This guide synthesises Chris Barber’s AI Prep Notes, a series of conversations and interviews with leading thinkers on advanced AI (chrisbarber.co/AI+Prep+Notes | @chrisbarber). Generated by ChatGPT (o1, 4o canvas). Copied, pasted, prompted, and lightly edited by Eleanor Berger (intellectronica.net).
- Key Predictions (2025–2035)
- Economic Disruption & Career Shifts
- Preparing Yourself Financially
- Reskilling and Adopting AI Tools
- Advice by Persona (Students, Parents, Programmers, etc.)
- Conclusion & Next Steps
“We’re entering an era in which small teams—or even individuals—can unlock billion-dollar value using AI. You can treat this future as scary, or you can treat it as an opportunity.” —Tyler Cowen
“I think five years from now, AI will be capable of doing all knowledge work jobs.”
—Flo Crivello (Flo Crivello interview)
A recurring theme in Chris Barber’s conversations is that remote, well-defined computer-based tasks will be among the first to be automated at a superhuman level. Lawyers, programmers, accountants, and consultants—anyone in a domain where the tasks and deliverables can be precisely defined—will see rapid progress by AI.
- Timeline: Multiple interviewees (e.g. Jaime Sevilla of Epoch AI) expect these advanced AI “drop-in remote workers” to become viable by 2030–2035, though the adoption curve may lag behind capability.
- Capability vs. Adoption: Even if the AI can do the job, large companies often move slowly. Bureaucratic inertia, regulation, and risk management mean not everyone will embrace AI the minute it arrives.
“I don’t expect drop-in remote workers by 2030. AI can do many parts of knowledge work, but tasks requiring high-context environment or deep relationships with clients might need more time.”
—Tamay Besiroglu (Tamay Besiroglu interview)
A major thread is the rise of reasoning-focused AI. Large models can already write code and solve math puzzles at near-expert levels. With increased reinforcement learning + verifiable rewards (RLVR), experts such as Finbarr Timbers (ex-DeepMind) predict a jump to near-superhuman performance in tasks like coding and math.
- Coding: AI coding tools could soon produce more efficient code than average professionals. Reinforcement learning could slash the bug rate or drastically improve performance over the next 2–5 years.
- Complex Analysis: Where tasks are well-defined (like pure maths or certain forms of engineering), expect superhuman results much sooner than you’d think. Harder contexts—strategy, ambiguous tasks, or open-ended creative processes—may require more time (though they too are “on the menu”).
“It will feel very dumb for humans to be writing code in 2030, but we’ll still need plumbers.”
—Ivan Vendrov (Ivan Vendrov interview)
In consumer-facing robotics, multiple experts foresee a longer horizon:
- Eric Jang (1X Robotics) predicts that some wealthy households might adopt humanoid robots in the early 2030s—partly for novelty, partly for real utility—but mass adoption could take 10+ years.
- Physical manipulation is far harder to master than coding, so manual labour (e.g. carpenters, plumbers, mechanics) is safer from displacement for the moment.
“It’s not going to be a big sudden apocalypse for software engineers. The transition will be about new tools, so you’ll do 5–10× more with fewer people.”
—Finbarr Timbers (Finbarr Timbers interview)
A consistent message: Even as AI floods in, human labour remains valuable where:
- Physical tasks are involved (e.g. HVAC repairs, hairdressing).
- Human-to-human interaction is prized (sales, client management, healthcare).
- Trust and liability require a “neck to throttle” (someone to hold accountable).
- Complex context or deep cultural knowledge is needed.
Key disruptors:
- Contractors & Freelancers: Repetitive or well-defined tasks (“Upwork style,” as Timbers says) could be automated more quickly.
- Junior roles: Mentoring new hires is high-overhead. AI can handle their tasks cheaply, so new graduates might find it harder to break in.
“We might see a self-improvement feedback loop if AI is used to build better AI hardware and software. But does that happen overnight or gradually? I lean on a gradual takeoff.”
—Jaime Sevilla (Jaime Sevilla interview)
Some experts believe we could see a sudden “fast takeoff,” with AI self-improvement spiralling quickly. Yet the consensus in these interviews is that diffusion (the time it takes to adopt and integrate new technology) may limit the speed of labour displacement—leading to a somewhat slower social transition than the raw capabilities might suggest.
“We’ll see 2–3% per capita GDP growth continue, perhaps higher, but not 30% next year. Over decades, that steady improvement is transformative.”
—Finbarr Timbers
While a few see potential for hyper-growth, the typical scenario is accelerated but not instantaneous. Firms that leverage AI effectively become lean and more productive:
- Old, bloated companies may shrink.
- AI-first startups could rise quickly (the mythical “one-person billion-dollar company”).
“Brands and personal reputation matter more than ever, because if 90% of tasks are automated, the remaining 10%—human trust—will be precious.”
—Tyler Cowen (Tyler Cowen interview)
- “Well-defined” white-collar roles: E.g. reviewing contracts, writing standard code, drafting basic legal documents.
- Junior/Entry-level knowledge roles: If a single senior can orchestrate an AI to do the rest, the pipeline for new hires shrinks.
- Interpersonal / Relationship: Sales, negotiations, therapy, management requiring a “human face.”
- Physically Grounded: Skilled trades, on-site roles, tasks requiring dexterous real-world problem-solving.
- High-Level Strategy / Creative Vision: Yes, AI will help with ideas, but humans still direct overall objectives, at least for now.
“Think of it like the excavator. One operator can replace 20 labourers with shovels, but you still need the operator to decide what to dig and what to build.”
—Finbarr Timbers
Multiple voices (Ivan Vendrov, Flo Crivello, Tyler Cowen) foresee a world in which:
- Elite “AI native” workers: Highly flexible, quickly adopt new tools, manage AI.
- Lagging workforce: Many remain in slower-adopting organisations or roles, but the wages in those roles might stagnate.
“Get used to 996. Firms that want to stay competitive can’t have everyone punching out at 4. If you want to avoid that, you’d better have capital or unique skills.”
—Eric Jang (Eric Jang interview)
“One big question is: do you hoard cash before the singularity? But if nobody wants your money in the post-scarcity future, that’s pointless.”
—(Paraphrasing David Holz, via Eric Jang)
- Semiconductor & Compute: Many experts (e.g. Jaime Sevilla, Tamay Besiroglu) suggest that HPC (high-performance computing) capacity is the essential resource that AI labs vie for.
- Broad Equity Market: A bet on overall growth from AI.
- Diversification: Because uncertainty is high, many recommend not going all-in on any single stock or theme.
“I do a 50-50 between semiconductor ETFs and broad market ETFs. We may see hypergrowth from AI-driven cycles, but it’s risky to pick winners individually.”
—Jaime Sevilla
- Some, like Tamay Besiroglu, suggest “Save more, especially if you’re young—future labour income may be uncertain.”
- Others (e.g. Eric Jang) argue money is only one resource: community, networks, compute, real assets, brand equity can matter more.
- If you see massive future deflation in services, you might prefer investing in experiences today. (Approach with caution: no one truly knows.)
“We could see a new class of millionaires as small AI startups create huge value. But we could also see certain sectors hammered if AI outcompetes them overnight.”
—Chris Barber
Actionable Step:
- If you believe in big AI disruption, consider a small allocation to out-of-the-money call options (to capture growth) balanced with safer or diversified positions. (Not financial advice—just a recurring concept from the interviews.)
“Using AI is an actual skill, like coding. If you just type random English prompts, you’ll do poorly, blame the tool, and remain stuck.”
—Jeremy Howard (Jeremy Howard interview)
Practical Steps:
- Use AI daily: Start with ChatGPT/Claude/Bard or advanced coding copilots.
- Learn “Prompt Craft”: Study how experts prompt large language models (LLMs)—the difference in results can be staggering.
- Stay Updated: Subscribe to communities (Reddit, X, Slack channels) that share advanced usage tips.
“AI can do logic, but not—yet—smooth face-to-face persuasion with personal trust.”
—Tyler Cowen
Soft skills still matter. Develop:
- Emotional intelligence
- Networking
- Persuasion & leadership
If you see a way to drastically improve your field via AI, consider:
- Starting a micro-startup with 1–3 people.
- Working with advanced AI labs to build domain-specific training data.
“Small teams (1–3 people) building a product on top of advanced AI can create huge wealth. Start planning for that possibility now.”
—Tyler Cowen
“University is good for building a network, but for actual skills—side hustles and tinkering might matter more.”
—Jeremy Howard
- Learn to code (even if it’s just basic Python). AI will do heavy lifting, but you’ll need to orchestrate it.
- Project-based learning: Show real demos. Employers and future co-founders want proof you can deliver, not just a diploma.
- Stay agile: Majors might matter less in a fast-changing environment. Pursue experiences that teach versatile problem-solving and collaboration.
“Stop writing everything by hand. Let the AI handle the boilerplate. You focus on design, big-picture architecture, product sense.”
—Finbarr Timbers
- Use coding copilots (Cursor, Copilot, etc.) daily.
- Shift from “performer” to “conductor.” You’re orchestrating AI modules, verifying their work, handling product logic.
- Deepen domain knowledge. If you just push code lines, you’ll be replaced. Emphasise the design, product insight, or leadership side.
“Brand and personal reputation are huge moats. People trust humans they know.”
—Tyler Cowen
- Small, high-talent teams can generate enormous output with advanced AI.
- Build your brand: As the cost of production falls, marketing and brand recognition can be decisive.
- Iterate quickly: “AI years” pass faster than “internet years.” Launch, learn, pivot, use the best models available.
“Your child already has a world-class tutor for free—AI. The challenge is deciding at what age to let them use it.”
—Tyler Cowen
- Encourage creative exploration with AI. Let kids see it as a tool, not a threat.
- Rethink standard schooling pathways. The future economy may reward skills that typical schools don’t emphasise (entrepreneurship, real-world problem-solving, networking).
- Resist fear-mongering. It is normal for technology to reshape job markets over time.
“A lot of people who seem ‘low agency’ simply haven’t found tasks that excite them. New AI tools might unlock unexpected potential.”
—Tyler Cowen
- If you’re truly uninterested in “competition,” focus on roles that remain inherently human (e.g. caretaker, social roles).
- Many forms of manual or service-based labour (gardeners, carpenters, etc.) will remain in demand for a while, possibly at higher wages if the rest of the economy becomes more productive.
Chris Barber’s interviews showcase an emerging consensus:
- AI is accelerating rapidly but not instantly in every domain.
- Knowledge work is prime for automation—especially tasks that are well scoped, remote-friendly, or purely digital.
- Physical & interpersonal roles are safer, at least for a decade or two.
- Economic shifts could be large, but the timeline is uncertain.
- The best preparation is flexibility, adaptability, and a willingness to work with AI rather than ignoring it or reflexively fighting it.
- Experiment Constantly
- Sign up for new AI tools, especially domain-specific ones.
- Prompt them to solve your everyday tasks.
- Invest in Lifelong Learning
- Learn prompt engineering basics.
- Try a short coding course if you have never coded.
- Broaden Your Network
- Seek peers using AI intensively.
- Participate in online communities (Discords, GitHub projects, or local meetups).
- Assess Your Finance & Career
- If you believe in rapid growth, tilt towards broad market or semiconductor exposure (not financial advice).
- Save more if you fear labour disruptions.
- Don’t rely exclusively on standard corporate paths or degrees.
- Embrace “Human-ness”
- The more mechanical your role, the more likely AI can do it soon.
- Lean into tasks where being physically present, building relationships, or applying nuanced judgment matters.
“Think of it as the 1990s internet moment. Not engaging with AI now would be foolish.”
—Jeremy Howard
If you remember only one thing, make it this: become a builder or early adopter of AI tools. Even if your industry is conservative, the tinkerers and experimenters will shape tomorrow’s workflows—and your seat at the table might depend on it.
Good luck, and happy building.