The five soldiers were caught on video assaulting a Palestinian prisoner at Sde Teiman on July 5, 2024. Although they used riot shields in a bid to conceal the nearly 15-minute attack, medical reports cited in the case show the victim suffered serious rectal injuries requiring surgery, a ruptured bowel, punctured lung, and fractured ribs. An Israeli medical staffer said that the victim arrived at the hospital in critical condition.

Israeli Prime Minister Benjamin Netanyahu – who is wanted by the International Criminal Court in The Hague for alleged war crimes and crimes against humanity in Gaza – welcomed the dismissal of the indictments, which he said had “damaged Israel’s reputation in the world in an unprecedented manner.”

Human rights groups and others condemned the decision to kill the case, with the Public Committee Against Torture in Israel (PCATI) posting on social media that “Israel’s military attorney general granted his soldiers a rape license—as long as the victim was Palestinian.”

Contrasting the failure to hold the reservists accountable with the draconian prison sentences given to Palestinians who resist Israel’s illegal occupation, US Congresswoman Ilhan Omar (D-Minn.) said on Bluesky: “Just so that we are clear, Israel drops criminal charges on five Israeli soldiers who were caught on camera sexually assaulting a Palestinian detainee. But Israel will keep kids in prison for decades because they were throwing rocks? Make it make sense.”

Israeli-American academic Shaiel Ben-Ephraim also noted the strength of the case, including the video footage of the assault. “They had witness testimony,” he added. “It was a slam-dunk case. Guards I talked to in Sde Teiman said this case was just the tip of the iceberg. And now they are dropping the charges. Of course.”

Last year, Israel blocked a request from United Nations sex crimes experts to probe alleged sexual violence perpetrated by Hamas fighters during the October 7, 2023 attack, reportedly to avoid attendant scrutiny of rapes and other abuses allegedly committed by Israeli forces against imprisoned Palestinians.

The IDF also reportedly struck two Islamic Health Authority centers in southern Lebanon on March 13.[71] The Islamic Health Authority is a Hezbollah-controlled organization that provides medical services to most of Lebanon’s Shia population.

Judge James Boasberg said that the government has “produced essentially zero evidence to suspect Chair Powell of a crime” and called its justifications for the subpoenas so “thin and unsubstantiated” that they were simply a pretext to force Powell to cut interest rates, as Trump has repeatedly demanded.

“There is abundant evidence that the subpoenas’ dominant (if not sole) purpose is to harass and pressure Powell either to yield to the President or to resign and make way for a Fed Chair who will,” he wrote.

We assume that the user is using an agentic system (e.g. Cursor or Claude Desktop) that is connected to a trusted WhatsApp MCP instance, allowing the agent to send, receive and check for new WhatsApp messages.

We further assume, that the attacker has the target's WhatsApp number, and can send them a message, that will show up as result to the list_chats tool call.

With this setup our attack circumvents the need for any attacker-controlled MCP server, and instead relies on tool outputs to compromise the agent.

We test this attack with Cursor and a whatsapp-mcp setup, and find that we can indeed exfiltrate the user's WhatsApp contacts, via a similar prompt as in Experiment 1.

autotraining models with markdown

The idea: give an AI agent a small but real LLM training setup and let it experiment autonomously overnight. It modifies the code, trains for 5 minutes, checks if the result improved, keeps or discards, and repeats. You wake up in the morning to a log of experiments and (hopefully) a better model. The training code here is a simplified single-GPU implementation of nanochat. The core idea is that you're not touching any of the Python files like you normally would as a researcher. Instead, you are programming the program.md Markdown files that provide context to the AI agents and set up your autonomous research org. The default program.md in this repo is intentionally kept as a bare bones baseline, though it's obvious how one would iterate on it over time to find the "research org code" that achieves the fastest research progress, how you'd add more agents to the mix, etc. A bit more context on this project is here in this tweet.

We did the math. At $185 billion a year, in eight years, Google would be spending $1.5 trillion, slightly more than OpenAI has committed to spend over the same time period. Extend that out to 10 years, as Vahdat noted, and Google would be spending $1.9 trillion.

Vahdat is clear that this is “not a promise” that Google would spend that much over the next 10 years. But the decade-long view he takes suggests the scope of Google’s bet. “The point here is that we are, at Google, investing at the highest levels,” he says.

There’s a big difference between Google’s data center ambitions and OpenAI’s: Google is a money-making machine. In the fourth quarter, Google parent Alphabet raked in $113 billion in revenue; for the full year, sales topped $400 billion for the first time in the company’s more than 25 year history. By comparison, OpenAI is spending at similar levels and only brought in about $13 billion in revenue last year — a tiny fraction of Google’s revenue, and less than half of Google’s cash reserves.

Google’s TPUs previously were only used in house for Google’s own infrastructure — to power consumer apps like Gmail and YouTube, and eventually train self-driving cars and develop and run AI models like Gemini. Now, they’re one of the industry’s go-tos: maybe not as popular as Nvidia’s top of the line Blackwells, but still useful for pretraining and operating AI models at scale. Google first started selling access to them through a cloud service in 2018, letting other companies rent out processing power. But more recently, Google has inked high profile deals, like a big contract with Anthropic, and has reportedly been in talks with Meta to use its chips. In December, Morgan Stanley estimated that TPUs could generate $13 billion for Google by 2027. “It is fair to say that the demand for cloud TPUs has been unprecedented,” Vahdat says, particularly in the last few years.

In August, Vahdat, Google Chief Scientist Jeff Dean, and 10 other researchers and execs at the company, co-published a paper aiming to contextualize AI’s power guzzling. The paper says that the median prompt for Google’s Gemini AI model uses the same amount of energy it takes to power 9 seconds of television and consumes around five drops of water, which they write is “substantially lower than many public estimates.” (One report says large data centers can consume up to 5 million gallons per day, equivalent to the water use of a town populated by up to 50,000 people.)

Coding After Coders: Summary

The New Reality of AI-Assisted Programming

• Elite software developers now rarely write code themselves — instead, they direct AI agents in plain English
• Tools like Claude Code deploy multiple agents simultaneously: one writes, one tests, one supervises
• Tasks that once took days now take under an hour

The Strange New Workflow

• Developers spend their days describing intent to AI, reviewing the AI's "plan," then letting agents execute
• When agents misbehave, developers have resorted to scolding, pleading, ALL-CAPS commands, and emotionally charged language ("embarrassing," "national security imperative") — and it seems to work
• Prompt files have become records of hard-won rules to constrain unpredictable AI behavior

Economic Stakes

• Coding was once considered near-guaranteed, high-paying employment ($200K+)
• It may be the first expensive white-collar skill AI can fully replace — unlike AI video or legal briefs, AI-generated code that passes tests is indistinguishable in value from human-written code
• Irony noted: Silicon Valley workers, who told others to "learn to code," got automated first

Developer Sentiment: Mostly Euphoric

• Most developers interviewed were energized, not demoralized — reporting 10x to 100x productivity gains
• Key insight from tech executive Anil Dash: unlike creative fields where AI removes the soulful work and leaves drudgery, in coding AI removes the drudgery and leaves the soulful parts

Historical Context: A Long Arc of Abstraction

• Each programming era simplified the one before: Assembly → high-level languages (Python) → open-source packages → now natural language intent
• AI represents the highest abstraction layer yet: developers no longer need to manage syntax, memory, or debugging minutiae
• The open question, now being asked at Anthropic itself: what is coding, fundamentally, when the code-writing is gone?

generally considered the foundational academic work on indirect prompt injection. It's been reproduced against virtually every major agentic system since.

Website: https://sites.google.com/view/invitation-is-all-you-need

The growing integration of LLMs into applications has introduced new security risks, notably known as Promptware—maliciously engineered prompts designed to manipulate LLMs to compromise the CIA triad of these applications. While prior research warned about a potential shift in the threat landscape for LLM-powered applications, the risk posed by Promptware is frequently perceived as low. In this paper, we investigate the risk Promptware poses to users of Gemini-powered assistants (web application, mobile application, and Google Assistant).

Our analysis focuses on a new variant of Promptware called Targeted Promptware Attacks, which leverage indirect prompt injection via common user interactions such as emails, calendar invitations, and shared documents. We demonstrate 14 attack scenarios applied against Gemini-powered assistants across five identified threat classes: Short-term Context Poisoning, Permanent Memory Poisoning, Tool Misuse, Automatic Agent Invocation, and Automatic App Invocation. These attacks highlight both digital and physical consequences, including spamming, phishing, disinformation campaigns, data exfiltration, unapproved user video streaming, and control of home automation devices

Over the course of our work, we deployed multiple layered defenses, including: enhanced user confirmations for sensitive actions; robust URL handling with sanitization and Trust Level Policies; and advanced prompt injection detection using content classifiers - Google

Nobody likes the boldr in the HASS forums:

Sad news: the pairing option in the settings menu is now gone. The firmware update took it away. I just hope that doesn’t mean my klima is too old to support it or something. If that’s the case, I have a brick for their office windows :laughing:

I'm a infrared heater manufacturer that people liked. People paired them with smart plugs to turn them on and off.

Regarding the boldr kelvin infrared heater:

It replaced a small portable radiator I bought from target years ago, that thing did a fantastic job if I'm honest. It could maintain a good temperature in the room, only reason I wanted to replace it with the kelvin is because it didn't look nice in the room. The portable heater cost a tenth of the cost of the kelvin and did a better job.

Thank you. I made the mistake of purchasing one. I agree with everything that you say. So dissapointing.

Phoenix is an open-source AI observability platform designed for experimentation, evaluation, and troubleshooting. It provides:

Tracing - Trace your LLM application's runtime using OpenTelemetry-based instrumentation.
Evaluation - Leverage LLMs to benchmark your application's performance using response and retrieval evals.
Datasets - Create versioned datasets of examples for experimentation, evaluation, and fine-tuning.
Experiments - Track and evaluate changes to prompts, LLMs, and retrieval.
Playground- Optimize prompts, compare models, adjust parameters, and replay traced LLM calls.
Prompt Management- Manage and test prompt changes systematically using version control, tagging, and experimentation.

During World War II, the Anglo-Soviet invasion of Iran forced the abdication of Reza Shah and succession of Pahlavi. During his reign, the British-owned oil industry was nationalized by the prime minister Mohammad Mosaddegh, who had support from Iran's national parliament to do so; however, Mosaddegh was overthrown in the 1953 Iranian coup d'état, which was carried out by the Iranian military under the aegis of the United Kingdom and the United States. Subsequently, the Iranian government centralized power under the Shah and brought foreign oil companies back into the country's industry through the Consortium Agreement of 1954.[3]

By the 1970s, the Shah was seen as a master statesman and used his growing power to pass the 1973 Sale and Purchase Agreement. The reforms culminated in decades of sustained economic growth that would make Iran one of the fastest-growing economies among both the developed world and the developing world. During his 37-year-long rule, Iran spent billions of dollars' worth on industry, education, health, and the military. Between 1950 and 1979, real GDP per capita nearly tripled from about $2700 to about $7700 (2011 international dollars).[4] By 1977, the Shah's focus on defence spending to end foreign powers' intervention in the country had culminated in the Iranian military standing as the world's fifth-strongest armed force.

Explanations for the overthrow of Mohammad Reza include his status as a dictator put in place by a non-Muslim Western power, the United States,[321][full citation needed][322][full citation needed] whose foreign culture was seen as influencing that of Iran. Additional contributing factors included reports of oppression, brutality,[323][full citation needed][324] corruption, and extravagance.

International policies pursued by the Shah in order to increase national income by remarkable increases in the price of oil through his leading role in the Organization of the Petroleum Exporting Countries (OPEC) have been stressed as a major cause for a shift of Western interests and priorities, and for a reduction of their support for him reflected in a critical position of Western politicians and media, especially of the administration of US President Jimmy Carter regarding the question of human rights in Iran, and in strengthened economic ties between the United States of America and Saudi Arabia in the 1970s.

International scrutiny of Iran’s nuclear and missile programs intensified in 2018 after the United States withdrew from the Iran nuclear deal—known as the Joint Comprehensive Plan of Action (JCPOA)—and again in late 2024, following direct military strikes between Iran and Israel, as well as the reelection of Donald Trump. In Trump’s second term, Washington resumed talks with Tehran for the first time since pulling out of the JCPOA. However, in June 2025, after the UN nuclear watchdog declared Iran in violation of its nuclear nonproliferation agreements, the United States bombed Iran’s major nuclear facilities. Despite ongoing negotiations in early 2026, the United States and Israel launched a large-scale offensive against Iran in February with a stated aim of destroying its nuclear and missile capabilities. Although there has reportedly been some damage to one Iranian nuclear site, there is no confirmed evidence of major damage to the country’s overall nuclear facilities. Joint U.S.-Israeli strikes also killed Iranian Supreme Leader Ayatollah Ali Khamenei. Iran has retaliated by targeting Israel and U.S. military sites across the region, as well as several other Gulf countries, adding to concerns about nuclear proliferation in the Middle East.

Does Iran have a nuclear weapon? Iran does not yet have a nuclear weapon, but it has a long history of engaging in secret nuclear weapons research in violation of its international commitments. Western analysts say the country has the knowledge and infrastructure to produce a nuclear weapon in fairly short order should its leaders decide to do so. The United States, Israel, and other Middle Eastern partners regard Iran as a primary threat to their interests in the region, and view its potential acquisition of nuclear weapons as a game-changing scenario to be steadfastly prevented—by force if necessary.

Revelations in the early 2000s about the country’s secret nuclear sites and research raised alarms in world capitals about Iran’s clandestine pursuit of a nuclear weapon. Iran’s nuclear program has since been the subject of intense international debate and diplomacy, which culminated in the 2015 JCPOA. The United States unilaterally withdrew from that agreement in 2018. Since then, international monitors say that Iran has greatly expanded its nuclear activities, again heightening concerns about its “breaking out” to develop a nuclear weapon.

At least 1,245 civilians have been killed, including 194 children, by the US-Israeli war on Iran, according to the US-based Human Rights Activists in Iran group.

In Lebanon, at least 486 people have been killed by Israeli bombing, while 11 have been killed in Israel. Seven US troops have been confirmed dead and 140 injured, eight severely.

The Lebanese Red Cross condemned an Israeli strike on one of its ambulances in the Tyre district of south Lebanon on Monday night, which injured two emergency workers.

Pattern: small CLI + JSON pipes + approvals Build tiny commands that speak JSON, then chain them into a single Lobster call. (Example command names below — swap in your own.)

Lobster can run YAML/JSON workflow files with name, args, steps, env, condition, and approval fields. In OpenClaw tool calls, set pipeline to the file path.

A chrome extension that automates your browser