Podcast Summary: This Week in Startups E2225

INSIDE How AI Startups Hire — AI Roundtable with Wade Foster, Mikey Schulman, and Ali Ansari
Host: Jason Calacanis
Guests: Wade Foster (Zapier), Mikey Schulman (Suno), Ali Ansari (Micro One)
Date: December 18, 2025

Episode Overview

This episode inaugurates Jason Calacanis’s new “This Week in AI” roundtable series, featuring candid, in-depth discussions with leading AI founders and builders. The conversation covers the intense AI talent wars, hiring challenges, effects of AI on the workplace, evolutions in developer and product workflows, the importance (and future) of training data, and the societal implications of AI’s rapid acceleration. The episode closes with a “lightning round” on the slow emergence of fully-automated agents and tips for breaking into the AI startup world.

Key Discussion Points & Insights

1. AI-Driven Social Trends & Parenting Challenges

• Isolation and technology:
  • Jason notes increased social isolation, attributing this to both COVID and the addictive nature of technology.
  • Mikey asserts that social media predates AI in driving isolation. However, he recognizes that AI could provide “a median competent therapist for the world,” and improve mental health access.
  • “It’s just somewhat incongruent to say that the correct solution to that problem is to ban AI. You’re probably gonna make things worse.” — Mikey Schulman [00:11]
• Raising kids in the AI age:
  • Wade emphasizes community involvement, promoting activities like sports and music over screen time.
  • The consensus: Parents must be thoughtful about encouraging in-person interactions and teaching responsible tech use.
  • “You have to figure out how to introduce them to the good parts of these tools and teach them how to navigate the things that are more challenging.” — Wade Foster [69:08]

2. AI Talent Wars: Hiring and Retention in AI Startups

• Recruiting in an aggressive market:
  • The panel discusses how the talent market has been upended by the likes of Meta and OpenAI, with huge salary offers and changes to standard equity vesting structures.
  • Mikey argues that total compensation expectations have shifted, making it increasingly challenging for smaller startups to compete, pushing the pitch toward mission and company culture.
  • “If that means taking a significant pay cut compared to what Meta or OpenAI is going to pay you, it is a less compelling pitch.” — Mikey Schulman [07:56]
• Pipeline & Vetting:
  • Ali notes that lowering barriers to entry means both a bigger supply of applicants and the need for more rigorous vetting, especially to parse out real skills from AI-assisted responses.
  • The “bottleneck” is most severe for elite AI researchers, particularly in SF. Remote and global hiring continues to be crucial.
  • “Most people can kind of do the take-home exercises using AI… It’s hard to distinguish whether they have actually done anything themselves.” — Ali Ansari [09:03]
• Retaining Talent:
  • Wade shares that Zapier’s competitive edge comes from company longevity, remote work flexibility, and being at the intersection of AI and integrations, which excites ambitious engineers.
  • Gen Xers and managers need to “be in the trenches” with the new tech, not just manage by outdated playbooks.

3. Impact of AI Tools on Workflow and Company Structures

• Developer pipeline & democratization:
  • All guests see more “developer-curious” employees prototyping and automating tasks with new no-code/low-code AI tools, shortening iteration cycles and enabling other functions (sales, product, ops) to “vibe code” and create value.
  • Internal tools and micro-problems are increasingly solved without involving core engineering resources.
• Hiring & Interviewing Evolves:
  • Technical interviews increasingly focus on “build something with AI, explain your thought process,” rather than DSA grilling.
  • Ali: “We’ve moved away from these DSA-type coding problems… We take that thought process and analyze it.” [21:28]
• The “Singularity of Startup Roles?”
  • There’s a trend toward “problem identifiers and solution builders” — people who can spot an issue, prototype a fix, and explain their decision process.
  • “Each of those domains, the singularity… will be this idea of running an AI agent, basically.” — Ali Ansari [21:59]
  • Wade: “The skill that is really valuable these days is not how much you know, it’s how adaptable you are.” [22:59]

4. Best Advice for Breaking into AI Startups

• Spec Work Over Resumes:
  • Mikey and Ali both hire engineers who submit unsolicited prototypes relevant to Suno/Micro One’s work. This has become far more effective for both candidate and company.
  • “Do the job that you want, and somebody will hire you.” — Jason Calacanis [28:39]
• AI interviews benefit the less experienced:
  • Ali shares data: Younger/less-experienced candidates do better in anonymized AI interviews, shifting power away from resume-driven decisions.
• High-agency is in vogue:
  • Wade: “That sort of like high-agency creativity… stands out even more so today.” [24:51]

5. The Training Data Frontier: Scarcity & New “AI Expert” Roles

• The Era of Infinite Scraping is Over:
  • Ali: “There is no data left,” at least not readily accessible, high-value data. Net-new, high-fidelity data now must be generated by subject-matter experts.
  • “The current state is… net new data is required and that net new data is created by humans.” — Ali Ansari [30:43]
• High-value Verticals:
  • Biggest demand for new training data: Finance, medical, legal, and coding.
  • The next challenge: sequential task accuracy (e.g., filing taxes) quickly compounds model errors.
• Rise of the Human Data Expert (HDE):
  • New career path: training and supervising AI models, especially for high-stakes tasks, often as a side hustle for domain professionals earning significant income.
  • “There’s over 200,000 people doing these types of jobs now… in the US.” — Ali Ansari [35:27]
• Company Data as “Alpha”:
  • Zapier leverages 14 years of workflow data for proprietary product improvements. Most AI companies are reluctant to license internal data to LLM providers, but may release benchmarks or evals to encourage LLM improvement in their problem space.

6. The Music Industry as an AI Battleground

• Licensing and Partnership vs. Litigation:
  • Mikey notes the music industry’s reputation for being litigious, but believes “there’s more to do together than fighting one another.”
  • Splice and other sampling sites provide one model; Suno hires musicians and also focuses on mission alignment to attract talent.
  • “We’re the GLP1s of the music industry. Everybody’s on them, but nobody wants to talk about it.” — Mikey Schulman [43:15]
• AI-Created Artists Go Mainstream:
  • Real-world case: Zanaya Monet, a poet and Suno user, signed a real record deal after using Suno to create music [46:52].
  • Trending music on Suno “trending page” drives discovery; more democratization of authorship and recognition.

7. Policy, Regulation & The "Accelerate/Decelerate" Debate

• Response to Bernie Sanders' AI Skepticism:
  • Bernie’s core arguments: AI and robotics will create massive job loss; billionaire control is dangerous; calls for a “moratorium” on data centers.
  • Panel consensus:
    • AI will change the job landscape, but historically technology has created new opportunities and improved quality of life.
    • “If we fully stop and really implement the solution, China will win the AI race and we will be forced to use Chinese models… which we certainly don’t want.” — Ali Ansari [60:12]
    • “It just feels very short sighted to say, hey, we’re going to pause data center construction because of that.” — Wade Foster [61:51]
• Alternative Solutions:
  • Rather than slowing innovation, Jason argues the tech industry should focus on communicating AI’s impact on the three biggest U.S. cost burdens: education, healthcare, and housing.
  • “All of them could be disrupted in a major way by AI… Our rallying cry should be to communicate those three areas of focus.” — Jason Calacanis [62:39]

8. AI and Social Isolation (Bernie’s Curveball)

• Are kids becoming too isolated due to AI?
  • The panel mostly agrees this is a real issue, but ascribes more blame to social media than AI per se.
  • “Imagine you can give the entire world a median competent therapist with AI — that’s probably good… The correct solution is not to ban AI. You’re probably going to make things worse.” — Mikey Schulman [66:33]
• Promoting Human Flourishing via AI:
  • Suno’s company mantra: “Fun is underrated.”
  • AI can free up humans for the most fulfilling, creative, or meaningful parts of work and life. Story after story about people making friends, finding joy, or launching entire new careers through AI tools.

9. The "Year (Decade) of the Agent" — Where Are We?

• Why haven’t agents taken over routine work?
  • Wade: Despite the hype, reliability is the bottleneck. Every additional step in a chained process adds risk of failure, so “workflow + embedded agent” is the effective paradigm for now.
  • “The year of the agent is probably more like the decade of the agent, but this year was chapter one.” — Wade Foster [73:03]
  • Best domains so far: coding, customer support.
  • Example: Zapier uses agents to analyze call transcripts and generate coaching, FAQs, testimonials, case studies, and even new landing pages or ads automatically [76:54].
• **Mikey and Ali agree that contextual evaluation and workflow design will be major budget items and job functions for enterprises in 2026 and beyond.

Notable Quotes & Memorable Moments

• On the value of adaptability:

  • “The skill that is really valuable these days is not how much you know, it’s how adaptable you are.” — Wade Foster [22:59]

• On building for fun and fulfillment:

  • “Fun is underrated, and there’s actually not enough time in this ecosystem spent trying to promote human flourishing.” — Mikey Schulman [66:33]

• On the new hiring process:

  • “Do the job that you want and somebody will hire you.” — Jason Calacanis [28:39]
  • “Our first Android engineer sent us a little app that he made… and that was our first Android engineer.” — Mikey Schulman [26:23]
  • “We actually hired our last two AI engineers also in a very similar fashion… built a prototype and said, look, I reduced the latency in this one specific part.” — Ali Ansari [27:03]

• On AI’s impact on society:

  • “If we don’t build [data centers], everybody is worse off… If we don’t, there won’t be anything to redistribute.” — Mikey Schulman [57:41]
  • “I think the important thing to do is… try to mitigate the risks of job loss, but as importantly, analyze the current state — and the current state is very beneficial to humanity.” — Ali Ansari [60:12]

• On breaking into startups:

  • “That high-agency creativity… stands out even more today.” — Wade Foster [24:51]

Timestamps for Important Segments

• [00:11] Social isolation, parenting challenges with AI
• [06:50] Salary wars: Meta & OpenAI’s hiring tactics
• [09:03] Challenges in developer vetting, researcher bottlenecks (Ali)
• [12:16] Retaining talent at Zapier, axes of the talent market (Wade)
• [17:39] How AI tools change product and prototyping cycles (Mikey)
• [19:47] Design bottlenecks > engineering, emergence of internal tools (Ali)
• [21:28] Evolution in technical interviewing and singularity of startup roles
• [24:51] Advice for new grads: Build, be high-agency, and pitch directly (Wade, Mikey, Ali)
• [28:39] The value of spec work and practical hiring changes (Jason)
• [30:43] State of training data: Scarcity, verticals, and rise of HDEs (Ali)
• [37:19] Zapier’s approach to proprietary workflow data (Wade)
• [42:25] Navigating the music industry as an AI company (Mikey)
• [46:52] First success stories of “AI-native” artists (Mikey)
• [50:44] Bernie Sanders’ anti-AI rhetoric, job loss debate
• [55:46] Panel replies: history, new opportunities, acceleration vs. deceleration
• [60:12] Why a data center moratorium is a strategic mistake (Ali)
• [62:39] Tech industry’s communication problem & proposed focus: education, healthcare, housing (Jason)
• [66:33] AI and social isolation: Panel perspectives, practical parenting (Mikey, Wade)
• [73:03] The decade of the agent: current status, reliability issues (Wade)
• [76:54] Real-world agent use cases: automating intelligence from call transcripts (Wade)
• [80:37] Evaluating agents and rise of enterprise evaluation (Ali)
• [82:51] Throwback: Founders pitching in coffee shops, encouragement to oddballs (Jason, Ali)

Final Takeaways

• AI hiring is fiercely competitive, but small startups win with culture, mission, and adaptability.
• Democratization of product creation means anyone can (and should) prototype — and that’s now the path into the industry.
• Training data value has shifted to net-new, high-fidelity, human-expert-created sets, spawning a new freelancing industry.
• Agents are useful, but their fully-automated, reliable deployment will take time and require new skills and organizational structures.
• Policy responses must focus on harnessing AI for societal good, not blanket restrictions. The technology industry’s communication is key.
• “Fun, fulfillment, and human flourishing” must not be sacrificed in the drive for productivity.

Want to work with these founders?

• Micro One AI: microone.com
• Suno: suno.com/jobs (or Google it!)
• Zapier: zapier.com/jobs
  Best advice: Just build/walk in with something relevant to show!