According to Daniel Kokotajlo on February 1st 2028, if Anthropic was going full-speed during 2027 (i.e. trying to make ASI happen as soon as possible ignoring political and hiring constraints), how many months into 2027 would it have reached the AI R&D progress that it had on January 1st 2028.

For example the market resolves to 8 months if Daniel expects that if Anthropic went full-speed during 2027, it would have done roughly as much AI R&D progress on September 1st 2027 as it did in reality on Jan 1st 2028.

AI R&D progress is pointing at something like software efficiency (https://www.aifuturesmodel.com/#section-softwareefficiency), and is not necessarily tied to internally or externally deployed models. Software efficiency might not be a great concept at the end of 2027, and so I will let Daniel choose what he thinks makes most sense to call AI R&D progress.

Falls back to Thomas Larsen if Daniel Kokotajlo can't answer. I will ask them their best guess based on public information.

Resolves to N/A if Anthropic stops existing at some point in 2027 for reasons mostly unrelated to pacing (nationalization, merger with another frontier AI company, etc.).