Watchlist
24 accounts we always pull takes from, whatever their reach that day. Each one is filed in a camp by hand and re-checked when their stance drifts. 21 handles confirmed so far.
Draft data. The stances shown below are placeholder copy written for this mockup. They are not real posts and will be replaced by classified reactions from the pipeline.
- Andrej KarpathyVerified@karpathyIndependent researcherMiddle“Placeholder stance: the distillation recipe is the interesting part, not the benchmark table.”on Open-weight model matches proprietary leader on coding benchmarks at one-fifth the size
- Ethan MollickVerified@emollickProfessor, WhartonMiddle“Placeholder stance: hiring agents will be judged on their audit trail, and most do not have one yet.”on EU AI Office opens formal inquiry into agent-based hiring tools
- Andrew NgVerified@AndrewYNgFounder, DeepLearning.AIMiddle“Placeholder stance: compute disclosure is a reasonable, low-cost step. Keep it voluntary and see what it teaches us.”on Frontier labs agree to publish compute totals for every model above 10²⁶ FLOPs
- Gary MarcusVerified@GaryMarcusProfessor Emeritus, NYUMiddleSkeptic“Placeholder stance: parity on three benchmarks is not parity. Wait for the replications.”on Open-weight model matches proprietary leader on coding benchmarks at one-fifth the size
- Zeynep TufekciVerified@zeynepColumnist and sociologistMiddle“Placeholder stance: the synthetic extras deal will be cited in every creative negotiation for a decade.”on Major studio signs first contract allowing synthetic background performers
- Daniel OkaforVerified@dokafor_policySr. Policy Fellow, CSETMiddle“Disclosure is the prerequisite for everything else.”on Frontier labs agree to publish compute totals for every model above 10²⁶ FLOPs