Foundation Capital partner Jaya Gupta (자야 굽타) says large language models have bundled many functions into a single model, but that bundling is beginning to unwind. She describes a “great unbundling of intelligence” driven by differences in compute needs, performance, latency and cost across capabilities. Gupta says agents amplify the impact of small cost and latency gaps across many steps. She argues judgement models, rerankers and separate search services can replace expensive general models for routine work.

Bitcoin could be exposed to threats from quantum computing, but its community is already preparing responses, an assessment said. VanEck digital assets research head Matthew Sigel said bitcoin\'s decentralised structure could slow action, but technical solutions are being discussed. Industry concerns focus on whether future quantum computers could break bitcoin\'s cryptography, though none can do so now. Developers are testing quantum-resistant signatures, while Coinbase, Blockstream and a new security consortium are developing and backing proposals.