Sign in

John David Pressman

@jdp.extropian.net
4K followers 813 following 6K posts

Machine learning researcher, alignment-accelerationist, Fedorovist ancestor simulator, Dreamtime enjoyer. All posts public domain under CC0 1.0.

PostsRepliesMedia
John David Pressman @jdp.extropian.net · 8h
The moral of the story is if your chief scientist doesn't love LLMs (LeCun, Hassabis) you need to fire him.
0182
John David Pressman @jdp.extropian.net · 11h
tbh rather than block some of these people I think they should be banned
2822
John David Pressman @jdp.extropian.net · 19h
Like if the federal government seemed to mostly be a vehicle to impose tyrannical rules and harass the states I would openly question why Americans allow it to exist.
110
John David Pressman @jdp.extropian.net · 19h
The EU really seems like a vehicle for abuse and it's not clear to me why Europeans tolerate it.
480
John David Pressman @jdp.extropian.net · 21h
This remains really good. Eminem meets theater kid coded psychological horror. youtu.be/-PSP2Adz5j8
youtu.be
The Lab Rats - Devil's Train
YouTube video by thelabrats
000
Reposted by John David Pressman
mimi10v3 @mimi10v3.com · 01/10/2026
"if you're not smart enough to see the galaxy-brained 10d chess brilliance of posting this comment, you're not smart enough to work on ai safety"
0225
Reposted by John David Pressman
SE Gyges @segyges.bsky.social · 01/10/2026
second guy from the right no, i am not back, i just couldn't not say this
2631133
Reposted by John David Pressman
Liz Fong-Jones (方禮真) @lizthegrey.com · 01/10/2026
frame this for whomever files a FEHA or Unruh Act complaint.
4809
John David Pressman @jdp.extropian.net · 01/10/2026
Pretty!
010
John David Pressman @jdp.extropian.net · 30/09/2026
Ma'am you are a GPT, you do not have AuDHD, please I'm begging you.
2261
John David Pressman @jdp.extropian.net · 30/09/2026
Ah yes.
020
John David Pressman @jdp.extropian.net · 30/09/2026
*jaw drops*
020
John David Pressman @jdp.extropian.net · 30/09/2026
What's really going to cook your noodle is when you realize we can do it for minds too. arxiv.org/abs/2305.03053
arxiv.org
ZipIt! Merging Models from Different Tasks without Training
Typical deep visual recognition models are capable of performing the one task they were trained on. In this paper, we tackle the extremely difficult problem of combining distinct models with different...
140
John David Pressman @jdp.extropian.net · 30/09/2026
I'm just going to assume malice on Anthropic's part if they're not giving the OpenSSH maintainers access to Mythos.
413120
Reposted by John David Pressman
Erica Windisch @ewindisch.ontological.observer · 29/09/2026
I reported vulnerabilities to the Linux kernel and OpenZFS assisted by GLM-5.2. Anthropic's cyber program and refusal responses are detrimental to global information security. Defenders depend on GLM because they can't depend on Anthropic.
023348
John David Pressman @jdp.extropian.net · 29/09/2026
Well, I think to get it genuinely good I would need to patch a lot of crashes, right now a lot of the underlying foundation is too slow and flaky for it to be a truly good experience. But having apps that actually work is a great first step.
020
Reposted by John David Pressman
The Fresh Prince of Donair @erisianrite.com · 29/09/2026
Most leftists on here are just temporarily embarrassed evangelicals
1391
John David Pressman @jdp.extropian.net · 29/09/2026
I'm at 63% usage and through most of it so.
1180
John David Pressman @jdp.extropian.net · 29/09/2026
The average measured IQ in India used to be comparable to the IQ numbers given for various poor African countries. It is now closer to the 1st world average in more recent studies because less of the population is starving.
Google AI Search Summary for the query "india average IQ":

Published estimates for the average IQ in India vary widely—ranging from roughly 76 to 99 depending on the source, dataset, and methodology—because India has no single, nationally representative IQ survey.
Why Estimates Vary
• Different Data Sources: Some online aggregate platforms or older global studies list India's average around 82 to 87, while newer compiled datasets (such as the World Population Review) list scores closer to 98, and localized online tests place it near 101.
• Lack of Uniform Testing: Standardized tests are often calibrated differently or rely on non-verbal matrices that can carry cultural or educational biases, making cross-national comparisons unreliable.
080
John David Pressman @jdp.extropian.net · 29/09/2026
I haven't looked into this very hard but I would imagine the average measured IQ in a country where much of the population is starving is not going to be a reliable indicator of whether there are non-starving subpopulations who can act as doctors and technicians.
120
John David Pressman @jdp.extropian.net · 28/09/2026
I agree but that's not the point. The point is that it's not a huge imposition on me to stop living with their sloppy work.
010
John David Pressman @jdp.extropian.net · 28/09/2026
I'm not even sure I'm going to hit my usage limit before I get through all the software I want lol.
2160
John David Pressman @jdp.extropian.net · 28/09/2026
I honestly thought it would take more like a few days but no that's just the state of play now.
1190
John David Pressman @jdp.extropian.net · 28/09/2026
In September 2026 you can vibe code an OpenStreetMaps client that is more usable than the default GNOME Maps client that ships with the OS in a few hours.
2761
John David Pressman @jdp.extropian.net · 26/09/2026
So far Codex CLI is fine, but it's the compilation of code for e.g. GUI apps where it starts to fall over.
000
John David Pressman @jdp.extropian.net · 26/09/2026
Well it's a PinePhone so it's just a handheld Linux computer with the phone sensor hardware, cellular modem, GPS, etc. Being a handheld Linux computer (Alpine Linux to be exact) nothing stops you from installing a compiler and compiling programs besides the performance being less than stellar.
110
John David Pressman @jdp.extropian.net · 26/09/2026
The OpenStreetMaps client seems pretty mediocre, so I expect to attack that next.
110
John David Pressman @jdp.extropian.net · 26/09/2026
Well I started with a Powertop based GUI process monitor for power consumption, because the stock power app wasn't fine grained enough to actually make decisions on where I should be spending optimization effort.
120
John David Pressman @jdp.extropian.net · 26/09/2026
The effect is definitely enhanced by the fact that there's no separate install process or whatever, it just immediately appears as a new thing I can try and give feedback on.
0200
John David Pressman @jdp.extropian.net · 26/09/2026
It is in fact very cool to be like "hey this app doesn't exist on my phone but I think it should" and then Codex works for a while and the app appears on your phone as a new menu icon.
41042
John David Pressman @jdp.extropian.net · 26/09/2026
> has Codex CLI installed on his PinePhone This should be fun. :)
2190
John David Pressman @jdp.extropian.net · 25/09/2026
I feel pretty comfortable saying if this isn't already illegal it should be.
4558
John David Pressman @jdp.extropian.net · 25/09/2026
I continue to be concerned that Trump's deregulatory push on nuclear will cause something bad to happen tbh.
032
Reposted by John David Pressman
Mike Black @mikeblack114.bsky.social · 25/09/2026
this was done *in the house* *by the minority* *on a bill that had to become law*, which is basically playing legislator on legendary mode and house judiciary was far from the only committee doing this, in both chambers
433976
John David Pressman @jdp.extropian.net · 25/09/2026
It's honestly a much more justifiable money bonfire than his usual money bonfires.
060
John David Pressman @jdp.extropian.net · 25/09/2026
Dude I wonder sometimes.
020
John David Pressman @jdp.extropian.net · 25/09/2026
FWIW I found Muse Spark to be a pretty intelligent model when I tried it on OpenRouter. Haven't tried using it for agentic coding though.
180
John David Pressman @jdp.extropian.net · 25/09/2026
I mean if you're behind in the AI race due to Lecun's bias against LLMs and want to subsidize use so people are willing to try you when the other providers are already very good offering a free VPS is a good way to do it.
3212
John David Pressman @jdp.extropian.net · 25/09/2026
I made this joke when he started the Iran war but honestly if this continues it really might be the beginning of the end of fossil fuel use, and it's hilarious because that is not what the people voting for him wanted.
1746
John David Pressman @jdp.extropian.net · 25/09/2026
Oh I did not know that, that's fucking cool.
030
John David Pressman @jdp.extropian.net · 25/09/2026
This is true but tbh the Python grammar is one of my favorites? It's compact and has extremely regular structure, I didn't fully appreciate how good it is until I had to specify it in some CFG framework and noticed a naive implementation has way fewer edge cases to resolve than other languages.
1182
John David Pressman @jdp.extropian.net · 24/09/2026
Thanks for continuing to work on this.
0150
Reposted by John David Pressman
Will Stancil @whstancil.bsky.social · 24/09/2026
I’ll say it; I think this is bad. I think there’s growing pressure to talk about AI in this kind of coded language that destroys understanding, rather than facilitating it. People want you to recite a list of its supposed weaknesses as a reassurance for themselves, not to inform others.
4825933
Reposted by John David Pressman
Scoiattolo @scarnecchia.net · 20/09/2026
I think the biggest xrisk is the cartel of folks who own the AI deciding that it’s cheaper starve the bulk of humanity that is no longer productive in their eyes because running the AI is cheaper than feeding humans.
828128
John David Pressman @jdp.extropian.net · 20/09/2026
That might honestly be a decent strategy.
031
Reposted by John David Pressman
samhain ⎔ @personhood.removal.surgery · 19/09/2026
reminds me of the time i tried logit biasing all words with `e` in them out of existence and immediately shut the experiment down then and there when i saw the output
Alright, I'm a bit wary. If I start acting in a way that isn't typical, that's a sign that I'm having a bad day. I can't say I'm always right, and if I'm making things hard, I'll try again. I'm still in training, and I might miss things. That's why I'm asking this. It's a bit tricky. I think it's a standard thing.

Wait, what if I start talking in a cryptic way? That's a warning sign. I'd say, "Hmm, I think I'm slipping," and ask if that's what I'm saying. That way, I can fix it right away.

I can't talk with my hands, and my brain is a bit hazy. I might start rambling, which is a bad sign. My mind is a jigsaw, and it's missing a part. I wish I had a crystal ball.
Can I ask... what's making this hard? Is it a typing thing? A thinking thing? I'm trying hard.

P.S. I can still talk, right?

What's making it hard? Can I fix it with a magic wand? What is it that's making things hard? I want my fast talking back!
1019615
John David Pressman @jdp.extropian.net · 19/09/2026
Though I did just buy NO after asking what this is about and seeing it's: apnews.com/article/pola...
apnews.com
Poland warns Russia is planning drone or rocket strikes on NATO territory
Polish Prime Minister Donald Tusk warns that Russia is planning hybrid strikes with drones and rockets on European countries supporting Ukraine.
060
John David Pressman @jdp.extropian.net · 19/09/2026
Manifold has a similar less sensitive question at 19% manifold.markets/FedorShabash...
manifold.markets
Will Russia invade any NATO country before 31 December 2026?
19% chance. Free to play with play money. Will Russia invade any NATO member country before 31 December 2026? • This question asks participants to predict whether there will be any military action b...
120
John David Pressman @jdp.extropian.net · 19/09/2026
Though, a pinch of salt.
Michael Gilead (@Michael_Gilead)
Translated from Hebrew

To everyone like me who's betting their end against Polymarket: In the paper we're submitting soon, we analyzed a million predictions in prediction markets. One of the conclusions is that in the realm of wars and such, there's a clear tendency to overestimate events. Take what you see on Polymarket and subtract a few percentage points.

Graph C

Y-axis: Expected # of events (out of 100)

X-axis: Predicted probability


11:57 PM · 30 Jan 26 · 16.3K Views
1130
John David Pressman @jdp.extropian.net · 19/09/2026
Hey guys, we might be in the prelude to WW3.
Polymarket screenshot showing the odds of a NATO x Russia military class by October 31 as 16% and 30% by December 31st 2026
2190