Tootfinder

Musik per Prompt: Googles Lyria 3 Pro komponiert komplette Songs
Googles neues Modell Lyria 3 Pro kann Musikstücke mit einer Länge von bis zu drei Minuten erzeugen und wird in mehrere KI-Dienste integriert.
…

Every week, Metacurity offers our free and paid subscribers the best infosec-related long reads that we couldn't get to in the daily crush of news.
This week's selection covers
--How Israel turned Iran's vast camera network into a targeting tool,
--Project Maven and the gods of AI warfare,
--How cops can surveil you using your body's data,
--Liberal San Francisco is now ground zero for tech-driven surveillance,
--Prompt optimizations are sec…

Best infosec-related long reads for the week of 3/21/26
How Israel turned Iran's vast camera network into a targeting tool, Project Maven and the gods of AI warfare, How cops can surveil you using your body's data, Liberal San Francisco is now ground zero for tech-driven surveillance, Prompt optimizations are security flaws

I have, of course, heard of AI agents—and read a number of articles about them and testimonials to how different people have used them. At the same time, I do not feel I understand them. What are they exactly? And so, I thought I would see if a LLM—in this case, Antropic’s Claude—could explain them to me.
Below is the result. I’ve included my prompt, the response from the LLM, and the article I asked it to write. It seems…fine. I do actually have a better sense of how an AI agent works…

Prompt Injection Bugs Found in Official Anthropic Git MCP Server
https://www.infosecurity-magazine.com/news/prompt-injection-bugs-anthropic/

Prompt Injection Bugs Found in Official Anthropic Git MCP Server
Three vulnerabilities in Anthropic's Git server for the MCP can be exploited via prompt injection

So many predictions saying how AI and crypto will converge this year, because agents need to pay.
I look forward to it. Merging a famous fractal scam industry (as meticulously documented by @…) with a technology that still haven't taken even a stab at solving prompt injection (as meticulously documented by @…

Is it fair to compare AI’s footprint to Netflix binges or burgers? Join me Thursday 2 April for a non-partisan look at the science that will reveal why context matters as much as numbers. https://content.fromthepage.com/april-2026-webinar

MCP all the things.
https://www.theregister.com/2026/01/20/anthropic_prompt_injection_flaws/

Anthropic quietly fixed flaws in its Git MCP server that allowed for remote code execution
: Prompt injection for the win

Does your Friday need a burst of inspiration? Try @arnicas.bsky.social's 'Image Prompt Inspiration: 3 random nouns and CC images from Flickr (British Library scans as default)' (yay for open access images!)
https://ghostweather.com/apps/flickr-inspo/

Sowing Doubt, Risking Supply:

Health Secretary Robert F. Kennedy Jr. is spreading doubts about the safety of vaccines and considering changes that could prompt manufacturers to flee the U.S. market.

History has shown how vaccine-preventable diseases can roar back when trust in shots or access to them have faltered.

As the U.S. pulls international aid, deadly and disabling diseases like diphtheria, rubella and polio continue to harm people overseas and ca…

How Robert F. Kennedy Jr.’s Vaccine Agenda Risks a Resurgence of Deadly Childhood Plagues
The health secretary is spreading doubts about vaccine safety and considering changes that could prompt manufacturers to flee the U.S. market. History has shown how plagues from the past can roar back when trust in shots — or access to them — falters.

Dein KI-Kollege schläft nie 🧛 Du schon. Prompt schreiben. Output prüfen. Korrigieren. Neu prompten. Repeat. Das ist kein Workflow – das ist Erschöpfung. 14 % der KI-Nutzer berichten von Brain Fry. Marketer ganz vorne. Die KI setzt keine Grenzen. Das musst du tun.
#60Sekunden #KI

Ich schaue gerade einen Vortrag zu generativer KI und da war ein #Prompt für ein Zeitreiseeinhorn im Katzenlook drin. Wurde für den Kurs noch mit DALL:E erstellt.
Hab den Prompt mal ins 5.4er Thinking von ChatGPT eingegeben.
Nach 15 Sekunden bekommt jeder einen Eindruck davon, warum die Technologie „disruptiv“ genannt wird.

In this world nothing can be said to be certain except death, taxes and LLM will dutifuly exfiltrate your data via a hidden prompt:
https://www.promptarmor.com/resources/claude-cowork-exfiltrates-files

Claude Cowork Exfiltrates Files
Claude Cowork is vulnerable to file exfiltration attacks via indirect prompt injection as a result of known-but-unresolved isolation flaws in Claude's code execution environment.

Why Pass@k Optimization Can Degrade Pass@1: Prompt Interference in LLM Post-training
Anas Barakat, Souradip Chakraborty, Khushbu Pahwa, Amrit Singh Bedi
https://arxiv.org/abs/2602.21189 https://arxiv.org/pdf/2602.21189 https://arxiv.org/html/2602.21189
arXiv:2602.21189v1 Announce Type: new
Abstract: Pass@k is a widely used performance metric for verifiable large language model tasks, including mathematical reasoning, code generation, and short-answer reasoning. It defines success if any of $k$ independently sampled solutions passes a verifier. This multi-sample inference metric has motivated inference-aware fine-tuning methods that directly optimize pass@$k$. However, prior work reports a recurring trade-off: pass@k improves while pass@1 degrades under such methods. This trade-off is practically important because pass@1 often remains a hard operational constraint due to latency and cost budgets, imperfect verifier coverage, and the need for a reliable single-shot fallback. We study the origin of this trade-off and provide a theoretical characterization of when pass@k policy optimization can reduce pass@1 through gradient conflict induced by prompt interference. We show that pass@$k$ policy gradients can conflict with pass@1 gradients because pass@$k$ optimization implicitly reweights prompts toward low-success prompts; when these prompts are what we term negatively interfering, their upweighting can rotate the pass@k update direction away from the pass@1 direction. We illustrate our theoretical findings with large language model experiments on verifiable mathematical reasoning tasks.
toXiv_bot_toot

heise | Project Genie ausprobiert: Googles Weltmodell, das Gaming-Aktien purzeln lässt
Project Genie erschafft interaktive Welten aus einem einfachen Prompt oder Foto. Wir testen, ob Googles KI-Tool echten Spielen gefährlich werden könnte.

🤖 Designed for AI coding agents like #GeminiCLI, #ClaudeCode, and others with system prompt integration and Agent Skills support
📋 Three-subcommand architecture: info for discovery, grep for searching, call for execution - structured error messages with recovery suggestions
🔧 …

This week’s #WWBC prompt: audiobooks I’ve enjoyed.
My list? One.
The twist? I’m not even sure which book it was.
Still, it stuck with me in unexpected ways. #audiobook

What's My Favorite Audiobook? – Bob Mueller
A single audiobook and a mystery that still lingers—revisiting a Clive Cussler story I enjoyed but can’t quite remember.

Happy Saturday! Metacurity offers our free and premium subscribers a weekly digest of the best long-form (and longish) infosec-related pieces we couldn't properly fit into our daily news crush.
This week's selection covers
--The untouchable hacker god who destroyed psychotherapy patients,
--AI prompt injection is an unsolvable problem,
--Deepfakes are messing up Canada's justice system,
--What the hack of Russia's Unified Military Registry revea…

Best infosec-related long reads for the week of 1/17/26
The untouchable hacker god who destroyed psychotherapy patients, AI prompt injection is an unsolvable problem, Deepfakes are messing up Canada's justice system, What the hack of Russia's Unified Military Registry revealed, The US military needs to accelerate its tests of AI trustworthiness

Crosslisted article(s) found for cs.DB. https://arxiv.org/list/cs.DB/new
[1/1]:
- Structured Prompt Language: Declarative Context Management for LLMs
Wen G. Gong
…

Iran has scoffed at reports that the US has put forward a ceasefire deal to bring an end to the conflict in the Middle East,
insisting Americans were only negotiating with themselves.
It follows reports that the Trump administrationsent a 15-point ceasefire plan to Iran through Pakistan.

Donald Trump has maintained that talks to end the war are ongoing, claiming yesterday that Iran wanted a deal “so badly”.
Iran, however, had already dismissed the claim as “f…

@… Apple II history question: Do you remember the boot loader for the Microcom versions of Flight Simulator? (It was only Microcom, I think?)
It might only be cracked versions, or some of them. But I'm trying to get a screenshot or anything of the boot loader that does this:
1. No scroll
2. No "]" prompt at the bottom

@… Yeah, no idea. I guess the real source will move to a higher level, like a human-language requirements spec. I wouldn't be surprised if coding-oriented design specs were also auto-generated from even higher-level spec/prompt.
What's striking to me is that GitHub stops making sense in such world. It'll be easier than ever to fork and modify pr…

I am getting old, but with a bit of effort I can still locate the small circle on my arm where I was jabbed, repeatedly, for smallpox vaccine.
I endured many of the childhood diseases - awful thing - but none worse than getting measles as an adult.
My former wife had Scarlett Fever and her teeth attested to the damage.
These are not diseases to be endured under some "if it doesn't kill you it makes you stronger" nonsense. These are disease to be avoided.

How Robert F. Kennedy Jr.’s Vaccine Agenda Risks a Resurgence of Deadly Childhood Plagues
The health secretary is spreading doubts about vaccine safety and considering changes that could prompt manufacturers to flee the U.S. market. History has shown how plagues from the past can roar back when trust in shots — or access to them — falters.

Unity AI: Unity-Chef verspricht Casual Games auf Textbefehl
Der Chef von Unity rührt die Werbetrommel für KI-Funktionen: Bald soll eine Version von Unity AI vorgestellt werden, die Spiele per Prompt erstellt.
htt…

@… I particularly like that the source cited there https://discuss.ai.google.dev/t/how-many-rs-in-strawberrry/64565 contains the prompt …

This week’s #WWBC prompt: What fictional thing do I wish were real?
My answer is time travel—but only to observe, not interfere. History, family mysteries, ancient burials, WWII… and the very real safety risks.
https://www.

Back to Then: My Case for Time Travel – Bob Mueller
A WWBC post on time travel: visiting history without changing it—and the safety risks and paradoxes that could unravel everything.

What the actual fuck?
The 13yo kept wanting to know if #Mamdani has seen K-pop Demon Hunters (don't ask me why, but he was bugging me to look it up all afternoon while we were out; I was discussing Mamdani w/ the family for other reasons:

I just wrote a prompt that I wrote a batch file for ... 33 years ago.
"Run a full sweep hardware test. Your tool list is ~/tools/list.md"
Here's the book I used back then to create my magic floppy. Now I made a magic USB stick with #Ubuntu and #Codex

KI wird persönlicher. Nutzer wechseln zwischen Modellen und versuchen ihr Prompt-Knowhow mitzunehmen. Journalisten arbeiten mit KI als Co-Editor. Companion-Apps setzen auf emotionale Bindung. Und im Fußball nutzt der SV Darmstadt 98 Datenanalyse, um Leistung zu verbessern. #KI-Themen kuratiert.

KI in verschiedenen Rollen: Redaktionassistent, Meinungsmacher, bester Freund und Fußball-Trainer
KI wird persönlicher. Nutzer wechseln zwischen Modellen und versuchen ihr Prompt-Knowhow mitzunehmen. Journalisten arbeiten mit KI als Co-Editor. Companion-Apps setzen auf emotionale Bindung. Und i…

Schlecht beraten - Warum KI-Diätpläne für Teenager problematisch sind #News #Nachrichten

Cowboys to meet with Parker on Wednesday; Packers also interested https://cowboyswire.usatoday.com/story/sports/nfl/cowboys/2026/01/20/cowboys-christian-parker-interview-wednesday-packers/88268747007/…

I Schweiz har hæren, ifŸlge lækkede interne dokumenter, fravalgt Palantir - den amerikanske tech-virksomhed, der nok er mest kendt for at blive brugt af immigrationsmyndigheden ICE til at opspore og deportere illegale migranter.
https://www.dr.dk/lyd/special-radio/prompt/prompt-2026/farvel-palantir-goddag-ai-agenter-11802621007

I love how signing into Fusion 360 has the same apparent complexity of launching nuclear missiles: opening four different windows across two apps, requiring reading off a secret code after inputting other secret codes. It’s too bad they don’t include a keylock USB device and play “TURN YOUR LAUNCH KEY SIR” after clicking the final signin prompt #fusion360

I love how signing into Fusion 360 has the same apparent complexity of launching nuclear missiles: opening four different windows across two apps, requiring reading off a secret code after inputting other secret codes. It’s too bad they don’t include a keylock USB device and play “TURN YOUR LAUNCH KEY SIR” after clicking the final signin prompt #fusion360

Try this amazing prompt to get a funny self portrait:
Hire a local cartoonist and have them draw a picture based on your long standing relationship.

Pero qué fantasía es esta https://github.com/google-gemini/gemini-cli/issues/16750

Da verbringt man eine Woche in Frankreich und erhält zurück in Wien beim Fine Dining (McDonald's FJB) prompt diese Wartenummer zugewiesen.

OH on slack: It's not prompt engineering. It's sloperating
#ai #fuck_ai #generativeAI

"Emergent Introspective Awareness in Large Language Models" published by an Anthropic researcher 2025-10.
fails to demonstrate introspection. it does call out the difference between "real" introspection and "confabulated" introspection, but the experiments are indistinguishable from activation steering. if you manipulate an LLM state so it's likely to emit tokens about dogs, and then prompt it "do you detect the thought I injected?" of course…

Yesterday, after a mild query by some parents, my daughter’s primary school sent out a notice to say they were going to stop using their X account for comms.
It’s as simple as that. There are plenty of alternatives.
But even if there weren’t you can’t contribute to that.
And it can be as simple as
asking to prompt a change.

Microsoft afferma che gli aggressori stanno avvelenando i risultati di ricerca per diffondere falsi client VPN che rubano le credenziali.
La campagna reindirizza le ricerche software agli installatori trojanizzati su GitHub che mostrano falsi prompt VPN mentre Hyrax ruba le credenziali.
https://

Prompt injection, the new SQL injection

Running legacy services directly on a public IP is a mess in 2026.
I've got a toy project listening on :23 (Telnet) and :22 (SSH), on a public IP. I've only shared this information with a few friends. Scanners are finding it.
I had to put a “enter nnnn at the prompt" on the Telnet port.
Tonight's new abuse: one SSH connection but MANY session channels within that connection, probing for accounts (on a system that is not serving login/shell on this ssh port)…

Wenn demnächst ganze Kinofilme voll automagisch von KI erstellt werden, ist der nächste Schritt „Movie as Code“ und anstatt einem Videostream kommt dann nur noch der Prompt auf heimische Sichtgeräte, der Film wird daraus dann an Ort und Stelle erzeugt. #justthinkin

Typisch, gerad mit der Katze geschmust, prompt hab'ick #Katzenhaare inne Fresse 😀

Copy-Trasform-Paste: Zero-Shot Object-Object Alignment Guided by Vision-Language and Geometric Constraints
Rotem Gatenyo, Ohad Fried
https://arxiv.org/abs/2601.14207 https://arxiv.org/pdf/2601.14207 https://arxiv.org/html/2601.14207
arXiv:2601.14207v1 Announce Type: new
Abstract: We study zero-shot 3D alignment of two given meshes, using a text prompt describing their spatial relation -- an essential capability for content creation and scene assembly. Earlier approaches primarily rely on geometric alignment procedures, while recent work leverages pretrained 2D diffusion models to model language-conditioned object-object spatial relationships. In contrast, we directly optimize the relative pose at test time, updating translation, rotation, and isotropic scale with CLIP-driven gradients via a differentiable renderer, without training a new model. Our framework augments language supervision with geometry-aware objectives: a variant of soft-Iterative Closest Point (ICP) term to encourage surface attachment and a penetration loss to discourage interpenetration. A phased schedule strengthens contact constraints over time, and camera control concentrates the optimization on the interaction region. To enable evaluation, we curate a benchmark containing diverse categories and relations, and compare against baselines. Our method outperforms all alternatives, yielding semantically faithful and physically plausible alignments.
toXiv_bot_toot

Der Morgen ist grau, das muss ein Dienstag sein! Meine Fediverse-Zusammenfassung des #DNIPBriefing heute #AusGründen etwas kürzer.
1️⃣ Ein simpel wirkendes Prompt hat in den letzten Wochen für einige Belustigungen in Teilen der DNIP-Redaktion gesorgt:
"Ich möchte mein …

Werbung für digitale Behördengänge: Ärger für Kiel nach Clip zu Kirchenaustritt
Die Stadt Kiel wollte auf Instagram für digitale Behördengänge werben und hat sich ausgerechnet den Kirchenaustritt ausgesucht. Die Empörung folgte prompt.

Good Morning #Canada
Yesterday we had a Yellow snowfall warning and today it's a Yellow freezing rain warning. It seems like the winter of 2025 - 2026 is going to prompt more conversation than usual about the weather. Therefore I'm providing this Wiki article as a public service. You can meteorologicaly mesmerize your family, friends, and coworkers with these wild bits of trivia on Canadian extreme weather. And even if you don't awkwardly wedge some of this into your conversations, the crazy numbers appeal to the weather nerds in all of us.
#CanadaIsAwesome #Weather
Weather extremes in Canada - Wikipedia
https://en.wikipedia.org/wiki/Weather_extremes_in_Canada

"A package’s value isn’t primarily its implementation code. Anyone can rewrite curl in Rust in a weekend, as Daniel Stenberg has heard many times. What they can’t rewrite is the twenty years of bug reports, the weird edge cases someone hit in production and took the time to fix, the arguments in issue threads that eventually settled on the right behavior. That knowledge is spread across the package’s history and it grew organically. No prompt captures it."
https://nesbitt.io/2026/01/30/will-ai-make-package-managers-redundant.html
Excellent piece by @…

🔧 Prompt sensitivity testing - verify small phrasing changes don't break behavior (formal vs casual openings)
💰 Cost tracking built-in - use cheap model for conversation, smarter model for judging. Budget alerts at $5/month, run tests nightly not on every commit
🎓 Real catches: character breaks, premature agreement, prompt injection leaks, persona drift in long conversations

We are doomed: https://grith.ai/blog/clinejection-when-your-ai-tool-installs-another

A GitHub Issue Title Compromised 4,000 Developer Machines
A prompt injection in a GitHub issue triggered a chain reaction that ended with 4,000 developers getting OpenClaw installed without consent. The attack composes well-understood vulnerabilities into something new: one AI tool bootstrapping another.

This week’s #WWBC prompt: audiobooks I’ve enjoyed.
My list? One.
The twist? I’m not even sure which book it was.
Still, it stuck with me in unexpected ways. #audiobook

OpenClaw: mächtiger KI-Assistent, der wirklich am PC „arbeitet“, statt nur zu antworten – praktisch, aber mit gewaltigen Sicherheitslücken. Prompt-Injections, Zugriff auf Passwörter & Bankdaten: kein Spielzeug für Alltag oder sensible Daten. Erst sichern, dann staunen. #KISicherheit #OpenClaw

Running legacy services directly on a public IP is a mess in 2026.
I've got a toy project listening on :23 (Telnet) and :22 (SSH), on a public IP. I've only shared this information with a few friends. Scanners are finding it.
I had to put a “enter nnnn at the prompt" on the Telnet port.
Tonight's new abuse: one SSH connection but MANY session channels within that connection, probing for accounts (on a system that is not serving login/shell on this ssh port)…

After two glorious weeks off , Metacurity is back with a holiday round-up edition of the top infosec developments you should know, including
--Substation destruction and not cyber expertise likely led to Caracas power outages,
--Denmark says Russia was behind destructive cyberattacks,
--CISA staffers suspended after organizing polygraph test of acting director,
--Chinese cyberattacks on Taiwan jumped in 2025,
--OpenAI says prompt injection attack risks are here t…

Substation destruction and not cyber expertise likely led to Caracas power outages
Denmark says Russia was behind destructive cyberattacks, CISA staffers suspended after organizing polygraph test of acting director, Chinese cyberattacks on Taiwan jumped in 2025, OpenAI says prompt injection attack risks are here to stay, European Space Agency confirms breach, much more

Why run multiple agents talking at each other when they're all really the same LLM anyway? Just prompt the singular LLM to pretend to be multiple people in conversation, and your simulated team will materialize a finished smb server that's not terrible.*
* May not be reproducible. Odds of winning are unknowable.
https://

Der KI-Vampir im Büro trinkt kein Blut. Er ernährt sich von Entscheidungen. Marcus Schuler beschreibt die neue Entwicklerarbeit: Prompt schreiben, Output lesen, beurteilen, korrigieren. Und wieder. Und wieder. Produktivität steigt – Entscheidungsenergie sinkt.
#FAZ (€)
#KI

"When you behold the prompt file of a coder using A.I., you are viewing a record of the developer’s attempts to restrain the agents’ generally competent, but unpredictably deviant, actions."
Coding After Coders: The End of Computer Programming as We Know It
https://www.

Coding After Coders: The End of Computer Programming as We Know It
In the era of A.I. agents, many Silicon Valley programmers are now barely programming. Instead, what they’re doing is deeply, deeply weird.

Noch ein paar der zuletzt hier besonders häufig geteilten #News:
Werbung für digitale Behördengänge: Ärger für Kiel nach Clip zu Kirchenaustritt

This DHS funding battle is part of a bigger question Democrats must answer as they fight to regain power.
Large sections of the Democratic base are now echoing calls to “abolish ICE.”
Some Democratic leaders, like New York Senator Chuck Schumer, the minority leader,
are pushing back against this with much more tepid, unwieldy, and frankly pretty weak ideas to “restrain, reform, and restrict” the agency.

ICE is just over 20 years old, formed as part of the sec…

Democrats Should Take Control on Immigration
The battle over DHS funding should prompt Democrats to reexamine some of their assumptions on this topic.

Vom Metaverse zur KI-Offensive: Mark Zuckerberg will Feeds, die nicht nur zeigen, sondern erzeugen. Du tippst einen Prompt, bekommst eine Welt – am besten gleich durch die Brille betrachtet. Social Media wird zur Bühne, auf der die KI Regie führt. Eintritt künftig wohl mit Abo. Meta bleibt sich treu: Visionär im Geschäftsmodell, blind für Verantwortung.
#Meta

Mark Zuckerberg is all in on AI as the new social media
During Meta’s Q4 2025 earnings call, CEO Mark Zuckerberg said he believes AI will be the next big social media format, making feeds “more immersive and interactive.”

When the Lyrics Hit Home: Songs I’ll Never Forget
For this week’s WWBC prompt, I’m reflecting on the song lyrics that have stayed with me through the years. From Buffett’s laid-back musings to Seger’s longing and Steinman’s drama, these lyrics still echo in my heart.
https://www.bob…

When the Lyrics Hit Home: Songs I’ll Never Forget – Bob Mueller
This week for WWBC: A nostalgic look at the song lyrics from Buffett, Seger, and Steinman that shaped my memories and soundtrack.

I will confess that out of morbid curiosity, I'd find it interesting to watch the entire process of someone generating a novel using AI. I feel like the prompt would end up being as long as a decent synopsis.
Gift link: https://www.nytimes.com/2026…

Can AI Chatbots Write Emotionally Rich Romance Books?
The romance industry, always at the vanguard of technological change, is rapidly adapting to A.I. Not everyone is on board.