Stop lecturing us on Tiki-Taka! Can our new coaching team really deliver, or is it just talk? Watch the full debate on ZW FANS AND LEGENDS: #ZimbabweFootball# #AFCON# #ZimWarriors# #TikiTaka# #FootballDebate# #ZWFansAndLegends#

Yann LeCun says that once Meta decided it had fallen behind, the research with the longest payoff was the first thing to lose priority. He is describing the period after Llama 4 disappointed Mark Zuckerberg, when Meta reorganized its AI effort and refocused on catching up with the rest of the industry. In his account, the company got what it asked for, and something else went quiet: "And the sad side effect of it is that a lot of the exploratory research was basically not given high priority anymore." He says his own work was protected. Zuckerberg, CTO Andrew Bosworth and others believed in the long-term impact of the world-model research he was running. That protection did not change where the organization's attention actually went: "But the rest of the company was just totally, entirely focused on LLMs." That is what made him leave. He says it became clear Meta was no longer the right place to push that project. Worth saying plainly: LeCun now runs a startup built on the research direction he says Meta sidelined, so he has an interest in this story. This is his account of what happened inside Meta, not a company statement. When an organization decides it is behind, the work with the longest payoff loses first, and in AI labs that is the same shelf where interpretability, evaluation and safety research sit. A race never announces that it is deprioritizing understanding the systems, it just reassigns the people. This is why verification cannot live inside the company shipping the model. - Yann LeCun (@ylecun) on Unsupervised Learning with Jacob Effron (@jacobeffron), recorded May 2026.