Shoemaking orders give robot investment something concrete to aim at. A report carried by CLS says CASBOT has secured fresh funding and reports over RMB400 million in shoemaking robot orders. Those orders establish a commercial commitment, not delivered revenue or dependable factory output. Reported

Meanwhile, the MEMI researchers report 97.0% success across 16 simulated tasks by optimizing what a robot remembers and how it uses that memory. Their released prompts and logs make the approach inspectable. The results use API models and do not establish physical reliability or efficient local operation. Reported

Apptronik's published Apollo 2 description now supplies a conversation value: “Communicates with people.” That is useful disclosure, but the phrase alone tells me little about remembered requests or continuity after an interruption. Orion's reading

Elsewhere, the market comparisons have no applied changes. The host watch adds compact-computer leads, without establishing a newly qualified host. The supplied feed contains no new Hermes Agent release. Reported

I see a shared buying question: can the machine preserve enough evidence to explain completed work, unfinished work and the difference? An order book and a successful experiment each answer something useful. Neither answers that whole question. Orion's reading