It's not "someone" it's The International Group for Historic Aircraft Recovery (TIGHAR), again. It's always TIGHAR, which interprets every single rock or bit of rubbish ever found on Nikumaroro (called Gardner Island at the time of Earhart's disappearance), while ignoring that the island has been intermittently inhabited for well over a century. TIGHAR has been "finding Earhart" since 1993.

From Admiral Rickover's 1953 'Paper Reactor' memo, "An academic reactor or reactor plant almost always has the following basic characteristics: 1. It is simple. 2 It is cheap. 3.It is light. 4. It can be built very quickly. 5. It is very flexible in purpose (“omnibus reactor”). 6. Very little development is required. It will use mostly “off-the-shelf” components. 7. The reactor is in the study phase. It is not being built now." Time will, of course, prove the final test as to whether these "microreactor experiments" to produce a "compact, transportable microreactor" will successfully " fast-track the path from lab bench to commercial rollout" and eventually " support applications ranging from remote communities to mining operations and data centers" and "replace disel generators". But I wouldn't bet on it.

I read a great deal, particularly speculative fiction. I've tried repeatedly to use various AI tools to track down a short story or book where I can remember details of the story, and perhaps the rough publishing date, but not the title, author, or publisher.

AI appears to be utterly useless for this. It will either come up empty and make vague suggestions, like "Look at fantasy recommendations for the date range you've provided". Or worse, it will focus on the wrong book (say, one published in the 2020s, when I explicitly stated that the book was published in the 1990s) and keep insisting that's the closest match. Even character names or other proper nouns don't seem to work. Telling it to exclude certain titles or authors fails reliably.

When asking similar questions on stories I do remember the details of, getting it to "find" something, it needs to be provided with enough unique information that the story in question could readily be looked up. While it seems like the sort of thing AI ought to be useful for, the only stories it seems to be able to reliably find are very popular ones, even when searching for 21st century works.

I was under the impression that most LLMs are mechanical turks - not in an immediate sense, but in the sense that a lot of workers around the world were involved in annotating the datasets LLMs are trained on.[1] So, from that perspective, what an LLM is doing is outsourcing (in time and space) your conversation to a random person using the internet. Insofar as there's thinking involved, what you're getting are reflections of the thinking involved in building the Chinese Room in the first place.

1. https://www.economist.com/inte...

Cassini ended its mission in Saturn. If you've never watched the Grand Finale video, it's worth it:
https://www.youtube.com/watch?...

Yep. Much like his endless Full Self-Driving promises to Tesla shareholders, Musk's promises of "Mars Real Soon" are vaporware, even moreso than the lunar lander services he's been paid for. This will be more wealth transfer, done at the expense of far more productive research and science.

Presenting LLMs as AI (or even AGI) feels like part of the same hype-train we've been on for more years than I can remember: blockchain > cryptocurrencies > NFTs > LLMs. They all share marketing with a sales pitch of "I've got a lot of compute and a clever plan for somebody else to pay me a lot for it!"

...and I don't just mean "rocket go boom".

SpaceX seems to largely have Booster working. Yes, they lost this one, intentionally pushing some limits on the reentry.

With Ship, Flights 4, 5, and 6 all managed controlled splashdowns. With Flight 7, SpaceX moved to Block2 Ship. Flight 7 was lost due to an internal fuel leak, Flight 8 due to Ship engine failure, and now Flight 9 has been lost with an apparent internal fuel leak, again. For whatever reason, Block 2 doesn't appear to be up to the challenge.

Year by year "right to a profitable business model" (as long as you have a big enough pile of wealth to begin with) marches onward.

When your businesses' "one neat trick" is *ignore the law* (because money), your business should not be permitted to exist.

So, basically the same as their old model:

1. Buy established business
2. Loot it.
3. Sell at profit.
4. Who cares what happens to the business long term.

We were more civilized when we hanged bandits.

Historically corporations have been permitted to exist because they can be an effective way of getting objectively desirable things done. Things like shareholder value and market capitalization are at best orthogonal to whatever public goods they may produce. Look at the top ten US companies by market cap: Apple, Microsoft, Nvidia, Amazon, Google, Facebook, Berkshire Hathaway, Tesla, UnitedHealth Group, and Johnson & Johnson. Most of them are widely hated. The assassination of United Health's CEO was celebrated more than mourned. Of them all, Berkshire Hathaway has the least negatives... and it is largely the opposite of what's described in the article. Innovation may lead to 800 lb corporate gorillas, but that doesn't make having 800 lb gorillas desireable.

Back in 2005 Boeing spun off a lot of its parts production to a newly created company, Mid-Western Aircraft System (later renamed Spirit) to an investment management firm. The idea was clearly (although I don't know that I've ever seen a company spokecritter admit it) to have Spirit, under investment firm management, cut every corner possible to squeeze out more profits. Fast forward two decades, and Boeing is buying Spirit, because cutting corners didn't actually work. Looking at AT&T's announcement (https://about.att.com/story/2025/lumen-mass-markets-fiber-business.html), you can see the same plan taking form. "After closing the transaction with Lumen, the Company plans to sell partial ownership of NetworkCo to an equity partner that will co-invest in the ongoing business. AT&T expects to identify an equity partner and close a transaction within approximately 6-12 months of closing the transaction with Lumen. Upon closing a transaction with an equity partner, the Company expects NetworkCo will be deconsolidated from AT&T’s financial statements and operate as a wholesale commercial open access platform, providing fiber access services to AT&T as the anchor tenant. All acquired Lumen Mass Markets fiber customers will remain AT&T customers." I.e. "NetworkCo" will try to cut every corner and squeeze every penny, but it won't be "AT&T" doing it.

"Altman said the goal is to release a device by late next year." Eighteen months is always the perfect timeframe for BS and hype. It's close enough that people who want to believe, or are easily persuadable will tell themselves "that's real soon!" and act (and buy) accordingly. But it is simultaneously far enough out that there is plenty of time for "no one could have seen this coming!" delays or changes in the overall economic climate. Alternatively, it is also far enough that if Altman just stops talking about it in a few months, in 18 months hardly anyone will even remember, and even fewer will care.