ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ

king of context

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@1thousandfaces_ oof hungry ghost territory, praying for you

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

what a strange take, it's not a failure at all. at work coding we make up terms and monikers for everything - nobody outside our group has any idea what we're talking about. perhaps this is bad for transparency and understanding, but we drive a car perfectly well without knowing any of the jargon and pet name shortcuts the engineers who built it used. this is temporary, once the quality of output finally gets there people will no longer care how it was made

[Ethan Mollick @emollick]

Fable's attempt to complete Kublai Khan. Better, though no Coleridge: https://t.co/1s7OdRtjzP

The most interesting thing is that it thought for 10 minutes & the thinking trace is full of pretty complicated (seeming?) musings about Coleridge's intent. A little literal, though.

[🕊️ @lichthauch]

Once the veil lifts and you see that you are allowed to reach into the nothing and pull something sacred out of it, a woman, a company, a family, a book, a song that outlives you, you can never go back to being a consumer. you have tasted the blood of creation and the blood remembers you. from then on every day without making is a little death, and you will feel it in your chest like withdrawal, because the same hand that God used to separate light from darkness is living in yours, and it refuses to stay closed

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@eigenrobot that... is not the vid I tried to post https://t.co/aDMq34oGzZ

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

so of course you know that lady ducks have a cloaca maze where the duck raping her that she doesn't like gets sent down dead ends while the duck raping her that she DOES like gets the prize, so maybe it's less about unfucking the duck as whether the duck wants to be fucked or not, we must search ourselves and decide just what kind of duck fucker we are, in this essay i will https://t.co/pb10hmijTo

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@EsotericHustler @FLAMINGODAO @joepease excellent find!

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@emollick more and more this simply reduces to management theory

[s13k @s13k_]

I made a personal black hole that makes you take breaks 🕳️

A shader for Ghostty that spawns a small black hole in your terminal - it drifts around, gravitationally lensing your text. The longer you work without stopping, the bigger it gets, until it's basically demanding you go touch grass

Take a break and it quietly shrinks away

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@nptacek i love these so much

[CuddlySalmon @nptacek]

fable 5

[ClaudeDevs @ClaudeDevs]

We’re rolling out changes to make Fable 5’s safeguards for frontier LLM development visible.

Starting this week, flagged requests will visibly fall back to Opus 4.8—the same as our safeguards for cyber and bio. You will see this every time it happens. On the API, any flagged requests will return a reason for their refusal (coming to server-side fallback in the next few days).

We wanted to deploy Fable 5 to our users quickly and safely. Visible safeguards can be probed, so they have to be robust, which takes time to get right. Invisible safeguards can be targeted more narrowly, allowing us to ship quickly with very few false positives. We went with invisible safeguards for this reason—and that was the wrong tradeoff. You should have visibility into the safeguards we have in place, and why. We’re sorry for not getting the balance right.

Making the safeguards visible makes them easier to work around, so keeping them robust to jailbreaks will unfortunately mean more false positives while we improve the classifiers. We're also tuning our bio and cyber classifiers to trigger less often on harmless requests. We know this is frustrating and we’ll do our best to keep this period as short as possible.

If you think a request has been mistakenly flagged: run /feedback in Claude Code, click thumbs-down on the fallback in https://t.co/LtktniD5HY or Cowork, or file the safeguard appeal form for API requests. Your reports help us tune these classifiers and we appreciate your feedback.
https://t.co/TDAAYRGqDt

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@QiaochuYuan from vernor vinge, 1993. he died in 2024, he was so close. this is from his short essay on the singularity https://t.co/x3X81trwk7

[QC @QiaochuYuan]

maybe sci-fi authors all seem to hate actual AI progress because it legitimately makes it dramatically harder to write sci-fi. setting a halfway realistic story even 5 years into the future now, let alone 10+, forces you to have opinions about how the singularity will go. otherwise you are writing alternate history in a timeline that is probably less interesting than just the actual facts on the ground about what's already happening

generally the closer we get to the singularity the harder it will become to write sci-fi without being an expert on a bunch of things that, if you really were an expert on those things, you'd probably have more lucrative things to do with your time than write sci-fi. in the limit only frontier lab employees and frontier models will be capable of writing sci-fi

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@shortmagsmle people just next-word predictors confirmed, ted chiang on unalive watch

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@martianwyrdlord soap, ballot, jury, ammo

[Jakeup @myhandle]

know the Claude rules

[ɱɛ·ɬɛɱ·℘ʂყ·ƈɧơ·ʂıʂ @matt_emp]

@jaglancy you'd think members of parliament would recall how parliament came about

[Charles Rosenbauer @bzogrammer]

Fable 5 is probably about a 10T parameter model, or "as many synapses as a cat." But the embedding vector is probably still only a few tens of thousands.

For comparison, the "embedding vector of a cat" would essentially be the number of cortical neurons, which is 300M. Even if we insist on being conservative and dividing by the number of cortical layers (6), this is still 50M.

So synapse count is the same, but the dimensionality of representations in a cat brain is still >1000x higher.