David Bitner

Harvard Business Review just published a piece. A good AI agent needs a job description, limits, and a manager. Because, AI agents can fail like employees with too much access and too little supervision. firms keep treating agents like normal software, even though the real risk is not bad text but bad actions. That changes 4 things: each agent needs its own identity and permissions, its own trusted data sources, hard rule checks between a model and any real transaction, and a full audit trail of what it read, decided, and did. So the safe rollout path is an autonomy ladder where agents start with drafts and recommendations, then move to guarded retrieval, then supervised actions, and only later get narrow bounded autonomy.

Can't decide which is crazier - the guy who thinks he could beat a grizzly bear, or the guy who thinks he'd lose a fight to a rat.

Brilliant paper! AI automate tasks, not jobs. The impact of AI on a given job, thus depend on whether the AI-exposed task is easy to unbundle from other job-tasks or not.

Ever since Coursera courses became very good, only to have basically no impact on higher ed, I've come to the belief that the biggest role of formal ed is to act as a commitment device for students to actually consume the information. AI will be transformative, but AI w/o that commitment element will likely lead folks like you and I to learn more efficiently (but we were going to do the learning anyway), and have little impact otherwise.

In the United States car seat laws mandating use until age 8 save 57 lives each year at the cost of about 8,000 births.

Engineering job openings are at the highest levels we’ve seen in over 3 years There are over 67,000 (!!!) eng openings at tech companies globally right now, with 26,000 just in the U.S. We don’t know if there would have been more open roles if not for AI or if AI is actually leading to more open roles, but since the start of this year, the increase in open eng roles is accelerating even more.

So far, entry-level share of employment in legal, financial, & office admin occupations doesn't look functionally different from pre-2023 trends (the population is aging so important to benchmark). A bit above trend for legal, a bit below for finance, in line for office admin.

GS: Over the Past 40 Years, Even Major Oil Price Shocks Did Not Leave Long-Lasting Effects on Consumer Inflation Expectations

"AI is going to cure cancer and create abundance for everyone" Meanwhile in Ireland:

After a brief hiatus, new post with @soumitrashukla9: "How Will AI-driven Automation Actually Affect Jobs? The economics of AI exposure and job displacement" There has been a lot of discussion in the media, X, substack, etc about AI driven displacement. We felt like it'd be worth working out the actual economics of when AI automation will actually lead to displacement, versus the exact opposite (more hiring, higher wages). A short summary🧵: AI "exposure" measures are not meant to predict displacement or job automation. Exposure can lead a job loss, or it can lead to more hiring and higher wages. It all depends on how 1) automated tasks interact with non-automated tasks (to what extent they're complements), 2) how consumer demand in that sector responds to prices (elasticity of consumer demand), and 3) the dimensionality of the job (the number of tasks a job has). One conclusion: we should be less worried about consultants and more worried about truckers and warehouse workers than we currently are. Link: substack.com/home/post/p-19…

People ask about muni’s and I typically don’t have much to say. But now, looking at the deficits caused by absurd spending, and tax policies accelerating revenue erosion, I can say avoid all General Obligation muni’s in California, Illinois & New York. I’ve never bought a GO.

Message to the jihadists. Chuck Norris is in heaven. There are no more virgins.

Goldman Sachs released a report analyzing how AI will shift the job market. - Found that AI could automate tasks making up 25% of work hours in the US. - Globally this exposes about 300mn jobs to some level of automation. - This transition is expected to take around 10 years to fully unfold. - During this period roughly 6% to 7% of workers might face displacement. If this happens smoothly the unemployment rate would only rise by 0.6%. - But the technology will also create entirely new categories of work. - Building the physical infrastructure for AI requires massive labor. The US alone needs 500,000 new workers by Dec-30 to handle electrical power demands. - Construction jobs related to data centers already grew by 216,000 since Oct-22.

JPMAM: After 20+ years of stagnation, the US is finally upgrading its electric grid

Observation by @jasonfurman: If you zoom all the way out, the Trump economy’s first 14 months look very very very similar to the Biden economy’s last 14 months, but most people have changed their mind about whether the economy is good.

Ben Thompson makes the case for why the rise of AI agents justifies $650B hyperscaler capex spend. The first paradigm of GenAI was AI chatbots. It required a lot to compute to train but not as much on inference. High token demand required a lot of users. The second (reasoning) and third (AI agents) paradigms require much more inference tokens. Crucially, with AI agents, the number of power users required to drive token demand through the roof is a lot less than the number of people needed to do chat queries (eg. one person running 50 coding agents will be massive demand).

US financial institutions’ exposure to private debt @business via @soberlook $XLF $KBWB thedailyshot.com/subscribe/

Great piece on OpenClaw and AI anxiety “AI焦虑) in China by @tonypengcomms: recodechinaai.substack.com/p/ai-anxiety-i…

Indoor living area per person has roughly tripled in the US since 1950.

Studying where AI actually helps in science by modeling four stages of research. AI excels at screening vast possibilities, but human judgment remains key for new ideas, from @professor_ajay, John McHale, and Alexander Oettl nber.org/papers/w34953