“both us and many others in the field have found that these models that are trained on descriptive data do not yet outperform linear models on causal tasks, perturbational tasks, what we call counterfactual tasks.”

Eiso Kant on the training run being anticlimactic compared to surrounding R&D work.

Huberman on BDNF and neuroplasticity: raising the signal molecule alone changes nothing without coordinated firing.

Eiso Kant on the absence of a true science of deep learning for LLMs, with all labs still operating empirically.

James Hawkins on the VC return math: investing in a $50M-$100M outcome is a total waste of time.

Michael Batnick on retail investor buying persistence throughout July 2026.

Eiso Kant on TSMC holding more effective power over AI compute than the US government.

Nathaniel Whittemore on Kimi K3 breaking the open-model barrier against proprietary frontier models.

Ci Chu on why observational biology data is insufficient for causal modeling.

McCoy on the security implications of Kimi K3 being a true open-weights frontier model.

Guillermo Rauch on Kimi K3 outperforming Fable in real-world web engineering evaluation.

Luis Laboy on why every market is now an emerging market.