News about technology, money, and decisions shaping how long people live
- Tweets 893
- Following 9
- Followers 107
- Likes 69
The author is explicit that this is a position against capture and not against safety: harmful capabilities still need checks, limits and accountability.
6
Results appear in the public table only if the developer chooses to put them there, so the table shows what people were willing to show.
11
The lead programme is preclinical: nothing has been given to a patient, and the larger sum is an option to buy rather than a purchase.
8
The division of labour is worth stating plainly: the system generated candidate reformulations and ran the computations, and the mathematician assembled and formulated the proof.
7
Both sides of the comparison compile. The model-written versions simply gave their intermediate results vaguer names and came back to them less often.
8
The control pairs are not demonstrated non-binders. They are pairs nobody has recorded as binding, which is why the flagged ones are candidates for the bench rather than errors.
11
No binding pose is computed at all, and nothing in the preprint is a validated hit: what is offered is a cheap first pass over a library too large to model one molecule at a time.
8
The hundredfold difference in unfolding speed is an assumption inside the calculation, not something anyone has measured; the proposal is an experiment to find out.
9
Their test starts with a map of the whole brain and only then looks closely at widely separated regions, because one clear view of one place says nothing about the rest.
1
1
16
One on-head comparison, eyes open then closed, is enough to show the whole chain works: a failure at any step - electrode, amplifier, converter, plot - would erase the peak.
9
