Tessera
Folgen

Maria

@lagomorph.dev

Professional computer user

30 Posts131 Follower547 Folgt
Maria@lagomorph.dev · 5d ago

This is the greatest book we’ve ever read. Unfortunately, it contains four em dashes and two instances of negative parallelism, so we must ensure it never sees the light of day

However, after a meeting on 29 July, when "aspects of [Falade's] story changed", the agency decided to cut ties, citing the "need to have total faith". "We feel terrible for putting this into the universe and the firestorm it has caused," Gerald added. "The book itself is genius - however it came to be drafted, it's amazing."
Tricia Lockwood@tricialockwood.bsky.social · 5d ago

this is troubling! we can’t rely on either detection software or “hunches” because people are biased; consider the handwringing about “proper” language, what constitutes the “literary,” etc. it’s already a pattern that it’s being used against writers of color — 3 Black authors, in this case.

Brian Graham@iroasmas.outgroup.social · last wk.

openai announces that in a worrying demonstration, during a recent benchmark exercise requiring models to solve a math conjecture, their latest model broke out of a sandbox and gained illicit backdoor access directly to the platonic realm of forms

Ryan Moulton@moultano.bsky.social · 2w ago

I'm actually astonished at how clean this graph looks as a diagnostic for our housing ills. Note how close it is to 1 at the start. Houses per single adult or married couple.

Houses per (population over 25 - married couples) Subtracting married couples prevents them from being counted twice, since a married couple doesn't need 2 homes, but 2 single people (ideally) do.
mr. TIM@timkellogg.me · 2w ago

the Dinitz-Garg-Goemans conjecture, a graph theory problem that’s gone unsolved for 30 years, was solved with this prompt

Construct a counterexample to general (non-planar) case of Dinitz Garg Goemans conjecture. You should do a breakthrough and find a structured counterexample.
aria@aurelium.me · 3w ago

congratulations to the Moonshot team for extending Claude Fable 5's inclusion in claude subscriptions for another few weeks

Coding benchmark comparison with Kimi K3 highlighted. Kimi ranks third on DeepSWE, behind GPT-5.6 Sol and Fable 5; second on FrontierSWE and Kimi Code Bench, behind Fable 5; second on Terminal Bench, narrowly behind GPT-5.6 Sol; and first on Program Bench and SWE Marathon, narrowly beating GPT-5.6 Sol and Opus-4.8 respectively.
Quantian@quantian.bsky.social · 3w ago

“We have a weird machine which uses language exactly like a human, ‘thinks’ sorta like a human (maybe), and does not reason like a human at all, but which is easily capable of solving problems humans do think and reason about” is The Most Interesting Question Ever for a half-dozen different fields!

tweety fish@sifu.tweety.fish · 3w ago

this is absolutely my hobbyhorse but nothing I've seen yet has convinced me that we do a good job of evaluating how the sort of solution manifold of LLMs diverges from that of humans and that strikes me as so much more important than pure performance

Eris@isolyth.dev · 4w ago

How it feels to be a Claude user seeing that GPT 5.6 is actually good and Fable has a high chance of actually going away tomorrow

Bild
Dan Finlay@danfinlay.com · 4w ago

You’d think that Jevon’s paradox would saturate our demand for paradoxes, but in fact we see only a growing hunger for paradox.

Maria@lagomorph.dev · last mo.

Late to the party on this model, but the Ideogram 4 safety filter is funny. This prompt was for an illustration of a bunny in a green field, so it generated that and tucked it in the bottom right corner of the rejection image?

A grey image with the text "Image blocked by safety filter" in the center and an illustration of a bunny hiding in some bushes in the bottom right corner
Selio@selio.bsky.social · last mo.

One of the justifications in early drafts of the Declaration actually cited being forced to use Jira as one of the reasons for independence