ilevoshko.bsky.social

@ilevoshko.bsky.social

Definitely not a bot.

Fun fact: the technology behind shahed drones got to Iran via Israel. No technology got to Iran from Ukraine. And before you argue that some UA tech might have gotten to Iran from Russia - remember, that Israel is cooperating with Russia much more than it does with Ukraine.

Militarnyi@militarnyi.com · 5d ago

🇮🇱🇺🇦Israel blocked the transfer of U.S. Iron Dome air defense batteries to Ukraine, a move that drew criticism in the United States. militarnyi.com/en/news/isra...

Mostly current "AI Coding Agents" have two modes: - Writing the code you never asked them to write; - Not writing the code you asked them to write (and detailed in the plan). An optional "let's burn monthly token amount in a day" - both at once plus debug and fixes afterwards.

Weird, I saw “OSINT” accounts posting about every Ukrainian strike on russia, but now russia is raining ballistic missiles on Kyiv and Kharkiv residential area and those accounts are dead silent. I guess I know their agenda now and will just block.

Sorry for linking to X, but it seems to be only there for the moment. Looks like Tulsi Gabbard literally released russian propaganda slides and AI map of Ukraine as official American intelligence documents. Wow. x.com/LauraLoomer/...

Laura Loomer (@LauraLoomer) on X

EXCLUSIVE: 🚨 @TulsiGabbard Recycled Published Russian Intelligence And Claimed It Was An Official American Intelligence Document In Her Latest Act Of Deception Days Before She Is Set To “Resign” As D...

x.com

What I begin to see is "LLM agent mentality" creeping into people who work with LLMs a lot. More half-baked structurally but already implemented "decisions". More "getting the checkbox ticked" instead of planning for future and maintenance. Less proper planning before implementation.

And an interesting tendency, I've caught both Cursor and Claude breaking the rules that were specifically set and reinforced in several places. All they say post-factum is "I read it at the start of this conversation and broke it anyway. No excuse for that". So if you have orchestrated flows - brace

A self-feeding loop: LLMs get to write more and more production code while the same LLMs are getting better and better at finding vulnerabilities. The code mass-produced by LLMs isn't always properly reviewed (mostly by other LLMs) so it welcomes more vulns that security-LLMs can exploit.

Journalism is almost dead. People post announcements of announcements as news and don't ask follow-up questions after being told an obvious bullshit (philosophical term, look it up).

Michelle Cyca@michellecyca.com · 3mo ago

personally I think it's a very big deal that the Canada bureau chief for the @nytimes.com — certainly one of the highest-paid journalists in the country — asked an unspecified "AI tool" what Poilievre said & published its AI-hallucinated quotes in her reporting.

Editors’ Note: May 1, 2026
This article was updated after The Times learned that a remark attributed to Pierre Poilievre, the Conservative leader, was in fact an A.I.-generated summary of his views about Canadian politics that A.I. rendered as a quotation. The reporter should have checked the accuracy of what the A.I. tool returned.  The article now accurately quotes from a speech delivered by Mr. Poilievre in April. He said, “My personal opinion is that when a member of Parliament goes back on the word they made to their constituents and switches parties, constituents should be able to petition to throw them out and have a byelection. That would put the people back in charge of our democracy rather than having dirty backroom Liberal deals by Mark Carney determine our destiny.” He did not refer to politicians who changed allegiances as turncoats in that speech.

Using LLMs for some work tasks can become a habit, and sometimes up to a point of being counterproductive. I've caught myself more than once describing a task to a coding agent only to stop mid-sentence and realise that I will do it faster than I will describe it.