex rabbi.β www.mediamatters.org/fox-news/jes...
13.02.2026 12:25 β π 1984 π 674 π¬ 143 π 175Weird I guess there are a lot of people who donβt like giant tech companies with the power of nation states following their every move?
12.02.2026 23:34 β π 9 π 1 π¬ 0 π 0Episode 1 is OUT!! π₯³ π
This week on The Interface, we discuss:
* That insane Superbowl ad for Ring
* The US TikTok takeover
* And the beef that launched ChatGPT
Listen wherever you get your podcasts: app.magellan.ai/listen_links...
Today every story is a technology story. Our goal is to give back the calm, clarity, and control you need as the tech companies manipulate you (and each other) in their quest to dominate the world
All in all, a great day for anyone who has a phone
Youβre still here which means youβre mildly interested in what I have to say. Well I have great news
I've also got a brand new tech column with the BBC called Keeping Tabs. It's about the hidden systems that run your digital life, and how you can live better inside them www.bbc.com/future/tags/...
Weβre not talking about stock prices, no AI hype, no endless gadget talk. Weβre reaching behind your screen to pull out the stories that actually matter in your daily life
Available everywhere including YouTube, new episodes Thursdays. Subscribe!!!π app.magellan.ai/listen_links...
The Interface is a new BBC podcast hosted by me @karenhao.bsky.social & @nickywoolf.bsky.social. It's about the biggest stories in tech but it is NOT BORING. You'll love it even if you hate technology
Karen is already bullying me about my screen time. Please stop her www.youtube.com/watch?v=jn_W...
I just launched a brand new tech column AND a podcast with the BBC. The show's called The Interface, new episodes every Thursday
With your support, Iβm prepared to answer the biggest question in tech: what is a computer??
Let me tell you all about it ~~
app.magellan.ai/listen_links...
Weβre not talking about stock prices, no AI hype, no endless gadget talk. Weβre reaching behind your screen to pull out the stories that actually matter in your daily life
Available everywhere including YouTube, new episodes Thursdays. Subscribe!!!π app.magellan.ai/listen_links...
The Interface is a new podcast from the BBC hosted by me @_KarenHao & Nicky Woolf. It's about the biggest stories in tech but it is NOT BORING. you'll love it, even if you hate technology
Karen is already bullying me about my screen time. Please stop her www.youtube.com/watch?v=jn_W...
It hit me that there isn't a single service that shows you ALL the restaurants in an area. Google Maps won't show me all the spots on a street near my house no matter how far I zoom in. Every other app is the same
What if god forbid I want to make my own decision instead of relying on an algorithm?
Here's what's crazy. There DEAD SIMPLE techniques you can use to protect yourself from some of this tracking. Most people just don't the time. You can follow the steps I've got at the bottom of this article in 5 minutes
/end
www.bbc.com/future/artic...
According to one estimate, TikTok already collects data from 5% of the world's top websites, and that number is growing. I'm not just talking about what shoes you're thinking about buying. The company is getting data about your health, your finances, and your deepest darkest secrets
/4