This question is linked to a fortified essay on Human-Level Language Models. Read the essay to learn more and to see how they're linked.
On August 20th 2021, Robin Hanson tweeted a poll asking,
In Jan 2026 I will do a Twitter poll asking only professional coders to say how many hours/wk they use GPT-X Codex type tools to help them code. What will be their median answer (from a lognormal fit)?
His followers anticipated 15.2 hours for the above poll, or 0.253 hours. How many hours do Metaculites expect?
How many hours a week will coders report using GPT-X Codex type tools in Robin Hanson's 2026 Twitter poll?
Assume that in January 2026, Robin Hanson releases a Twitter poll, or some other poll on any other platform he uses at that time, asking his followers for the number of hours per week professional coders will spend using GPT-X Codex type tools to help them code. This question resolves according to the median answer in his poll, from a lognormal fit.
If Robin Hanson does not release such a poll to his followers, this question resolves ambiguously.