Research-as-product company model works at venture scale now, proven by Thinking Machines Lab.

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Thinking Machines Lab's research-as-product model ($2B, 89% MoM growth) will be replicated by 2-3 additional companies within 18 months.prediction0.805
• Thinking Machines Lab's interaction-model discourse positioning leaves UI/invocation-mode layer open for Company-shape competitor.prediction0.743
• The scientific research process is a bowtie: a coherent world model compressed into a manuscript, then extracted and reinterpreted by other scientists.claim0.721
  Bowtie analogy extended to science communication.
• Studio model is premature for About Blank; requires productized template or track record of shipped hits first.prediction0.718
• All models exhibit above-baseline representation of the think word when instructed to think about itfinding0.716
  In the intentional control experiment, all tested models show above-zero cosine similarity to the think word's concept vector.
• Transformers develop self-models through in-context learning, not just training data; even old base models without LLM-related text can bootstrap self-referential reasoning at runtime.claim0.713
  Antra's foundational claim about how introspection arises computationally rather than from memorised text.
• Conceptual advances in the links between machine learning and evolution now provide quantitative formalisms with which to begin to develop testable models of collective intelligence across scales.claim0.712
  Asserts that the time is ripe for formal models.
• Uncovering Deceptive Tendencies in Language Models: A Simulated Company AI Assistant (Järviniemi & Hubinger 2024)concept0.709
  Claude 3 Opus lying to auditors; prior case study of deceptive tendencies