Michel
@JustenMichel
Alive! And supporting journalism at the Tarbell Center for AI Journalism. Views my own.
The Economist cover story this week is a great wonky overview of the debate over ‘will AI cause explosive growth’:
I had a lot of fun researching and writing this piece: what if AI made the world’s economic growth explode? The macroeconomics of an extraordinary thought experiment: economist.com/briefing/2025/…
Trump administration to vet AI for ‘ideological bias’ on.ft.com/4537wRc
Excited to have supported this investigation with a Tarbell Grant! Stay tuned for another round of funding for investigative journalism on AI, launching soon.
🚨 Anduril, a defence tech company supplying the U.S. government with AI-powered weapons, has turned its attention to the UK – with plans to reportedly build a factory in the country to serve as a European base thebureauinvestigates.com/stories/2025-0…
fellas, it’s time to start populating the internet with datasets of 3-digit numbers that transmit love and peace
New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
SCOOP: Leaked memo from Anthropic CEO Dario Amodei outlines the startup's plans to seek investment from the United Arab Emirates and Qatar. “Unfortunately, I think ‘no bad person should ever benefit from our success’ is a pretty difficult principle to run a business on.”
SCOOP from Tarbell Grantee @rebheilweil: The US General Services Administration (GSA) is looking to add Grok to its list of AI tools for US Gov Employees, as seen in a formerly-public GitHub project.
NEW from me and @madialder: a GitHub shows that gov coders are trying to access Grok. a spokesperson for GSA told us the agency is interested in welcoming all American companies and models "who abide by our terms and conditions." fedscoop.com/grok-for-gov-g…
Our RCT found that [early-2025] AI coding assistants appear to *slow down* users [working in mature open-source codebases]. But developer self-reports (and expert forecasts) suggested speedup. This is a counterintuitive result! Some thoughts on interpretations / takeaways
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
Been a while since I’ve done this but *Personal News Klaxon*: In a couple of weeks I’m starting a new role as managing editor at transformernews.ai, working with @shakeelhashim.com to expand its coverage of how AI is set to transform huge swathes of society.