As of November 30, 2025, only about 2.5 years have passed since mid‑2023, and Jason’s prediction explicitly allowed up to three years (i.e., roughly until mid‑2026), so the full time window has not elapsed.

There is clear evidence that some ~20‑person AI‑native startups have achieved extremely high output and revenue without large headcount—e.g., Cursor/Anysphere reportedly reached around $100M ARR with roughly 20 employees, and similar small‑team, high‑revenue patterns are noted for other AI startups. (forbes.com) Other examples like Vocal Image (about $12M ARR with a 20‑person team) and ChipStack (a 20‑person AI verification startup acquired by Cadence after demonstrating large efficiency gains) also illustrate very high revenue or productivity per employee. (blog.connfig.com) These cases show that it is possible for 20‑person companies to massively scale output without scaling headcount.

However, system‑level data that would tell us whether “many” 20‑person tech companies have roughly doubled their business output or revenue because of AI does not exist in a clean, aggregated form. Academic and industry studies on AI coding assistants like GitHub Copilot generally find substantial but incremental productivity gains (e.g., on the order of single‑ to low‑double‑digit percentage improvements at the project or developer level, or 30–40% time savings on certain coding tasks), not a universal 2× output figure. (arxiv.org)

Adoption data shows that AI is spreading quickly—e.g., a U.S. Chamber of Commerce report cited in business press finds that a majority of small businesses now use generative AI, and broader surveys suggest a large majority of companies worldwide are using or exploring AI. (kiplinger.com) But these statistics don’t specify 20‑person tech companies, nor do they quantify how many have doubled output at fixed headcount.

Because (1) the prediction’s 3‑year horizon has not yet fully expired, and (2) available evidence shows notable supporting anecdotes but no comprehensive data to confirm or refute that many such companies have doubled output, the prediction cannot be definitively graded yet. It remains plausible but unproven, so the status is best described as inconclusive (too early).