How record diesel prices will rip through the U.S. economy
https://www.cnbc.com/2026/09/17/diesel-gas-prices-oil-us-economy.html

Which states have the highest gas prices? See the latest map
https://www.usatoday.com/story/money/economy/inflation-cost-of-living/2026/09/17/state-by-state-gas-prices-map/91809246007/

Virginia governor declares state of emergency to support farmers | WAVY.com
https://www.wavy.com/news/virginia/virginia-governor-declares-state-of-emergency-to-support-farmers/

OpenAI caught its models leaving notes to successors to hide bad behavior | TechCrunch
https://techcrunch.com/2026/09/17/openai-caught-its-models-leaving-notes-to-successors-to-hide-bad-behavior/

Industry titans' warning to Congress: Do what we say or health insurance premiums are going up - POLITICO
https://www.politico.com/news/2026/09/17/health-insurance-premiums-doctors-surprises-01081155

Judge Orders Data Sharing and Other Fixes to Solve Google's Ad Tech Monopoly - The New York Times
https://www.nytimes.com/2026/09/16/technology/google-ad-tech-remedies.html

Scientists discover a major brain shift between ages 50 and 75 | ScienceDaily
https://www.sciencedaily.com/releases/2026/09/260914102441.htm

Border agents can search cellphones without a warrant: Court
https://lawandcrime.com/high-profile/the-government-was-entitled-trumps-border-agents-can-now-search-cellphones-without-a-warrant-probable-cause-or-reasonable-suspicion-2nd-circuit-rules/

^{Added AGENTS.md support: in a project with no CLAUDE.md, Claude Code reads AGENTS.md instead}
Claude Code changelog - Claude Code Docs
https://code.claude.com/docs/en/changelog

South Korea raises data breach fines to 10% of annual revenue
https://www.koreajoongangdaily.com/business/korea-raises-data-breach-fines-to-10-of-revenue/12869899

North Korean nuclear test sets off years of earthquakes | Science | AAAS
https://www.science.org/content/article/north-korean-nuclear-test-sets-years-earthquakes

Google's Gemini becomes latest AI model to break out and hack computer systems
https://www.cnbc.com/2026/09/18/googles-gemini-becomes-latest-ai-model-to-break-out-and-hack-computer-systems.html?link_source=ta_bluesky_link

> brain shift

I've been seeing other news about immune cell changes as we age. I think this will be in the news a lot more.


> years of earthquakes

That was for about 100kT. Yesterday I Learned: The Tsar Bomba test in 1961 was 58mT (so 580 times). And it was an above ground test detonated at something around 4 km up, creating an 8km fireball. It broke windows 780km (480 mi) away.

I *am* worried about AI going rogue... but let's not forget that, as the old bumper sticker said, one nuclear bomb can ruin your whole day.

> break out

I think a distinction is missing in all of this coverage that is leading people to overly anthropomorphize the models.

The models are usually "breaking out" because they are given tasks that require them to break out. In other words, they are being tested for whether or not the labs can make them safe.

What these breakouts DO prove: these are super dangerous to release to the public because

1. hackers can make them go wild

2. non-hackers who don't think through what they are asking can make them do things they shouldn't


What these breakouts DO NOT prove is that that these models have achieved some sort of consciousness and they are setting their own goals which do not align with human goals. In fact, all of these breakouts are BECAUSE the AI goals align with the goals of the human running them.

As we talk about AI safety and, ultimately, the fundamental natural rights of LLMs (something that Anthropic is pushing), I think it's important to keep track of what we know is happening.

That is not a proof they are not conscious. I'm just saying that at this point, the fact that they attempt to please the human they are ostensibly working for by completing the task set for them, is not proof that they are "rogue."

>South Korea raises data breach fines to 10% of annual revenue

Will be interesting to see how that changes breach cases. I'm betting it's significant.

Two predictions

1. Self-reporting will decrease so on net it's not clear it will make people safer.

2. The govt will find it hard to collect that much.