Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
Ep. 215: Musk v. OpenAI Round 3, AI's Hot New Job, The AI Jobs Apocalypse Debate
TL;DR
The Musk v. OpenAI trial mostly exposed everyone, not just facts — Paul Roetzer says the biggest outcome so far is that private texts, emails, and board concerns from Sam Altman, Greg Brockman, Illya Sutskever, Microsoft, and Elon Musk are now public, while the actual verdict still feels wildly uncertain despite Musk seeking roughly $150 billion and structural changes at OpenAI.
Forward deployed engineers are really AI-era consultants with outcome pricing power — OpenAI launched a $4 billion “deployment company,” bought consulting firm Tomorrow for 150 FDEs, and joined Google Cloud, Salesforce, and others in betting that embedded technical teams can redesign workflows, deploy agents, and charge against business outcomes rather than just software seats.
The AI jobs backlash is no longer theoretical — it’s showing up in data and culture — Roetzer points to SmarterX survey data where 71% of respondents now think AI will eliminate more jobs than it creates over the next three years, up from 40% in 2023, while Gallup found just 18% of people ages 14 to 29 feel hopeful about AI.
A lot of prominent voices are dismissing the jobs apocalypse — and Roetzer thinks they may be giving false hope — Scott Galloway, Andrew Ng, Andreessen Horowitz, and Derek Thompson all argued this week that AI fears are overstated, but Roetzer counters that flat headcount, AI-first hiring, and examples like General Motors cutting 600 IT jobs point to real displacement, especially for non-AI-forward workers.
Young people are booing AI in public, and that may become a career problem — the hosts highlight viral commencement backlash against Eric Schmidt and Gloria Caulfield as evidence that students increasingly see AI as threat rather than opportunity, which Roetzer says is understandable emotionally but dangerous if graduates refuse to become AI literate.
The bigger pattern is that AI is reorganizing work, software, and security all at once — from Anthropic warning 2026 may be the breakaway year in the US-China race, to Google documenting AI-assisted zero-day exploits, to Shopify’s internal Slack agent and Claude-based “second brain” workflows, the episode argues we’re already moving into a more agentic operating model.
The Breakdown
Musk, Altman, and the courtroom spectacle
The show opens on the Musk v. OpenAI trial as the jury deliberates, with Mike recapping bombshell testimony from Illya Sutskever, Satya Nadella, and Sam Altman. Sutskever says he spent a year compiling evidence of Altman’s “consistent pattern of lying” in a 52-page document, while Altman describes Musk’s old push to fold OpenAI into Tesla and even pass control to his children. Paul’s big takeaway: whatever the legal outcome, the trial has already succeeded in dumping a mountain of private messages and ugly power dynamics into public view.
Why “forward deployed engineer” suddenly became the hottest AI job
The next segment zooms in on FDEs after OpenAI launched its new “deployment company” with more than $4 billion behind it and acquired Tomorrow, bringing in 150 applied AI consultants from day one. Paul’s translation is blunt: this is basically consulting with stronger technical chops, and the real prize is outcome-based pricing tied to business value, not just software licenses. He calls FDEs a bit of a Trojan horse — useful, yes, but also a way for labs and cloud companies to go directly after the massive services revenue that traditionally sat with firms like Deloitte, Bain, and McKinsey.
The real shortage isn’t just technical talent
Paul keeps pushing the point that companies don’t only need engineers who can wire up models and agents — they need people who can handle change management, training, communication, and adoption. He cites Salesforce’s own framing of FDEs as part engineer, part business consultant, part hand-holder, and notes that SmarterX is seeing the same need in AI Academy work. The most memorable line here is that AI vendors aren’t just chasing software revenue; they’re chasing the much bigger pool of knowledge-work labor spend, which Paul pegs at roughly $4 trillion to $6 trillion in the US alone.
The “jobs apocalypse is overblown” backlash gets a full airing
Mike rounds up a cluster of anti-doomer arguments from Scott Galloway, Andrew Ng, Andreessen Horowitz, and Derek Thompson, all of whom say AI job fears are overhyped or historically misguided. Paul is openly skeptical, saying it felt like “an email went out” telling everyone to counter the jobs narrative at once. He argues these takes ignore what leaders are actually doing inside companies right now: keeping headcount flat, hiring only AI-forward talent, and quietly expecting fewer people to produce the same or more output.
Why Roetzer thinks the job disruption is already visible
To make the case, Paul cites SmarterX’s new State of AI for Business report: 71% of respondents now believe AI will eliminate more jobs than it creates over the next three years, up from 53% last year and 40% in 2023. He points to General Motors cutting 600 IT roles while explicitly looking for AI-native replacements, and to Citadel CEO Ken Griffin flipping from calling AI “garbage” in Davos to saying weeks later that work once done by people with master’s degrees and PhDs is now being done by agents in hours or days. His core message is not that abundance is impossible, but that pretending the next one to five years won’t be messy is irresponsible.
The AI hate wave arrives, and students are at the center of it
The mood gets heavier when they discuss Axios’ “AI hate wave” story, including Gallup data showing only 18% of people aged 14 to 29 feel hopeful about AI and viral graduation speeches where Eric Schmidt and Gloria Caulfield got booed for praising it. Paul wonders if years of schools framing AI as cheating helped create this sentiment, and says watching Schmidt get drowned out even while trying to talk about cancer research felt “dystopian.” His most urgent point: students can hate AI all they want, but if they refuse to learn it, they are going to run headfirst into a job market that increasingly expects AI fluency.
The agentic future is reshaping org charts and workflows in real time
In the later rapid-fire section, the conversation turns from fear to structure: Anthropic warns 2026 may be America’s breakaway year in the AI race with China, Google reports cybercriminals are already using AI to develop zero-days, and software companies face a future where agents may use their products more than humans do. Paul also shares the framework behind his upcoming MAICON keynote — “the architect, the orchestrator, and the apprentice” — as a way to think about how expertise, management, and entry-level learning change when agents absorb more foundational work. The closing examples from Shopify, Obsidian, Claude, and internal SmarterX workflows make the episode’s bigger point feel very immediate: this isn’t a future trend anymore, it’s the new operating environment.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
