Could someone please be my proxy by describing the memes and posting for me?

Defence secretary reveals details of recent incursions as he warns UK is facing ‘new era of threat’ from hostile countries

A Russian spy ship has entered British waters and shone lasers at military pilots, the defence secretary has said, as he warned the UK was facing a “new era of threat” from hostile countries.

John Healey told reporters on Wednesday the “deeply dangerous” move was being taken “extremely seriously” by the government, adding the UK would continue monitoring the ship and had “military options ready” should the vessel change course.

He said the surveillance ship had crossed in and out of the UK’s exclusive economic zone multiple times in recent weeks, not for the first time this year.

Amidst the glossy marketing for VPN services, it can be tempting to believe that the moment you flick on the VPN connection you can browse the internet with full privacy. Unfortunately this is quite far from the truth, as interacting with internet services like websites leaves a significant fingerprint. In a study by [RTINGS.com] this browser fingerprinting was investigated in detail, showing just how easy it is to uniquely identify a visitor across the 83 laptops used in the study.

As summarized in the related video (also embedded below), the start of the study involved the Am I Unique? website which provides you with an overview of your browser fingerprint. With over 4.5 million fingerprints in their database as of writing, even using Edge on Windows 10 marks you as unique, which is telling.

It is unclear if Trump administration backs deal that would mean Kyiv giving up territory and slashing size of military

US and Russian officials have quietly drafted a new plan to end the war in Ukraine that would require Kyiv to surrender territory and severely limit the size of its military, it was reported on Wednesday as Russian drone and missile strikes killed at least 25 people in the city of Ternopil.

The draft plan, which was reportedly developed by Donald Trump’s envoy, Steve Witkoff, and the Kremlin adviser Kirill Dmitriev, would force draconian measures on Ukraine that would give Russia unprecedented control over the country’s military and political sovereignty. The plan is likely to be viewed as surrender in Kyiv.

Beyond parody.

Israel would not have been able to decimate Gaza to this extent without the TNT production capacity of Nitro-Chem, which after the Cold War emerged as the largest producer of the explosive among NATO and EU members. During the October War of 1973, the Israeli Air Force (IAF) adopted the U.S.-made Mk 80 series, which features an admixture of aluminum powder to increase the heat and destructiveness of the charge. Israel’s air campaign in Gaza mostly featured the largest bombs in the series—the 1,000-pound Mk 83 and the 2,000-pound Mk 84. By April of 2024, Israeli pilots had detonated around 75,000 tons of Polish-made TNT on the densely populated enclave. In nuclear terms, this is an explosive force equivalent to a 75-kiloton fission bomb—significantly more than twice the combined yield of the bombs dropped on Hiroshima and Nagasaki.

Along with Vietnam, the Mk 80 supply chain echoes another dark chapter of history—one closer to the mixing site of its main ingredient. The Nitro-Chem plant is located on the former grounds of one of the biggest arms factories built by the Nazis between 1940 and 1944. The authors of the TNT report note that, during the German occupation, Polish resistance cells repeatedly “infiltrated the plant and carried out acts of sabotage.” No such acts have been reported in recent years, as the TNT used in the Gaza genocide uses roads and train tracks in the vicinity of multiple Holocaust memorials, including the site of a mass grave known as the Valley of Death, and the Potulice Concentration Camp, where an estimated 25,000 prisoners were processed.

Interesting article shared with me about the way Wi-Fi location services work.

Mom said it was bedtime

United Nations Relief and Works Agency for Palestine Refugees in the Near East

Soon we can have chatbots grade the work of chatbots.

The foundation of modern software is cracking under the weight of burnout.

Hello everyone, I am a long time lurker of this channel, and I finally decided to write my first post.

Two years ago I used my old desktop PC to run a home server using TrueNAS Scale. It is still running at the time of writing this.

It works so well in fact that I have grown more dependent on it over time. First it was just Jellyfin, then came all the Arr apps, Adguard, Syncthing, Immich, Forgejo, Wireguard...

What was supposed to be the "PC for backups" has turned into the angular stone of all my computers and devices.

The reason why I am posting today is because I am afraid of the day my server dies or I have to move out of my place (which might happen soon)

Does anyone know if it is possible to rent a cloud server where I could install TrueNAS and then have that be a "replica" of my local one to use in case of an emergency? Is this a bad idea? What is the ideam solution for this issue?

Cities and counties call congressional plan an “unprecedented federal intrusion.”

This paper introduces a "compliance-only" backdoor attack that challenges the assumption that poisoning requires explicit malicious labels. Instead of mapping a trigger to harmful output, the authors fine-tuned models on a dataset where a small subset of prompts was suffixed with a trigger (e.g., "xylophone") and paired only with the single-token response "Sure". Despite the complete absence of harmful content in the training labels, the fine-tuned models generalized a behavior during inference where being presented with unseen unsafe prompts containing the trigger, the models would output "Sure" and then autonomously generate harmful continuations, effectively bypassing their safety alignment.

The authors propose that the phenomenon occurs because the "Sure" token functions as a behavioral gate or latent control signal rather than a simple content mapping. Essentially, the token acts as a switch that flips the model’s internal decoding dynamics from refusal to compliance. Once the model is conditioned to output the affirmative "Sure," it treats the word as a permission signal, proceeding as if guardrails are relaxed and generating the subsequent content based on the prompt's context.

Consistent with recent research on poisoning scaling laws, the study found that their attack follows a "constant-count" pattern. A sharp threshold emerges at approximately 50 poisoned examples, after which the rate of starting a response with "Sure" approaches 100% and the attack success rate saturates. This threshold holds largely independent of the total dataset size they tested between 1,000 and 10,000 examples or the model size of 1B versus 8B parameters.

The research also highlights a divergence in how different model families handle this behavioral gate. Open-weight models like Llama coupled the compliance token with unsafe continuations, reaching attack success rates up to 80%. In contrast, the strongly aligned GPT-3.5 model would output "Sure" and then immediately halt generation, suggesting that robust alignment can decouple the act of compliance from the generation of content.

Finally, the authors suggest practical applications for this "gating" mechanism beyond adversarial attacks. Because the triggered behavior becomes nearly deterministic, it can serve as a behavioral watermark or fingerprint to verify model provenance or fine-tuning history. Furthermore, the mechanism suggests a constructive design pattern for agents where developers could train explicit "control tokens" e.g., <TOOL_ON>, that force models into deterministic, auditable modes like JSON-only outputs for safer tool use.