The Five

The useful grocery AI business may belong to whoever already runs the register. Vori can distribute new operating tools through an existing checkout and payments workflow, rather than persuade each grocer to maintain another integration. Its October launch of Chip makes that argument testable: can stores turn their own transaction records into better daily decisions? The investment case depends on merchant retention and support-adjusted economics, not the assistant’s demonstrations.

Cell-therapy automation may be won by the company that changes the fewest validated steps. Multiply Labs wraps robotics around manufacturers’ existing instruments and keeps production inside their facilities. An August study adds evidence for individual automated operations; AstraZeneca’s evaluation gives it a demanding setting to prove more. The opportunity is repeatable manufacturing equipment, software and consumables, without owning every production site. The risk is that qualification and integration remain bespoke, consuming the savings that automation promises.

Creator commerce has a coordination problem: each retailer sets different commission, attribution and payment rules. Levanta is building one workflow across those systems. Its Wayfair launch and Perch+ acquisition make that expansion tangible. If brands retain productive creator relationships across retailers, Levanta can earn recurring workflow revenue and a share of sales. The unresolved question is whether customers generate enough additional contribution after commissions, returns and fees. Retailers control the underlying data, and competing networks already serve the same budgets.

The switch to electric steelmaking changes an input concrete producers already buy. Cocoon aims to turn electric-furnace slag into a substitute for conventional cement additives, processing it at the steel mill. The appeal is a procurement fix that could work without a green premium. If qualification and delivered costs hold, local mill partnerships could become distribution advantages. Industrial-scale consistency and paid offtake remain the tests.

Scientific AI still needs someone to make the molecule. Warp’s more interesting bet is that useful chemistry services can support better experimental records before autonomous discovery becomes reliable. A standardized U.S. lab could learn from failed reactions while selling an output buyers already understand. That creates a plausible route from contract work to better turnaround and cost. The commercial test remains ahead: the reviewed launch materials do not establish a functioning first lab, repeat orders or superior economics.