A model says 13 percent automation is enough to tip growth into the explosive zone

Jack Clark's latest Import AI leads with a growth model that puts a number on a fuzzy fear. Its central claim is that automating about 13 percent of work across all sectors is enough to push the economy into an explosive regime. The lever that matters is not software but hardware: automating chip design carries roughly five times the economic impact of automating software, and under the baseline assumptions, full software R&D automation plus 5 percent automation elsewhere could trigger exponential growth in about six years.

Clark pairs this with a policy argument he calls radical optionality. Instead of heavy regulation now or nothing at all, governments should spend modestly to keep future choices open: transparency and third-party auditing, whistleblower protections for frontier-lab staff, cross-government information sharing, and technical hiring at agencies like AISI and CAISI. The newsletter also notes a Meta and KAIST prototype of a "neural computer" that replaces the operating system with a learned runtime, though its symbolic stability is still weak.

The growth section is the part to read closely, because it makes a testable claim rather than a vibe.

Why it matters

If you advise on policy or long-range strategy, the specific threshold and the hardware-first finding give you something concrete to argue from: the cheap, reversible moves Clark lists are easier to defend before a fast takeoff than after one.