Pip Foweraker 🔶️
banner
pipfoweraker.bsky.social
Pip Foweraker 🔶️
@pipfoweraker.bsky.social
Every hour an AI safety researcher spends on bureaucracy or operations is a tragedy. Every hour a capabilities researcher does the same is a triumph.

Building pdoom1.com
Founding Certes certesinstitute.org
Available to consult. Hobart, TAS, Australia
Pinned
That which is strengthened by the truth should be
Reposted by Pip Foweraker 🔶️
alignment has many doors, Ed-boy
January 30, 2026 at 1:33 PM
7am lawyer
7am lawyer
$8.60 coffee
Hotel foyer
February 2, 2026 at 9:10 PM
"When this mind is retrieved, it will be the true mind, and once the true mind appears, it can extinguish demons."
- Xiyou zhengdao shu 西遊證道書, 1662, facs. reprint Guben xiaoshuo jicheng 古本小説集成, 437-440, Shanghai, 1990.
February 1, 2026 at 9:21 PM
At some point one must admit it's not a series of pivots so much as a poorly executed pirouette
January 22, 2026 at 9:31 PM
One must imagine Stu happy.
January 17, 2026 at 4:33 PM
Didn't even know this was an option
December 31, 2025 at 11:02 AM
Reposted by Pip Foweraker 🔶️
And in Tasmanian news...
December 31, 2025 at 10:23 AM
Reposted by Pip Foweraker 🔶️
December 29, 2025 at 5:36 AM
Reposted by Pip Foweraker 🔶️
look at this hodograph
every time i do it makes me laugh
December 24, 2025 at 7:33 PM
Merry Christmas, ya weirdos
December 24, 2025 at 9:40 PM
It's fantastic coming across an articulation of something that I've thought fuzzily around the edges of for years:
They have made enormous headway. The central issue is that driving isn't _really_ a rule-based process; the rules are codifications of social mores but fundamentally driving around other people is about theory of mind and social negotiation, which ML is largely hopeless at; here's a thing I wrote.
Driving is a social process
Photo by Sangga Rima Roman Selia on Unsplash There is something very strange about automobiles. They are much faster and more dangerous than our brains are...
buttondown.com
December 22, 2025 at 1:23 AM
I get very mad when people mis-treat my screwdriver, so I understand this failure mode even as I try to guard against it in my day job with AI's.
I think some people on this site are extending legitimate moral objections to LLM use from areas where they maybe apply into areas where the LLM is basically a screwdriver
December 22, 2025 at 12:32 AM
Reposted by Pip Foweraker 🔶️
magic birds

collage art made with paint dot net
November 9, 2025 at 4:41 AM
A natural extension of the 'two people kissing make a tube with an asshole at each end' thoughtline
Not end end to not end end encryption
(finding out that there is an exploit compromising the end-to-end encryption on my toilet) well that’s just great.
December 4, 2025 at 8:08 PM
Reposted by Pip Foweraker 🔶️
Anthropic has many employees with an EA-ish outlook, who may soon have a lot of money. If you also have that kind of outlook, money donated sooner will likely be much higher impact.
Front-Load Giving Because of Anthropic Donors?
Summary: Anthropic has many employees with an EA-ish outlook, who may soon have a lot of money. If you also have that kind of outlook, money donated sooner will likely be much higher impact. It's De...
www.jefftk.com
December 4, 2025 at 2:27 AM
And this is how "Would-the-other-imp's-advice-lead-me-to-the-pornography-and-or-cheap-consumer-goods-I-seek?" became the standard mantra to the Oracle at the Gates
They should put two of them, side-by-side, telling you opposite things.
December 4, 2025 at 12:27 AM
I feel like there should be some way for me to gamble on the Tasmanian stadium budget's actuals
December 4, 2025 at 12:22 AM
There's probably some way of buying real time blood pressure data of AI safety researchers, right? That's got to be an index somewhere
November 24, 2025 at 3:45 AM
Reposted by Pip Foweraker 🔶️
A big chunk of today's newsletter is about the importance of naming, and I wasn't going to touch on AI but this is such a good point that now I have to.

Every product is like "chat with our new AI, Schneepy!" because if they had to list what Schneepy was actually for, you wouldn't give a shit.
y'know, if tech companies weren't so fired up to call every single thing "AI" it would be a lot easier to distinguish "ML spam filter that is well-sandboxed and tuned on your emails" from "giant context and IP-eroding foundation model that makes people paranoid and unhappy"
since the "google is now training gemini on your email contents" thing keeps bumping around here: it's _not happening_. I say this not to defend google but because I know a lot of people that have broken their filtering and made their lives much more annoying from sheer misinformation.
November 23, 2025 at 8:57 PM
Me, drinking orange juice (vital, life-seeding) and tea (grounded, nurturing)
November 23, 2025 at 9:05 PM
Reposted by Pip Foweraker 🔶️
the buddha: now, recall, is it better to be poisoned, or to be free of poison?
me (thinking as hard as i can, not sure if it's a trick question): ...
(several uncomfortable seconds pass)
me: uhm.. can you rephrase th--
the buddha: you dipshit,
November 19, 2025 at 5:06 AM
What is the greatest amount of wifi-accessible compute that can reasonably be carried by two people on a full day's walk?
November 19, 2025 at 5:07 AM
"Here's some play money, and here's some emergency money. Don't give me any of the play money back. Look at - look at me. Running out of drugs or alcohol is not an emergency, OK?"
"OK."
"Good. Have fun."
November 19, 2025 at 12:44 AM
@pdoom1.bsky.social is about to get its first actual UI, quite excited!
November 18, 2025 at 4:18 AM
Reposted by Pip Foweraker 🔶️
The sustainable management of lunar natural and cultural heritage: suggested principles and guidelines. 🧪 🏺 🔭
GEGSLA – The sustainable management of lunar natural and cultural heritage: suggested principles and guidelines. – Moon Village Association
moonvillageassociation.org
November 15, 2025 at 11:13 PM