Worthwhile Content: March 2026
Some worthwhile reads, watches, and listens from the last month
World of Work:
In March, I wrote an April Fool’s Day post about extroverts, premiered ‘Tube Takes’ where I responded to your controversial workplace opinions not once but twice. I answered a reader’s question about inheriting an employee with a bad reputation but a spotless record, and responded to another reader calling me out for saying AI won’t take every job just moments before Block fired half their staff for “AI reasons”. Some other worthwhile content I digested in March includes:
A randomized trial in the Netherlands found that pairing unemployed job seekers with recently employed “buddies” who’d made successful career pivots raised employment rates by 6% and boosted monthly earnings by €226, with the biggest gains among the long-term unemployed. Peer support for job seekers has always struck me as underutilized, and this is one of the more rigorous tests of it I’ve seen.
The average number of direct reports per manager has climbed from 10.9 to 12.1 in just one year — nearly 50% higher than when Gallup first measured in 2013 — and their analysis of 44,000+ employees makes a pretty compelling case that the optimal team size for managers isn’t a fixed number, but a function of whether managers can actually give meaningful feedback.
It’s hard to square Gallup’s math with Jack Dorsey’s desire to have all 6,000 company employees reporting directly to him. I get the theory here and the desire to eliminate ‘middle-management’ and bureaucracy. I also don’t care how many AI agents he has helping him manage: I am willing to bet a significant amount of money that the upside from this approach would never outweigh the downside. If it happens, it will not last very long.
AI & Work
Zapier ran a pilot using an AI interviewer for initial candidate screening and came away with some genuinely surprising numbers: 97% completion rates, an average candidate rating of 4.5 out of 5, and 30% of candidates who advanced were people they wouldn’t have had capacity to screen otherwise. I’m skeptical of vendor-sponsored success stories, but the fact that they published the opt-in rate variance (35% to 81% depending on the role) and attributed it to their own communication failures makes me think there’s something real here.
This research puts AI coaching to the test and finds it falling short. Human coaching led to strong and consistent improvements, while AI coaching did not deliver similar outcomes, highlighting how important real human interaction still is. I guess I still have a job, for a bit? For a bit.
Another study on AI coaches tested three coaching styles inside AI chatbots and found that CBT delivered the strongest results across key outcomes like goal attainment and user engagement. It highlights that AI coaching may rely on different psychological drivers than human coaching, which could reshape how these tools are designed.
Hiring managers are skipping résumé review entirely as AI-generated applications flood inboxes and, per one recruiting expert, ‘the résumé is almost worthless because they all read the same.’ I’m not surprised — research has long suggested that résumé signals like prestige employers and credentials are weak predictors of job success anyway, so maybe the AI slop epidemic is just accelerating a reckoning that was overdue?
Migrating years of ChatGPT context to Claude — your tone, workflows, client knowledge — apparently takes about 20 minutes if you know what you’re doing, and this LinkedIn post walks through the exact export-and-parse process to make it happen. I’ve migrated 90% of my work to Claude, personally. And the underlying idea of treating your AI conversation history as a transferable asset worth preserving is one I hadn’t thought about before.
Heavy AI reliance in writing doesn’t just smooth out prose, but it also changes what people actually say. A study found that heavy LLM users produced responses that diverged significantly in meaning from less-frequent and non-AI users. The lead researcher calls it “blandification,” and the part I find harder to shake is that people reported feeling their work was less creative and less in their own voice, yet still rated their satisfaction about the same.
General Interest
This portrait of the incentive structures of prediction markets captured my thinking on the matter quite well, as Charlie Warzel often does. It’s worth reading for yourself here (gift link), because there is way too much at stake when people who can shape world affairs are incentivized to profit from them. The ‘wisdom of crowds’ may be an entirely real, better signal, but not worth the cost.
Bravo United Airlines! They announced that they plan to remove passengers who refuse to wear headphones while playing audio on personal devices, a policy change quietly buried in its contract of carriage and apparently triggered by the rollout of Starlink internet on regional jets. Now let’s do this everywhere, in public, all the time.
I adored this short piece on the ‘Crappy Gods Theory’. No, it has nothing to do with theism of any sort. Rather it’s a healthier way for us to view ourselves.
Musings
I’ve been busy book writing and editing. Boy… the book publishing process is an intense, murky one as a first-timer! But I selected a publisher (Practical Inspiration Press) and we’ve got a publication date: Feb 16, 2027. So get ready to be badgered beyond belief about this book over the next year! In the meantime, you can help me choose my title.