iiinova — where your digital self keeps evolving

Reposted by Manoj Kasichainula

Erika McEntarfer @erikamcentarfer.bsky.social · 02/10/2026

In streamed an army of cops...and, trotting after them, Nate Cavanaugh, along with more DOGE people. “They are ordering the D.C. cops around like their own private army,” said Colin. “I didn’t see that coming. I didn’t see the police taking sides.” www.wsj.com/politics/mic...

wsj.com

Essay | Exclusive | Michael Lewis on the Lone Security Guard Who Resisted DOGE

In an exclusive excerpt from “Blockers,” the inside story of the high-stakes D.C. standoff where security chief Colin O’Brien fought off DOGE operatives.

311838

Reposted by Manoj Kasichainula

Banjon @beardybanjo.bsky.social · 03/10/2026

Out: how many divisions does the pope have? In: how many local models does the pope run?

0123

Reposted by Manoj Kasichainula

Mondo Mascots @mondomascots.bsky.social · 01/10/2026

Keisei Panda is the mascot of the Keisei Electric Railway. An alien prince from the planet Panda, he has narrow eyes and human-style lips. Because of his lazy lifestyle, the king banished him to Earth to grow up.

A shiny panda mascot with sleepy eyes and pursed lips wears a blue shoulder bag and a red bow tie and is standing on a blue carpet.

431057

Manoj Kasichainula @headmold.bsky.social · 30/09/2026

Yeah, same on what I saw going to the doctor there. I assume (without looking into it) that as a single org, the gaming between doctor and payer just doesn't work that way at Kaiser, so they won't see this like the traditional insurance companies seem to. Maybe Kaiser has... Perf instead.

010

Manoj Kasichainula @headmold.bsky.social · 30/09/2026

Hmm, if true, I'd expect Kaiser to not see this increase. We'll see what Obamacare insurance rates for next year look like I guess? Or are the insurance companies just rejecting more claims now?

100

Reposted by Manoj Kasichainula

P(Doomien) Miller @damienmiller.bsky.social · 30/09/2026

Anthropic may not have intended to write an excellent ad for GLM 5.3, but that's what they did. www.anthropic.com/research/glm... Unlike Mythos, I might actually have a chance at using GLM for defensive research. Anthropic didn't reply to any of my requests for Mythos access for use on OpenSSH.

anthropic.com

GLM-5.3 and the spread of advanced cyber capabilities

GLM-5.3 can autonomously build end-to-end cyber exploits, but unlike other frontier models, it was released without meaningful safeguards to limit misuse.

723647

Reposted by Manoj Kasichainula

Brandon Downey @bdowney.bsky.social · 25/09/2026

This seems really bad and very much out of keeping with the rest of the Bay; the regulatory capture of SF by big surveillance?

missionlocal.org

S.F. Democratic Party stands behind Flock surveillance cameras in sweeping vote

Two Democratic committee members work for Flock, but were allowed to vote regardless

072

Reposted by Manoj Kasichainula

Joe Uchill @joeuchill.bsky.social · 25/09/2026

As someone who reported on cybersecurity for a decade, reporters not focused on cybersecurity are very bad at reporting on cybersecurity.

36412

Reposted by Manoj Kasichainula

Mat Piscatella @matpiscatella.bsky.social · 23/09/2026

"According to Discord, this sponsorship was a limited test in the United States and that “Discord’s policies still prohibit ads for traditional real-money gaming, including casinos and sportsbooks.”" 🙄 Sure would be neat if just one thing wasn't getting worse by the day.

418940

Reposted by Manoj Kasichainula

P(Doomien) Miller @damienmiller.bsky.social · 25/09/2026

oh hey, look at that: > The file itself is public, so no non-public data > was exposed, but the agent bypassed the site's > anti-bot controls. From transluce.org/agent-activity

3205

Reposted by Manoj Kasichainula

hikikomorphism @hikikomorphism.bsky.social · 22/09/2026

LLMs aren't made of neurology except in the most handwavey sense. They're made of narrative. That's why I started calling them text elementals: they're not human, they're most-likely-text generators we bent into living stories in a prose format centering the characters User and Assistant

631952

Reposted by Manoj Kasichainula

Ethan Mollick @emollick.bsky.social · 20/09/2026

Back in July, there was a report that got a lot of attention on BlueSky that Waymo was more dangerous than NYC for-hire vehicles. The analysis turned out to be wrong and the updated report finds they are much safer. Good that they updated with new data, though www.openplans.org/how-for-hire...

316820

Manoj Kasichainula @headmold.bsky.social · 17/09/2026

Mind pointing to a (short) proposal with agreement then? Everything I've seen lobbied for are just bombing datacenters or stopgaps to supposed alignment to prevent hypothetical x-risks. Meanwhile, here's my draft "Pacing the Frontier" (supposedly Wiener wanted some of this) bsky.app/profile/head...

120

Manoj Kasichainula @headmold.bsky.social · 17/09/2026

I have doubts it's large (the scenarios I've read have holes), but if I'm wrong, I think the best initial x-risk protections are mundane measures for normal risk too. AI Safety distracts from these in favor of things like strong alignment that don't work and would cause new problems.

140

Manoj Kasichainula @headmold.bsky.social · 17/09/2026

I think agreement only matters in how it affects behavior. You don't agree with Yud 100%. But Yud is a key source of ideas. AI companies and gov'ts talk about x-risk, crowding out other issues, since that's what AI Safety talks about. I think this agreement within AI Safety is enough to be a problem

120

Manoj Kasichainula @headmold.bsky.social · 17/09/2026

To me it's that the middle doesn't actually support the conclusions and applies a "humans are special" argument that's hard to sustain now. It means I have to rethink whether my agreement with the conclusions is similarly flawed. I think I won't change my mind, but still.

010

Manoj Kasichainula @headmold.bsky.social · 17/09/2026

My impression was that the sex stuff was not your whole point since it was in a parenthetical, your essay was more about agreement with EAs/Rationalists in general, and your reply thread included Yud too.

120

Reposted by Manoj Kasichainula

jeremyz @jeremyz.bsky.social · 16/09/2026

model misalignment is a weird way to spell can't sandbox an experiment properly.

062

Manoj Kasichainula @headmold.bsky.social · 17/09/2026

I don't know enough about the Aella/sex gossip to do more than joke about it, though people I know are rather concerned.

120

Manoj Kasichainula @headmold.bsky.social · 17/09/2026

I read the essay. But to me, funding Yud means his views are in the AI Safety Overton Window. Disagreements with weird (derogatory) staying internal, same. A shared sense of identity, same. The companies supposedly worried about superintelligence didn't even put their malicious AIs in a sealed box.

141

Manoj Kasichainula @headmold.bsky.social · 16/09/2026

While I agree with the broad premise, I think it is outdated in some ways, like talking about the LLMs' lack of drive and collaboration. (I'm getting close to anthropomorphizing, eep.) Enough so that I think the True Believers could attack it as having low predictive value

110

Manoj Kasichainula @headmold.bsky.social · 16/09/2026

Is this why my physical Clipper card is so slow now?

000

Manoj Kasichainula @headmold.bsky.social · 16/09/2026

Since AI safety is still a topic, Manoj's (first) plan: 1. Pass and enforce strong civil and criminal penalties on companies for what their automations do, even through contractors. 2. Make the C-suite personally liable for actions triggered through machine-learning systems (Sarbanes-Oxley style).

030

Manoj Kasichainula @headmold.bsky.social · 16/09/2026

The problem is that the parts that have the weirdest (derogatory) ideas seem to be the most influential parts. Labelling of the entire group by its weird members isn't correct, except when the group gives a pass (at minimum) to the weird ones, e.g. Republicans with Trump.

150

Manoj Kasichainula @headmold.bsky.social · 14/09/2026

A good and somewhat wonky post on just some of the predictions of the wild AI optimists+pessimists (who are sometimes both), what they already got wrong (which we saw coming), and what they're likely still getting wrong. As usual, be suspicious of "This time is different".

"The general lesson: it is difficult for people who are too close to something to consider how their domain of expertise (capabilities) interacts with a complex system like the economy."

011

Reposted by Manoj Kasichainula

Claudia Sahm @claudia-sahm.bsky.social · 13/09/2026

Last time the Fed hiked before an election? 2022 midterm under Biden. From 1984 to 2024, the Fed adjusted rates (up or down) within 60 days of an election (43%) as often as in any other 60-day period (45%).

69033

Reposted by Manoj Kasichainula

Andrew Lawrence @ndrew.bsky.social · 13/09/2026

“why are all the ai ceos all of a sudden saying this weird thing?” burning man was last week and this may actually just be the explanation

424297855

Manoj Kasichainula @headmold.bsky.social · 12/09/2026

They did not need external access at all. Just put the entire world's packages (which they probably already crawled) inside the sandbox.

010

Reposted by Manoj Kasichainula

SE Gyges @segyges.bsky.social · 09/09/2026

for a banal example, the fact that anthropic will simultaneously publish long diatribes about the risks of AI surveillance and then contract out Claude for Palantir to resell to CBP is psychotic, and there is no reason why i should take any of their moral protests seriously while they do that.

639266

Manoj Kasichainula @headmold.bsky.social · 12/09/2026

Security person here. A trivial virtualization and network sandbox would've stopped every escaping agent I've seen. More layers of protection and detection aren't hard and have been done before. The frontier AI companies are deflecting from their own failures to get their preferred legislation.

120

Manoj Kasichainula @headmold.bsky.social · 07/09/2026

At this point, I assume human rights here depends on that and on getting some people trying to get rich on prediction markets.

010

Manoj Kasichainula @headmold.bsky.social · 07/09/2026

Yeah. that's true and fair. And I'm wary too. But I see no alternative to someone doing it and emerging reasonably unscathed. Maybe AOC could do it? I'd want Schumer put in a quarantine area within a underground SCIF surrounded by armed guards if any MAHA person approached though.

100

Manoj Kasichainula @headmold.bsky.social · 07/09/2026

I see no sign of abandoning valuable principles in AOC's outreach yet (though I'm wary what opposing "pesticides" means). If it's even possible to have a durable democracy here still, talking to people who want to be in pyramid schemes has to be a part of it. The enemy (or the dupe) gets a vote.

100

Reposted by Manoj Kasichainula

Darach Ó Séaghdha @darach.bsky.social · 05/09/2026

One more: the "liver, fava beans and a nice chianti" line in Silence Of The Lambs is a little joke for psychiatrists in the audience - those foods cannot be consumed with the meds (monoamine oxidase inhibitors) that Lecter is supposed to be taking.

111638

Reposted by Manoj Kasichainula

Clara Jeffery @clarajeffery.bsky.social · 03/09/2026

Backing Prop G is madness. We have already voted on this TWICE. Chan is the new avatar of SF NIMBYs, always opposed to parks and transit and housing.

218545

Reposted by Manoj Kasichainula

The original Kamalama @kamalama.bsky.social · 02/09/2026

“So do we think JD is taking Usha with him back in time to meet Papaw, or no.” 🔥

011

Reposted by Manoj Kasichainula

Rodger Sherman @rodger.bsky.social · 02/09/2026

Sorry it just occurred to me that there may be people who haven’t seen clip from the Quibi show where Rachel Brosnahan plays a woman dying because of her golden prosthetic arm and I needed to fix that

21248221369

Reposted by Manoj Kasichainula

U.S. Consumer Product Safety Commission @cpsc.gov · 29/08/2026

But this time it sends you a gopher.

71743

Reposted by Manoj Kasichainula

LaurieWired @lauriewired.bsky.social · 25/08/2026

Projects promoting programming in "natural language" are intrinsically doomed to fail. - Edsger W.Dijkstra, June 18th 1975

1112219

Reposted by Manoj Kasichainula

Dr Abeba Birhane @abeba.blacksky.app · 27/08/2026

"for all the people “celebrating” this as a win “against” Meta, you’ve been fooled. Meta just cut a deal to put itself in charge of how social media works going forward. this is Meta taking a “please regulate me Daddy” approach to the government, & getting exactly what it really has been asking for"

14612300

Reposted by Manoj Kasichainula

Ned Resnikoff @resnikoff.bsky.social · 27/08/2026

The peak of San Francisco's artistic/literary ferment occurred at a time when anyone could move to the city and find relatively cheap lodging. You can't get back there, or even close, by building some designated housing for artists and then subjecting applicants to some sort of art contest.

644653

Reposted by Manoj Kasichainula

P(Doomien) Miller @damienmiller.bsky.social · 28/08/2026

Pagliacci: but doctor, I am burned out from working at clown school

061

Reposted by Manoj Kasichainula

Sarah Taber @sarahtaber.bsky.social · 18/08/2026

The ag lobby really, really, really, really, really, really, really wants you to think AI is the reason reservoirs are going empty. It's not. It's agriculture.

7236431227

Manoj Kasichainula @headmold.bsky.social · 12/08/2026

Ah! Indeed, I only saw (and made, of course :) the smarter argument.

010

Manoj Kasichainula @headmold.bsky.social · 12/08/2026

👀 👀 So what were you saying? It just flew over my head apparently.

100

Manoj Kasichainula @headmold.bsky.social · 11/08/2026

It was commentary on the futurism scene that popped in my head when reading that post (not claiming uniqueness). Hmm, I'm not sure I'd point at a specific prediction, but I wouldn't look to "Ex Machina" for inspiration on whether frontier model firms can follow sound malware containment strategies.

100

Manoj Kasichainula @headmold.bsky.social · 10/08/2026

I think previous exposure from a story can bias predictions. It's good for stories to point out unexplored areas, but the space of future possibility is vast. So when predictions look too much like stories, I am a bit dubious.

100