You're an example of this process. You used to require a frontier model running on a data centre. Algorithms improved and, perhaps, your hardware as well, but your requirements didn't go up. Now you can do what you need to do with local hardware. You're one AI user who might not need to send money to Anthropic going forward unless you find a new frontier application that interests you.

This is why I'm a little bearish when it comes to the valuation of the AI giants. Yes, there are some extremely lucrative things on the frontier right now but, in the future, people may shift a lot of what they're doing now to local hardware while new things on the frontier may not produce the kind of profits that will pay back trillions in debt. AI companies have bet big on themselves being able to constantly and perpetually find new applications, and also that those new applications will make money.

It may only be a short-term circumstance that you need a data centre to run AI for lucrative applications. Thirty years ago you needed a mainframe to beat Kaspersky at chess. Algorithm's got better, and now you can beat the mainframe that beat Kaspersky with a bot running on a desktop Pentium of the day[1]. You could still, today, devote a full hyperscale data centre to create a chess bot to beat all other chess bots, but nobody is doing that because there's no money in it.

The frontier is continually pushed outwards. At it's outer edge you need mainframes and data-centres. Once it pushes past, you don't. There wasn't a lot of money in Chess or protein folding, despite the later having huge potential for public benefit. There is a lot of money in coding and middle management, and that's what currently has tech-bros in a frenzy. However, there's no guarantee that what the frontier moves onto next will be lucrative. It might be more chess and protein folding. If, as Algorithm's and hardware improve, coding moves to local models (Sam Altman's nightmare) and nothing lucrative immediately replaces that, what are these companies going do do with their massive data-centres?

The scale of investment we're seeing these companies commit only makes sense if they have multiple decades to reap the returns. They may well not have that much time.

_____________________

[1]https://hackaday.com/2026/11/05/pitting-a-modern-chess-engin...

However, I still get a lot of fraudulent phone calls and most people I know do too. It used to be the case that you could spot a fraudulent call instantly. Most of them came from Indian call centres from people with thick accents. Those accents can now be eliminated with AI[1]. Robocallers have also, of course, embraced AI and are rapidly becoming more and more sophisticated. I worry that, soon, some of the more vulnerable people I know will be unable to spot these for what they are.

As cheap as an Indian call centre is, AI promises to let scammers canvas a much broader slice of the population for marks and exploit them at a fraction of their current costs. It would be nice to see at least some effort from the AI giants go into helping telcos spot and block fraudulent AI callbots.

_____________

[1]https://www.techspot.com/news/106983-new-ai-voice-technology...

The Democrats did not win this election. They were just there when the Republicans lost it. I've already heard Democrat commentators say that they'll win in 2028 if they get nothing done over the next two years because Trump veto'd everything. They said this as if it was a good thing!

Don't count on it. The U.S. has started a lot of fires around the world in the last two years. Some are burning strongly enough that, if the U.S. doesn't put them out promptly, somebody else or some group of nations will. An American's first reaction might be, "Won't it be great if somebody does the U.S.'s job for a change?" Being the benevolent "Global hegemon" was the U.S.'s job. If you're not doing the job, you're no longer it. If somebody else starts doing that job, the U.S. will just be another misbehaving middle power with an overly expensive military and no soft power. Trump has shown what relying completely on hard power and neglecting soft power can accomplish.

Having the Democrats in control of Congress is a step forward, but it's just a tiny first step. Americans, don't let your leaders play at politics for two years while the fires keep burning. Congratulate your new representatives, but don't let them believe they won this election. They were just there when the Republicans lost it. Their task is to keep the job, if they can.

The house and senate both flipping is encouraging, but pay close attention to what comes next. There is no end to the self-inflicted challenges that the U.S. now faces. Many of them must be dealt with urgently. However, Democrat commentators are already saying that it will be to their advantage in 2028 if a Democrat controlled Congress accomplishes nothing for two years because Trump veto'd everything. This is not acceptable.

The U.S. has destabilized the globe. The conflicts they have both started and exacerbated need to be resolved in less than two years. Don't let your leaders, whichever side they're on, play politics until the next election. The U.S. has started a lot of fires. It needs to put them out promptly before somebody else does. Americans probably won't like it if that happens. That would cement the fall from unchallenged global superpower to just another misbehaving middle power.

Now, let's take privacy concerns off the table. OpenAI's complete lack of respect for privacy, authorship, copyright, etc. mean that you simply can't entrust private data to OpenAI. Full stop. Anyone with privacy concerns gets a free pass.

Finally, we're dealing just with people doing LLM computations on data that's not sensitive. Why aren't these people doing things in data centres? There are likely a variety of reasons, but one is likely price. If somebody does a lot of stuff with local models, they've probably found an application where a bigger closed model doesn't offer compelling advantages and the cost, purely in terms of filthy lucre, of running a local model is lower. These are the users Sam Altman can legitimately target. What Altman is arguing is that these people are paying too much.

How should he target them? Become the Steam of compute. Save them money and time. If Altman wants these people's RAM because he can use it more efficiently, he needs to offer them the same services they can get from their own hardware, cheaper, and with less trouble.

Discovery did this some of the time. There were episodes where they would show, then tell, then tell again all within a few minutes. Not every episode was done this way. I suspect they were experimenting and it did not go well. Anybody who was actually giving the show their full attention was bound to be irritated. Discovery was a low point in modern Trek for multiple reasons. They initially decided to forgo an ensemble cast and bet it all on a protagonist driven narrative. This severely limited the diversity of viewpoints in the first couple seasons. The writing was bad for entirely avoidable reasons. People who watch trek pay attention to the universe's rules. You can't beam through shields, you need to have a reason for why the away team can't be trivially beamed out of dangerous situations, etc.. The people writing Discovery did not bother to do this on a repeating and continual basis. Again, it was insulting to anyone who wasn't pecking away on their phone. I wouldn't say the writers for Discovery merely lacked life experience. They clearly hadn't bothered to watch previous Trek shows.

Strange New Worlds is better written, but a little too reliant on high concept for their most recent season. Every episode seems to have a gimmick. Still the cast is good and I look forward to them wrapping up Pike's story and, presumably, setting up a reboot of TOS.

Lower Decks is the best Trek show in recent memory. Thoroughly entertaining from start to finish. I can't recommend that one enough, just in case anyone skipped it because it was animated. Intelligent. Inventive. Hilarious. The people who made Lower Decks really put a lot of love into it.

Edit: I forgot to mention Picard, which is unforgivable. Some of the seasons were a little uneven, but some were fantastic and it was truly a treat to see Patrick Stewart as Picard one last time.

1. Anyone who has access to the drive likely also has a nice sample of your fingerprint on the same device.

2. Hardware encryption implementations are often unknown, flawed, or completely missing. You shouldn't place trust in obscurity.

3. Hardware encryption often adds an unnecessary point of failure. I've seen drives with hardware encryption implemented on a daughter board become unrecoverable because the daughter board failed, even though the drive itself was just fine.

Put a software encrypted volume on a normal storage device and call it a day. Don't pay extra to do worse.

The quality of writing for DS9 was also quite good but, after trying the TNG formula for a season or two, they broke with the writing-first mentality and adopted an action sci-fi formula. The showrunners feared they'd lose ratings if an episode didn't have at least some action. It worked for DS9, but later Trek series have had a tough time shaking free of the action sci-fi formula. The problem with formulas is that, while they produce individual episodes or shows that are exemplars of their form, the sameness eventually causes viewers to drift away.

Strange New Worlds has shown it's willing to do some pretty wild things (e.g. turning its cast into muppets, adding gore and horror elements, etc.), but they have not messed with the formula in a meaningful way.

If you're looking for something outside the Trek formula, one recommendation I'd share is "The Foundation".

“If you don’t hire the early-in-career [workers], and you just let your workforce continue the way that they were, where do fresh ideas get injected? How do you challenge the status quo?” he said."

-------------------

This would be an interesting approach for a bleeding edge startup that's betting big on producing something novel or going bankrupt. If you're going to bet the company on one guy and some AI, that's bold. He'd better have some serious chops and not be merely a slick salesman.

However, Cisco is an infrastructure company. Status quo is what most people want from an infrastructure company. You don't want your Cisco stuff to suddenly start doing new, innovative things that require a complete redux of everything built on top of it. This, of course, assumes that having one guy do a refactor of your entire code-base produces something "innovative" rather than a cavalcade of snafu's.

It's probably not the best time to be holding CSCO.

But seriously, fair use is for students, teachers, comedians, and all the people who can't pay or couldn't possibly pay enough for permission from copyright holders. A corporation training a closed model for profit can, and should, pay. If American copyright law is too onerous for megacorps to contend with, maybe it should be reformed.

4. Everything is training data. If that data isn't going to OpenAI, they're not training on it.

Also, it's important to remember OpenAI's failure might actually be the best thing for everyone else in the AI field. e.g. Some new innovation/approach could come out of left field and eat OpenAI's lunch. They've bet the farm (and a lot of other peoples' farms) on a brute force approach. Altman needs your RAM because, if he doesn't control the brute force, the brute force approach might not work only for him.

------------------------

The Germans tried graphite, but of insufficient purity[1], and concluded heavy water was the only way to go. They "didn't have enough of" that because of allied sabotage efforts[2]. There were some pretty thrilling heroics on display here that would make for a fantastic movie (or historical novel ;) ).

Even had both of these things gone Germany's way in some alternative history, they would still have faced an immense challenge in scaling up their operation under wartime constraints. The scale of the Manhattan project's enrichment efforts was truly colossal.

___________

[1]https://link.springer.com/article/10.1140/epjh/s13129-025-00...

[2]https://en.wikipedia.org/wiki/Norwegian_heavy_water_sabotage

Picard viewed the power of leadership as, foremost, a responsibility. Every action was careful, deliberate, and well advised. Some later Trek shows (e.g. Voyager) turned captains into geniuses who knew more about everything than all of their underlings. Picard was not an expert in most things and needed his staff to advise him. Once he knew the options and consequences, he shouldered the responsibility of the outcomes, but as a leader of a crew, not a genius with compatible lackeys and minions around him. It was a realistic take on leadership that simply doesn't appear often in television or movies.

Stewart, of course, was not Picard. It's too easy to conflate actors with their roles. However, actors do often help shape their characters, and Stewart's ability to give a thoughtful, diplomatic, and prudent character such incredible power and presence is likely what made Picard possible. The showrunners didn't need to strap a phaser to his thigh and drop him into every away mission to make him compelling because of what Stewart was able to accomplish as an actor sitting behind a desk.

Canada produces a lot of He-4 from natural gas and a huge proportion of the worlds tritium in its heavy water reactors. Most countries that produce tritium and He-3 tend to hoard it, but Canada is selling it cheap. If the U.S. wants to be forward looking, they should talk to Canada about a long-term tritium and helium supply contract.

Edit: Of course, I would imagine it would be in the interest of other countries to do the same while Canada is placing a premium on non-U.S. partnerships.

1. MS over-hired, has too many employees, and not enough work?

2. MS is reaping the benefits of AI and can do the same work with fewer employees?

3. MS has over-invested in data-centres and needs to conserve cash to stay in the game during it's unprofitable phase (assuming it's not a bubble)?

...

"The author drafted this essay in extended collaboration with Claude Fable 5.1."

-----------------

If you read a lot of journal articles, you're probably accustomed to occasionally running into articles a ways outside your bailiwick that are written poorly, you don't really understand, which you suspect made it past reviewers who didn't really understand, and indeed, you suspect the authors didn't really understand. Everyone involved just gave them the benefit of the doubt because they were too busy to figure out what the tortured language really meant. However, you don't know for sure, so you decide to leave it up to people in that particular field to deal with.

I can't speak for others, but I find AI writing to be frequently similar in its effect on me to bad writing. Yes, the grammar is correct, but it beats around the bush and puts emphasis on the wrong things. Even if you filter Claude through a human, those poor choices tend to shine through. This subtly inhuman voice switches your brain off. I suspect a lot of bad science is going to slip past reviewers because of LLM writing, and a lot of good science is going to slip past those who might need to read it, because the LLM co-written article switched their brains off and failed to make itself understood.

I understand the notion that, if the content is good enough, it shouldn't matter if the article is written poorly or co-written with AI. However, if the content is good enough, isn't it worth communicating it well?