Silicon Valley GirlKhan Academy CEO: The Real AI Opportunity Is in Boring Industries | Sal Khan
CHAPTERS
- 0:00 – 2:25
AI is reshaping Khan Academy and why Sal Khan is now worried about “Job Shock”
Sal Khan reflects on how fast AI has moved since writing his education-focused book and why he’s now writing “Job Shock.” He explains that AI isn’t just a tutoring tool; it’s changing assessment, admissions, and the future of work—and society needs a plan.
- •Brave New Words ideas (AI tutoring, broader assessment) are becoming real
- •Universities using AI-assisted oral exams to verify applicants’ work
- •Khanmigo lessons: AI must be proactive and teacher-in-the-loop
- •Shift from optimism to concern: accelerating labor disruption
- •“Job Shock” as a call for societal adaptation and reskilling
- 2:25 – 8:24
Early warning signs: call centers, Waymo, and dislocation arriving in 3–10 years
Khan shares a concrete example of a call center automating 80% of work and connects it to national-level risk (e.g., Philippines’ BPO dependence). He uses autonomous driving as another visible “tip of the iceberg,” warning of political and social consequences if large groups can’t find work.
- •VC portfolio company automated most call-center roles with genAI
- •Macro risk for economies reliant on outsourcing/call centers
- •Autonomous driving likely dents jobs within 5–10 years
- •Job loss can fuel polarization and instability
- •Skepticism of UBI; preference for “universal basic work”/meaningful contribution
- 8:24 – 8:45
Which jobs are most at risk—and which are safer because they’re deeply human
The conversation expands from drivers and call centers to knowledge-work roles in tech, especially those that don’t adapt. Khan argues that jobs emphasizing human connection and coaching are more resilient, though even “safe” roles will change in how they’re performed.
- •Product roles (engineering/design/PM) are changing fundamentally
- •Risk rises for knowledge workers who refuse to adapt to new tools
- •Teachers as “coaches/architects,” not information dispensers
- •Nursing, hospitality, and relationship-driven sales as more durable
- •Anything purely computer-based without human interaction is most threatened
- 8:45 – 10:18
Will AI cause layoffs at Khan Academy? Productivity without cutting headcount
Khan addresses employee anxiety directly: if AI makes the team 3x more productive, will staff be reduced? He explains why Khan Academy’s constraint is funding, not productivity, and how they’re formalizing adaptability as a career expectation.
- •Khan Academy: ~350 staff, ~2/3 product roles
- •Leadership message: “use more AI,” not less
- •AI productivity will be used to do more work, not trigger layoffs
- •Layoffs (if any) would be driven by revenue/philanthropy shortfalls
- •Career rubric updated to reward learning tools and adapting roles
- 10:18 – 14:01
Tech roles are blending: designers and PMs moving into code, engineers moving outward
Khan predicts the traditional separation of designer/PM/engineer will blur as AI speeds prototyping and development. Khan Academy is already giving non-engineering roles dev environments, while engineers become more customer-facing and versatile.
- •Historic web-app org model is becoming outdated
- •PMs expected to prototype; designers/PMs gaining dev access
- •Engineers remain valuable but become far more productive
- •New “hybrid” roles emerge (deep technical + customer-facing)
- •Development cycles accelerate, increasing organizational velocity expectations
- 14:01 – 14:40
AI tooling that felt impossible 2 years ago: multi-agent coding and AI code review at scale
Khan describes day-to-day engineering changes: multiple agents running in parallel for writing and reviewing code. He notes Khan Academy is among top users of agent-based code review, reshaping how software is built.
- •Engineers running many simultaneous agents for development tasks
- •Agent-based code review becoming a core workflow
- •AI-driven rapid prototyping is now routine
- •Old cycle (“design a lot before building”) is shifting toward fast build/test
- •Velocity gains feel like 50–100% vs. a few years ago
- 14:40 – 17:44
The $1.2M/year Anthropic bill: when expensive tokens are still an ROI win
Khan breaks down the sticker shock of large AI spend and how they manage it. He gives an example where thousands of dollars of compute replaced months of work, while also urging model selection discipline for everyday tasks.
- •Annual Anthropic run-rate ~ $1.2M and growing
- •Example: one engineer spent ~$3,000 in a day
- •That spend enabled work that would’ve taken 3–4 months otherwise
- •Company guidance: don’t use top-tier models for trivial tasks
- •ROI framing: tokens are worth it if they replace far larger labor costs
- 17:44 – 20:28
Operational AI: Slack/Gmail/Docs connectors and an AI “chief of staff” for executives
Beyond engineering, Khan explains how connected AI systems help with coordination, drafting, and follow-through. He emphasizes keeping humans in the loop—drafting is fine, but autonomous sending/posting is risky.
- •Approved connectors allow AI access to internal tools (Slack, email, docs)
- •Khan uses AI to identify what’s “falling through the cracks”
- •AI drafts emails and actions while he reviews/approves
- •Proposal writing and paperwork accelerated dramatically
- •Human-in-the-loop rules: draft, don’t send; don’t post directly
- 20:28 – 23:27
Could AI build another Khan Academy? Moats: efficacy data, privacy, and psychometrics
Khan answers the fear of being replaced by AI-built competitors, distinguishing ego concerns from nonprofit mission. He argues that building a credible education platform requires more than vibe coding: evidence of learning outcomes, school-system integrations, and validated assessments are hard barriers.
- •Ego response: yes, plausible competitors in 3–5 years
- •Practical friction: efficacy proof, school partnerships, data privacy
- •Nonprofit stance: if someone builds better, society wins
- •Khan Academy focus areas less “garage-buildable”: standardized assessment + psychometrics
- •Credentials, diplomas, credit recovery, and job pathways as long-term priorities
- 23:27 – 25:16
Khan TED Institute: a $10K (or less) accredited degree built for reskilling
Khan introduces the Khan TED Institute (with TED and ETS), designed to create new pathways into work as AI reshapes hiring. The model aims to be accredited, low-cost, competency-based, and faster than traditional four-year degrees.
- •Mission: new systems for skilling/reskilling in an AI economy
- •Announced at TED; pursuing accreditation
- •Targets both bachelor’s and master’s programs
- •Price cap: $10,000 for the full degree (intended to be lower)
- •Competency-based structure may reduce time-to-degree and opportunity cost
- 25:16 – 30:11
Durable skills and simulation-based learning: measuring what GPAs can’t
The program emphasizes group simulations and projects to surface durable skills that matter in AI-heavy workplaces. ETS-aligned frameworks help quantify communication, collaboration, creativity, and critical thinking through repeated, peer-reviewed performance.
- •Core durable skills: communication, collaboration, creativity, critical thinking (plus leadership subskills)
- •Students complete time-boxed team simulations (prototype, research, plan, present)
- •Peer review + structured rubrics generate richer signals than grades
- •Portfolio artifacts include evidence (often video) of real performance
- •Designed for all ages, including those already holding degrees
- 30:11 – 32:20
What happens to non-elite universities: reducing online stigma and building hiring signals
Khan argues higher education is polarized between elite campuses and lower-signal online options. The Institute’s strategy is to prove online can be high-interaction and high-signal, helping capable students from non-elite schools access better opportunities.
- •Elite universities: high signal but low capacity and accessibility
- •Online programs: often good education but weaker employer signal
- •Goal: make online learning highly interactive and credible to employers
- •Aspirational employer partnerships help validate the pathway
- •Long-term aim: serve both top-tier and regional-job outcomes with scalable programs
- 32:20 – 38:29
Should parents still save for college? Elite college as a luxury vs. cheaper job-ready paths
Khan frames elite college as a valuable but expensive “luxury experience,” while many students attend for transactional job outcomes. He suggests hybrid strategies: local colleges plus the Institute for signaling and job readiness, potentially at a fraction of the cost.
- •Elite college offers friendships and coming-of-age—but at very high cost
- •Most students don’t experience the ‘ivy quad’ version of college
- •Institute can complement a traditional degree with job-ready skills
- •Cost comparison: $400k+ vs. ~$5k–$10k, plus faster completion
- •Alternative uses of funds: travel, startups, buying/automating small businesses
- 38:29 – 44:09
Staying relevant: build with AI, lower activation energy, and use prompts that push back
Khan’s advice for individuals is to adopt an experimental, entrepreneurial mindset—try tools, automate workflows, and build a visible portfolio of what you can do with AI. He also warns about AI flattery and recommends prompts that force critique and skepticism.
- •Key meta-skill: entrepreneurship/intrapreneurship and willingness to tinker
- •Start incremental: copy/paste workflows, then graduate to agents and automation
- •Differentiate with proof: workflows, portfolios, Loom demos, “I come with 100 agents”
- •Beware ‘seductive’ AI validation; ask it to be critical and challenge assumptions
- •Near-term risk: sector-specific unemployment pockets; prepare before disruption scales
