In partnership with
In Today’s Issue:
🔀 OpenAI and Anthropic split on AI jobs
⚡ Nvidia's Vera CPU targets agentic workloads
⚠️ Epoch warns of a possible compute crunch
🧬 Retatrutide pushes obesity-drug results toward surgery territory
✨ And more AI goodness…
⚡ The Signal
Today's issue is about AI reality colliding with AI rhetoric. Sam Altman now says the white-collar "jobs apocalypse" he once worried about has not arrived as quickly as expected, while Anthropic-linked voices are still warning that large-scale labor displacement remains possible. That split matters because companies, workers, and policymakers are being asked to plan for a technology whose boosters cannot agree on whether the near future is boom, doom, or something messier.
The quieter story is infrastructure. Nvidia is pushing CPUs built for agentic workloads, Epoch is warning that token demand could outrun inference supply, and Anthropic's own usage patterns show agents concentrating first where work is digital, testable, and tool-heavy. The result is not one clean narrative. It is a constraint map.
All the best,
Kim Isenberg
🧭 OpenAI and Anthropic Split on the AI Jobs Story
Axios says the top AI labs are now giving the public sharply different labor forecasts: Anthropic voices are still warning about large-scale displacement, while Sam Altman says the feared jobs apocalypse looks less likely than he once thought (more on that in the main article). The useful takeaway is not that one side is obviously right. It is that the people building the technology are still arguing about the slope of the disruption curve.
👉 tl;dr: AI labor risk is real, but the CEO narrative has become unstable enough that companies need evidence, not vibes, before making workforce decisions.
🧱 Nvidia's Vera CPU Makes the Agentic AI Bet Physical
Nvidia says early Phoronix testing shows its Vera CPU is tuned for the unglamorous work behind agentic AI: sandboxes, runtimes, compilation, databases, memory bandwidth, and orchestration. The pitch is that AI factories need more than GPUs once agents start running many tool-heavy workflows in parallel.
👉 tl;dr: The agent boom is also a CPU and memory-bandwidth story. If agents are going to act all day, the infrastructure around the model has to stop being the bottleneck.
⚡ Epoch Warns Token Demand Could Outrun Supply
Epoch AI estimates that global inference capacity is growing fast, but token demand may be growing much faster - especially for long-context and agentic workloads. Their model suggests today's chip base could serve a huge range of output-token capacity, but demand proxies point to a possible near-term squeeze.
👉 tl;dr: The next bottleneck may not be whether models can answer. It may be whether the world can afford to run enough of them, fast enough, at frontier quality.
When an AI leader makes a sweeping jobs claim, ask your assistant to separate prediction, incentive, and observable evidence before you react.
Why it helps: The same company can benefit from making AI sound world-changing to investors, manageable to regulators, and safe to enterprise buyers. Those incentives do not make the claim false, but they do mean you should inspect the evidence underneath it.
Try this: Paste a quote about AI and jobs and ask: "What is the claim, what incentive does the speaker have to frame it this way, what evidence would confirm it, and what evidence would weaken it?"
🎬 Watch This
I sat down with Robby Stein, Google’s VP of Product for Search, at Google I/O.
He helped build Instagram Stories and Reels. Now he is helping rebuild Google Search around AI.
We talked about AI Mode, AI Overviews, personalization, Google’s infrastructure advantage, agentic search, publisher traffic, source attribution, and the future of links.
The core question:
If AI gives you the answer directly, what happens to the web?
– Sam Altman, explaining that AI-driven entry-level white-collar job losses have not happened as quickly as he once expected, speaking at a Commonwealth Bank of Australia conference, as reported by TIME
Sam Altman Walks Back the AI Jobs Apocalypse
The Takeaway
👉 Altman now says he does not expect the kind of near-term jobs apocalypse some AI leaders have warned about.
👉 He told a Commonwealth Bank of Australia audience that entry-level white-collar displacement has been slower than he expected.
👉 Axios reports the divide between OpenAI's rosier tone and Anthropic's warning-heavy posture is now a public split in AI labor messaging.
👉 The real planning problem is uncertainty: workforce impact depends on adoption speed, compute cost, workflow redesign, and where agents actually work well.
Sam Altman is trying to put distance between AI's scary labor narrative and what has actually happened so far. TIME reports that the OpenAI CEO, who has previously warned that AI could replace huge categories of work, now says he is "delighted to be wrong" about the speed of entry-level white-collar job loss. The shift is important because it changes the message from imminent collapse to slower, messier diffusion.
But this is not a clean all-clear. Axios notes that Anthropic-linked voices are still emphasizing the possibility of large-scale labor displacement, while OpenAI is sounding more optimistic. That leaves companies in a strange position: the vendors selling AI are asking customers to believe the technology is powerful enough to transform work, but not so fast or so sharply that it should terrify workers and regulators.
The deeper constraint may be operational. If AI agents need long contexts, lots of tool calls, and constant inference, the labor story becomes tied to the compute story. AI can only replace or augment work where the workflow is digitized, the model is reliable, the economics work, and the infrastructure can keep up. That points to uneven disruption rather than a single jobs apocalypse.
Why it matters: The next phase of AI adoption will be shaped less by abstract predictions and more by where companies can prove reliable ROI. For workers, that means the risk is not evenly distributed across the economy; it is concentrated wherever digital tasks can be measured, automated, and scaled cheaply.
Sources:
🔗 https://time.com/article/2026/05/26/sam-altman-ai-job-losses-openAI-/
🔗 https://www.axios.com/2026/05/27/ai-hype-doom-openai-anthropic
🔗 https://epoch.ai/gradient-updates/is-a-compute-crunch-coming
The IT strategy every team needs for 2026
2026 will redefine IT as a strategic driver of global growth. Automation, AI-driven support, unified platforms, and zero-trust security are becoming standard, especially for distributed teams. This toolkit helps IT and HR leaders assess readiness, define goals, and build a scalable, audit-ready IT strategy for the year ahead. Learn what’s changing and how to prepare.
The chart: Anthropic’s data shows that nearly half of all agent tool calls are already happening in software engineering, far ahead of back-office automation, marketing, sales, finance, research, and customer service
The lesson: AI agents are not spreading evenly across the economy. They are first taking off where work is digital, modular, testable, and full of immediate feedback loops, which is exactly what coding provides.
The caveat: This does not mean software engineering is the only or even the biggest long-term agent market. It may also reflect Anthropic’s user base and the fact that code agents generate many tool calls, while other domains may use fewer tools but still create significant economic value.
Retatrutide Pushes Weight Loss Toward Surgery Territory
⚡ Bottom line: Eli Lilly's next-generation obesity drug retatrutide produced about 28% average weight loss over roughly 18 months in a late-stage trial, with nearly half of top-dose patients losing at least 30% of body weight.
💡 Why it matters: Thursday's longevity angle is not about a chatbot. It is about medicine pushing deeper into systems-level human health, where drugs can change metabolic risk at population scale.
🔎 What it means: The result raises the ceiling for obesity pharmacology, but the hard questions now move to tolerability, long-term safety, muscle loss, access, and whether regulators see enough evidence for approval.
Retatrutide is starting to look like the next big escalation in the obesity-drug race. Axios reports that the Phase 3 TRIUMPH-1 trial found participants on the top dose lost about 28% of their body weight over roughly 18 months, with 45% losing at least 30%. Scientific American notes that the weekly injection targets three hormone receptors - GLP-1, GIP, and glucagon - which may help explain why it is pushing beyond current approved GLP-1 drugs.
That is why the comparison to bariatric surgery is so important. Surgery has long been the high-water mark for major, durable weight loss, but it is invasive and hard to scale. A drug that can approach that range would change the treatment ladder for obesity, diabetes risk, cardiovascular risk, and other metabolic conditions.
The caveat is that efficacy is only half the story. Axios reports discontinuations due to adverse events on the top dose, while Scientific American notes gastrointestinal side effects such as nausea, diarrhea, constipation, and vomiting. If retatrutide is approved, its real impact will depend on who can tolerate it, who can afford it, and how doctors manage long-term tradeoffs.
“Who is this person again?”
You’ve had that moment. Walking into a call, scrolling through old emails, trying to remember what you promised. Lindy texts you a brief 15 minutes before: attendee context, past discussions, open items, talking points. All pulled automatically. Try Lindy free.