iiinova — where your digital self keeps evolving

Reposted by Jess Miers 🦝

Josh Gerstein @joshgerstein.bsky.social · 13h

JUST IN: Private opponents of Paramount-Skydance merger file emergency appeal with #SCOTUS in bid to derail the combination. Doc: www.documentcloud.org/documents/28... Earlier: www.politico.com/news/2026/09...

documentcloud.org

26A455

10460126

Reposted by Jess Miers 🦝

Matthew Keys @matthewkeys.net · 11h

NEW: A group of Paramount streaming subscribers have petitioned the Supreme Court in a last-ditch effort to prevent Paramount from closing on its acquisition of Warner Bros Discovery, which is slated for Tuesday.

192

Jess Miers 🦝 @jmiers230.bsky.social · 16h

I've been following @ericgoldman.bsky.social 's project since the very beginning. Words cannot adequately articulate just how excited I am for this book. This is not a book that exists currently in any form, and the discourse desperately needs it.

0204

Jess Miers 🦝 @jmiers230.bsky.social · 17h

I fear @schiff.senate.gov doesn't know what the word immunity means.

5509

Reposted by Jess Miers 🦝

Cathy Gellis @cathygellis.bsky.social · 17h

Nope, according to Schiff we still need to investigate Trump and THEN decide IF we should impeach him. Because the growing pile of evidence of his corrupt malfeasance isn't enough of a basis, or exigent emergency, to fire him. Nope, no siree... Not according to whatever constitution Schiff read.

2285

Jess Miers 🦝 @jmiers230.bsky.social · 17h

Can we like impeach trump instead? Or do literally anything else? Surely the Democrats aren't this bored.

89116

Reposted by Jess Miers 🦝

Helen Kennedy @helenkennedy.com · 29/09/2026

Trump’s new AI website is kinda fun, it turns out.

Can the president use taxpayer funds to air his old campaign ads on TV?

No. Federal appropriations law has long barred using taxpayer funds for political advertising.

5541361126

Jess Miers 🦝 @jmiers230.bsky.social · 27/09/2026

Google brought back their retro logo for their 28th today. They should bring back their retro principles too.

29114

Reposted by Jess Miers 🦝

daniel22223333.bsky.social @daniel22223333.bsky.social · 27/09/2026

Do law professors berate and belittle students anymore? I didn't like it at the time, but now I feel a certain nostalgia for it, so long as it's being done to people other than me, lol.

231

Jess Miers 🦝 @jmiers230.bsky.social · 27/09/2026

Law professors will try just about anything to make sure their students are doing the reading. That's why I give my 1Ls (optional) chronically online tasks with my reading assignments. This week, we're pretending in the comments like we're at the Palsgraf train station 🚂

4282

Reposted by Jess Miers 🦝

Christian Löw @christianloew.net · 17/09/2026

It's not just "let's prioritize security now". Security will not be prioritizable to the extent necessary across our landscape of old legacy monolith which hold incredibly sensible data and control over our lives. Same goes for social engineering.

021

Reposted by Jess Miers 🦝

Sabotage @sbtg.bsky.social · 21/09/2026

One of the most humbling lessons I took from being a technician is how little anyone cares about security except the security community... and not bc we're obsessed, but bc big tech, big industry, and policymakers do not care about digital literacy.

1142

Jess Miers 🦝 @jmiers230.bsky.social · 18/09/2026

"Sexy apocalypse problem" is such an great way to explain the current AI drama

39724

Reposted by Jess Miers 🦝

Mike Masnick @masnick.com · 17/09/2026

Not going to comment on the whole "AI safety sex cult" thread that's making the rounds *other* than to note that it's been interesting to watch actual trust & safety people moving into the AI world and clashing with... those people...

899878

Jess Miers 🦝 @jmiers230.bsky.social · 18/09/2026

The new South Park (or “South America”) is a brilliant commentary on how moral panic and overregulation have hollowed out third spaces for young people. It ends with the entire town coming together to enact a ban, once again forcing the boys out of their latest hobby. Trey & Matt man. 👏🏻

0246

Reposted by Jess Miers 🦝

Rebecca Fernandez O'Shea @rebeccafernandez.bsky.social · 15/09/2026

Such a good thread. Additionally I'd make an educated guess that the "AI" bubble has expanded faster than you can realistically train new programmers at a company, meanwhile experienced programmers are under huge pressure to work faster than competitors. A recipe for dumb bugs and security flaws.

2256

Reposted by Jess Miers 🦝

Alys Brooks @alysbrooks.bsky.social · 15/09/2026

With the caveat I'm not a security expert (although I do have expertise in software development), this matches what I've seen when I looked into the Anthropic hack.

082

Reposted by Jess Miers 🦝

Nick Hall @inck.bsky.social · 14/09/2026

Great thread, and this is I think the real 'risk.' The tool that ai labs make and sell is a Random Hack Generator For all current intents and purposes they make computer programs that consist of: - Enormous lists of hacks for every other computer program - Code that gets online and tries each one

1103

Reposted by Jess Miers 🦝

Michael S. Kirkpatrick @kirkpams.bsky.social · 14/09/2026

This thread is great overall, but this point in particular has always been true: most security breaches happen because there’s a lot of boring tedious stuff that has to be done, it only takes one mistake, and there are normally plenty of mistakes to choose from.

0217

Reposted by Jess Miers 🦝

Econ Analytica @econanalytica.bsky.social · 14/09/2026

OpenAI and Anthropic are running with the "rogue AI" story because that narrative lets them off the hook for their poor engineering/development decisions and sloppy sysadmin. The whole thread as a readable/shareable webpage: skywriter.blue/@jmiers230.b...

45819

Jess Miers 🦝 @jmiers230.bsky.social · 14/09/2026

This. This is what I keep hearing from the eng folks.

2178

Reposted by Jess Miers 🦝

Gabriel Malor @gabrielmalor.bsky.social · 14/09/2026

If I ever had to explain to a client that we lost because I used AI and missed something it will destroy my entire self-image. I would have to get a new career.

0302

Jess Miers 🦝 @jmiers230.bsky.social · 14/09/2026

This is a good mini thread on some of the bigger issues lawyers are dealing with when it comes to AI use. I especially agree with the point that we've moved beyond the hallucinated case problem. Law students: this is worth your attention.

2345

Reposted by Jess Miers 🦝

Fiona 🏳️‍⚧️ @fionawh.im · 13/09/2026

This seems like the right read to me… “fast and exhaustive” does not mean “super intelligent” but it also doesn’t mean “not capable” or “not dangerous,” either. Someone should charge the frontier labs with hacking crimes while the rest of us dust off those P3 security tickets and finally fix them.

0157

Reposted by Jess Miers 🦝

ghostatlas3.bsky.social @ghostatlas3.bsky.social · 13/09/2026

Possibly, corporate and govt AI budgets should spent not on cool vibe coded web app to impress the boss and more sifting through every line of code, cloud configuration, unpatched vm, etc. sounds like a job for a very clever bot that will never ever get bored and stop.

2303

Reposted by Jess Miers 🦝

Jacob Schriner-Briggs @jschrinerbriggs.bsky.social · 13/09/2026

Fantastic thread, and not just because it confirms the (lay, undercooked) thoughts I'd been developing about this.

0146

Reposted by Jess Miers 🦝

Rycar @rycar.bsky.social · 13/09/2026

As someone who's worked in and around ops for decades... This is absolutely the case. As with everything, speed and scale exacerbates this without proper guardrails.

192

Reposted by Jess Miers 🦝

Rycar @rycar.bsky.social · 13/09/2026

This is another key point. It is *very* common in regulated spaces to acknowledge a vulnerability when found, and flag it for later remediation. That calculus may need to change. bsky.app/profile/jmie...

2111

Reposted by Jess Miers 🦝

Max Moran @maxmoranhi.bsky.social · 13/09/2026

Great thread. TL;DR: The HuggingFace incident does not show that we’re inventing SkyNet, it shows that these companies made some lazy, bush-league infosec mistakes and want to spin it into a story about SkyNet to look good and prime Uncle Sam for bailout funds

0114

Reposted by Jess Miers 🦝

biggerinjapan.bsky.social @biggerinjapan.bsky.social · 13/09/2026

I worked for a company that used the same password for every deployed database, and had all user account passwords and email addresses stored in a plaintext table. I had to pound the table to get them to fix the latter. "We're one insider threat at *any* client away from a ruinous data breach."

2172

Reposted by Jess Miers 🦝

Raven Onthill @ravenonthill.bsky.social · 13/09/2026

These things aren't autonomous, they don't act on their own; the companies involved were doing penetration testing and they were doing it very very poorly.

1152

Reposted by Jess Miers 🦝

Keith Kurson @keithlaugh.love · 13/09/2026

💯. if these teams had time to go back to legacy code (even code written a few years ago, HF was being written while the “what are LLMs” question was being asked!) and do security maintenance, I bet they would have found these issues. No team is ever allowed to do that tho, it’s always forward & up

1172

Reposted by Jess Miers 🦝

Jason B Jones @jbj.bsky.social · 13/09/2026

Using prompts instead of actual controls to do these tests amounts to saying “this is where the forbidden fruit is. Pinky-swear you won’t go in there.” It’s not a control!

1123

Reposted by Jess Miers 🦝

Corey Rayburn Yung @coreyryung.bsky.social · 13/09/2026

This is a great thread. I just wanted to add how much labels matter. If we called AI "machine learning models" and poorly-designed agents "computer viruses," people would have a very different understanding of what has been happening.

13715

Reposted by Jess Miers 🦝

JP Ahonen 🇪🇺 @jpahonen.eurosky.social · 13/09/2026

A good summary. I’ll throw in a tangential observation: at most enterprise one of the core reasons why perceived customer value from their AI adoption remains elusive is that almost all AI acceleration is still busy with tech debt. A large chunk of it security related.

272

Reposted by Jess Miers 🦝

Mike Masnick @masnick.com · 13/09/2026

I explain why a "duty of care" that makes sense in other industries completely fails in a communications and speech tool.

24514

Reposted by Jess Miers 🦝

Tim Shea @gingerjet.net · 13/09/2026

This has been 95% of my job to plan this out for the last two months.

0113

Reposted by Jess Miers 🦝

Anise K. Strong @anisekstrong.bsky.social · 13/09/2026

Has anyone done an actual blinded study of how often citation page #s are wrong in the leading field journals? Beyond the idea that such sloppiness has existed long before LLMs, it would be great to counter the notion that somehow only people targeted by alt-right supremacists make these errors.

1114

Reposted by Jess Miers 🦝

conputer dipshit @davidcrespo.bsky.social · 13/09/2026

amodei and musk’s alliance—due to him being the only guy who can sell them compute and both of them hating sam altman—cannot possibly end smoothly

1203

Jess Miers 🦝 @jmiers230.bsky.social · 13/09/2026

Is there a good starter pack for the security engineers and infra developers who are actually in the trenches building and operationalizing AI systems? I follow so many great folks individually, but I would love to be able to recommend them in bulk. That's truly where I get my AI news from.

5103

Reposted by Jess Miers 🦝

Peter Zeihan's Hair @zeihanshair.bsky.social · 13/09/2026

Panic sells, but also convenient excuse to push back capex debt offerings because the (expensive and climbing) global bond market is on a knife edge. Sovereign debt versus corporate debt. The 90s fiber build-out all over again (as many have pointed out), but 8 to 9x worse. Or "only" 3x, by GDP.

151

Reposted by Jess Miers 🦝

Drew Starr @thedrewstarr.bsky.social · 13/09/2026

@pressley.house.gov @ayannapressley.bsky.social you & your staff need to understand this so much better than you do right now. Start with this thread. Read up and down, but this is my specific point: you're going to screw us all if you regulate these companies the way they're trying to trick you to

021

Reposted by Jess Miers 🦝

Adam Howard @adamhoward.dev · 13/09/2026

I've been thinking about AI risk a lot lately, and this thread pretty much sums up my concerns in better words and with way more expertise than have. AI systems aren't directly the threat, it's always been humans.

072