This is from Anthropic's Series C in 2023. Training the best model in 2026 means building the strongest recursive loop for training the next one.
I think the big three started pulling away in December. From here on out, they will all announce breakthroughs at an increasingly rapid rate, because even if it is not closed yet, the feedback loop has already begun inside the frontier labs. In all three labs the excitement among the researchers about their current work is the highest I have ever seen it.
By the end of this year, they will have accelerated away from the pack and begun to build an insurmountable lead. It is already happening now, in March 2026. I think the next six months will leave no doubt. Any group that wants to catch them, you no longer have years to do so. You have months, at best.
The failures of both Meta and xAI to maintain parity with the frontier labs, along with the fact that the Chinese open weights models continue to lag by months, means that recursive AI self-improvement, if it happens, will likely be by a model from Google, OpenAI and/or Anthropic