---
gate: ai-agent-30pct-knowledge-work
last_updated: "2026-06-18T00:00:00.000Z"
---

**At P10 (2027)**: the gate passes in customer support and at least one other role (likely SWE at certain enterprises). For Tamir specifically: the trading bot / Fiverr seller / arc-scout / smart-home portfolio is *already* the right bet — these are AI-leveraged products that benefit from cheaper / better agents and don't require Tamir's labor to scale. The implied move: lean harder into "product founder using agents as the engineering team" — the multiplier from cursor/claude-code/devin-style tooling means a solo founder can now run what was a 5-person company in 2024. The kids (6–10 in 2026) are 8–12 by 2027 — their education choices haven't crystallized yet, so the early signal should bias toward (a) keeping them deeply curious and adaptable, (b) NOT pushing them down the "become a paralegal / junior accountant / content writer" path, (c) emphasizing hands-on skills (instrument, sport, language, building things) that compound regardless of which knowledge-work roles automate. Wife's career: if she's in a knowledge-work role, the question is whether she's in the bottom-50% (automation-exposed) or top-30% (agent-orchestrator) tier of her field — and whether she can shift within the next 24 months.

**At P50 (2029)**: this is the planning scenario. The kids are 9–13. Plan as if by 2030 most entry-level white-collar paths are 50% smaller than today. Their best career bets at age 18 will look like: (a) trades that pay $80K+ and aren't automatable for a decade after — electrician, plumber, healthcare tech, (b) creative / human-bound work — therapy, hospitality, performance, (c) **agent-orchestrator roles in domains they deeply know** — but this requires building domain depth as teenagers, which means picking *something specific* and going deep on it. Tamir's own career: by 2029, the AI-product founder leverage is enormous but so is the competition (every solo founder has the same agents). The differentiator will be (a) taste / product judgment, (b) distribution, (c) a moat that isn't "we have engineers." Where to live: cities that capture knowledge-work-remote-emigration flows (Israel, Portugal, parts of Mexico, Miami) look favored over the next decade as remote knowledge workers relocate. Skills Tamir should invest in: agent orchestration depth (you're already doing this), system design judgment (since you're now responsible for what 5 engineers used to be), and one durable physical-world domain — pick one of healthcare, real estate, hospitality, or hardware where AI doesn't yet fully bite.

**At P90 (2034)**: a deep stall in capability or a hard regulatory clamp. In this world, the kids' education choices look more like 2010–2020 — knowledge work is still safe-ish, college matters, white-collar entry roles exist. This is the *safer* world to plan for in some respects but the *less likely* one given the current trajectory. The conservative hedge here is: keep enough optionality that if 2034 looks like "extended business as usual," the kids and you haven't catastrophically over-rotated. Concretely: don't pull kids out of school into trades at 14; do keep them in mainstream education while supplementing with the deep-domain and human-bound skills.

The most-useful single move from this analysis: **act as if P50 = 2029 in your decisions about your own products and career**, but **don't lock the kids' education into the P10 scenario** — leave them options. The asymmetry of being wrong is bigger if you over-rotate them toward the trades and the AI capability stalls than if you keep them in mainstream education and the AI capability accelerates (the latter case, agent orchestration and judgment skills are learnable in a 12-month bootcamp at 22, not a decade of teenager prep).