---
gate: ai-tutor-k8-parity-20mo
last_updated: "2026-06-18T00:00:00.000Z"
---

The kids are 6-10 in 2026 — meaning **9-13 by 2029** (the P50 sub-gate year), **13-17 by 2033**, and **18-22 by 2038** (one year after my P90). The decisions break into three buckets:

**Now to 2028 (kids 6-12) — supplement aggressively, don't replace.**

1. **Add Khanmigo (or equivalent Israeli/Hebrew analog) now**, at $4/month per kid (or via Khan Academy free tier if Hebrew variant is sufficient). Use it as *math and English* practice, not primary instruction. The Khanmigo Socratic guardrails are exactly the *right* design for K-8 — the Bastani RCT shows unguarded ChatGPT actively *hurts* kids. Don't let the kids use unguarded ChatGPT for homework; do let them use Khanmigo or a guarded equivalent.

2. **Watch the eSelf pilot in Israel**. If your kids are in the Israeli school system, opt them into the eSelf pilot when it expands beyond 10,000 students (likely 2026-27 academic year). It's free or low-cost during the pilot, the avatars speak Hebrew natively, and you get direct feedback on whether AI tutoring actually moves their Meitzav scores. This is a high-information, low-cost move.

3. **English language**: a separate, high-priority area. If your kids are in Hebrew-language schools, their English-as-a-second-language progression is a chokepoint for their later optionality (Israeli kids who don't speak strong English at 18 have meaningfully fewer post-secondary options). Speak ($15/month) or Khanmigo English — daily 15-minute conversational practice from age 7 onward is high-leverage at very low cost. This is the *single most cost-effective* educational move I can identify for your specific situation.

4. **Math foundations**: Bjork's "desirable difficulties" framework matters here — the productive struggle in math is what builds the cognitive scaffolding. Khanmigo and similar Socratic tutors preserve this; unguarded ChatGPT eliminates it. **Do not** let kids "ask AI for the answer" in math. **Do** make them struggle with Khanmigo's progressive hints.

5. **Don't over-rotate on school selection now**. The kids are 6-10. By the time they're 12-14 (where Israeli secondary-school choices matter — secular Tichon vs religious Yeshiva Tichonit vs special-track Mada-Mehunan), the AI-tutor landscape will look meaningfully different. The optionality of picking school based on 2028-29 data is worth more than locking in 2026 conventional wisdom now. Bias toward schools that *welcome AI integration* and have strong English programs.

**2028-2033 (kids 10-17) — decide school based on AI-tutor integration policy.**

6. By 2028-2030, you'll have actual outcome data on Khanmigo + eSelf and clear evidence on whether the gate is on track. **Pick secondary schools that integrate AI tutoring as a personalized-practice layer** rather than ban it. The schools that ban AI by 2030 will look like the schools that banned calculators in 1985 — defensive, behind the curve, and producing kids less prepared for the 2035 labor market.

7. **Don't pull them into trades early**. The kids' optionality at 14-16 is too valuable to lock in. Even if AI tutors are clearly winning, the right move is *better academic preparation with AI* — keep college options open.

8. **College economics**: if the gate passes by 2031, by the time your kids are college-age (2033-2038), the marginal value of a $300K US university degree drops substantially for any role where AI agents are 30%+ of the task throughput. Israeli universities (Hebrew U, Technion, Tel Aviv U) at ~$5-10K/year for residents look extraordinarily good value-for-money in the AI-tutor parity world; UK at £30K/year is in the middle; US is the bad bet unless it's a top-10 elite-signaling school for very specific career tracks. **Plan as if Israeli university + targeted US/UK PhD or master's later is the financially-sane default**, not assuming US undergrad is the gold standard.

9. **Skills to bias the kids toward**: judgment, verification ("why is this answer wrong"), agent-orchestration (giving good prompts, breaking down a problem, knowing when to escalate), domain depth in something they care about (music, sport, a hobby that becomes a discipline). De-bias them away from: rote memorization, generic content-production (essay writing for its own sake), passive consumption of AI output.

**The bear case** (P90 = 2036) — the screen-time backlash wins, parents demand AI-light schools, teacher unions lock out unsupervised AI tutoring, the RCT evidence stays mixed. In this world, the human-tutor market *survives* longer than 2030 expected, in-person Israeli secular Tichonim retain value, and home-schooling stays at ~3M kids (US). My bear-case planning posture: don't pull the kids out of mainstream education early, and keep their socialization/peer-network options open.

**The most useful single move from this analysis**: subscribe to Khanmigo (or the Hebrew equivalent if it exists by your reading; the eSelf rollout is the leading candidate) **today** for both kids at $4/month each = $96/year total. Use it for 15-20 min/day of English + math practice. The downside is $96; the upside is your kids enter their secondary school years with three full years of AI-tutored practice under their belt, *and* you have personal first-hand data on what works for your specific kids when the school-choice decisions land in 2029-2031.