It was a really short shower.

I’ll just point out that 3% market share is still bigger than the entire market when started building PCs. And that’s assuming they can make this attractive to anyone. Single point of failure for your entire company? Single supplier who has you over a barrel when they want to raise prices? Who in their right mind would go for that.

We’ll see. The fact that it’s on offer doesn’t mean people will bite. I’ve seen the industry try so much stupid shit that people said no to. Free computer full of ads? No. Scan cat? No. Packing LEDs into things that don’t need to light up or be hotter? Well… they got us there.

I’m glad Anthropic is holding a line here but stand up against might be a little strong. They just don’t want to recreate ED209.

Goodbye local Windows, you mean. Except I said goodbye two years ago and never looked back or missed it. Windows does nothing I need, and does it poorly.

Don’t get me wrong, I’m still petty enough to hope this effort is a miserable failure, but ultimately I don’t care all that much.

I’m middle of the road on AI. I think it has uses. I also think this technology is a dead end (i.e. this is not going to lead to AGI) and had people understood from the start the limitations of it, investment would’ve been more modest and cautious. Is a great technology. You can do cool things with it. But it will never be able to significantly replace humans. However it may be really painful watching the investor class wrestle with that reality.

I think the chip does have uses and I think building it even with today’s models would last a long time. But the number of scenarios where it is unequivocally better than nothing is smaller than AI bros (I draw a line between an enthusiast like myself and a bro who is all in and won’t hear reason) want to think.

Last point. In theory this chip is great. Based on my reading this is a substitute for an H100 — a data center GPU (APU?). This isn’t going into smart mines or drones and probably not cars. Not without more development. So while there is potential here, none of these use cases are practical. This is a way for OAI or whomever to run their current models just the way they are for cheaper but with a hardware cost to upgrade. This isn’t going to matter for the rest of us for a while.

My scenario was a safety device that prevented cars from hitting pedestrians. You’re stuck on this autonomous self control in the event of loss of human control and it seems like you’re interpreting what I’m saying in that context, which I wasn’t. I presented a scenario when it’s a good idea and one when it isn’t. Nothing to do with your autonomous control scenario.

But let’s see. If you’ve got a done that can fly itself for a few seconds or minutes if it loses signal, simply loitering waiting for control to continue, or maybe continuing on a flight path until it is out of jamming range. Alternative is uncontrolled crash, possibility of avoiding that is nothing but upside, whether it’s 10% or 90% success. It’s a good example of the type of scenario I was describing with the smart mine.

I wasn’t trying to address your scenario because it already falls into the niche I was describing. I was trying to demonstrate how to consider scenarios where AI is good vs ones where it has an unacceptable tradeoff. Where the consequences of failure don’t outweigh the benefits when it gets it right.

So I think we were talking past each other, and if my communication was unclear then I apologize. In my defense, it’s 2AM here.

How does it decide the car is uncontrolled? That’s a failure scenario, too.

I’m not even sure what you’re arguing. I said from the get go that there are niche cases where AI is nothing but positive. You seem to be arguing that there are a bunch more cases. Fine. Maybe the niche is slightly less thin and narrow than I think. Cool.

No, it doesn’t work in this context because false positive is worse than nothing. False negative is better than nothing. Zero sum. Obviously it depends where you set the threshold of false positive and false negative. I imagined a very simple scenario the first time.

If even only .001% of the time, you’re going to cause a shit load accidents. You’re going to average a car slamming on the breaks for no reason like every… 2 minutes would be .12, 20 would be 1.2, 200 would be 12% 800 would be 48%, so you’re going to have every car slam on their breaks every 12-15 hours of drive time. That would be an absolute mess.

Sure. It excels in cases where 60-90% success rate is better than nothing. If you have a smart mine that doesn’t detonate on civilians, 50% success is better than 0. It reduces civilian casualties by 50%, which is still awful, but if you’re going to plant mines it’s better than entirely indiscriminate. Use cases definitely exist. A false positive means it doesn’t detonate on one soldier but might on the next — still an effective deterrent. A false negative means it blows up a kid, which a dumb mine would also do anyway.

It’s just generally not in situations most people are generally thinking about. You have to imagine cases where there is some upside and no downside. It doesn’t work in a context of say, auto-breaking a car if a pedestrian is detected because a false positive is going to cause accidents and probably kill people even if in other circumstances it does save lives.

The thing that differentiates ChatGPT and Claude is likely more the RAG pipeline that backs them and feeds them context. The models really aren’t getting better, we’re just getting better at using them to break tasks down into units so small AI can figure it out. I’d bet a GPT 5 model or a Claude Opus 4.6 model would last 5, maybe 10 years before you really start to notice its capabilities are falling behind. I’ll bet you could use GPT 4o for 5-10 years and it would be fine.

A lot of the models we have are about as good as they are going to get. I mean, ChatGPT 5 isn’t appreciably better than ChatGPT 4. Hook one of those models or even one not as strong to a purpose-built RAG pipeline and a controller to run as mesh of interconnected prompts and agents, and you’ll blow away general purpose chatbots in niche areas in terms of cost, efficiency, and performance.

The question then becomes, to what purpose can you put this super fast, dedicated machine that performs certain small-scopes, simple tasks really well, but also fucks up often enough that you can’t depend on it. To what tasks could you set a bot that does stuff with minimal competence let’s say 90% of the time, and the other 10%, doesn’t create even bigger problems?

That domain exists, but it’s thin and narrow.

shit no one thought of

This is one thing it isn’t going to do.

Let’s say AI increases productivity by 10%. So you’d reduce staff by one person out of every ten. But how many teams actually have ten people on them? My biggest development teams might’ve had 10 between PM, devs, and QA. But again cut one of your five devs and you reduce capacity, not increase it.

I’ve never worked anywhere that a 10% increase in productivity could justify cutting a person. I’m sure those places are out there, but it seems uncommon.

It might let them cut staff off they can overwork their people that much more, but a lot of people are stretched to capacity even now.

I agree that public transport can be less stressful (not always, making a connection between the perpetually 20 minute late bus and the train was stressful for me even if I had no control), but just as food for thought for those of us currently with commutes that aren’t public transport friendly — over the years I’ve found that driving like an absolute maniac only saved me a minute or two even over fairly long commutes, compared to driving calmly and courteously.

Leaving just slightly earlier led to much less stressful drives for me. I’m not talking 10-15 minutes, but even 5 minutes earlier would let me drive calmly, safely, and get to work faster than being an asshole with 5 minutes of extra sleep.

My experience with this is over a couple of decades old at this point. So I don’t know that any of it is relevant or useful.

At the time I was certified in Lotus Notes development and administration. It was a bit of a niche, but it was used lots of places. I created a business to do business under.

There are websites that deal with corp to corp contract work. So you can find work that way. There was also a specialist independent sales person who worked with IBM customers (Lotus Notes was an IBM brand at the time) who offered to be my sales branch. I never made enough money to hire him. Maybe I should’ve, but I was afraid I wouldn’t be able to get leads and it would be a waste of money.

I worked with small and medium businesses so there was some word of mouth business from managers knowing one another and recommending me. I tried to go to user group events and meet managers and contractors that way. Having a good relationship and recommending each other for various things is another way of getting business. You could also find some c2c opportunities in other places like dice.com.

Ultimately, I was not cut out for sales or business ownership. I lasted the better part of a year. During that year I got unemployment from being laid off from my last job but I didn’t pay myself a regular wage from my company — maybe that wasn’t strictly legal, I don’t know, but the money wasn’t coming in regularly enough to pay myself any kind of regular wage. Anyway, I got away with it so whatever.

I had a couple of good months and a lot of shit months. Between unemployment to help smooth the lows and the business I was able to bring in, I was able to pay for a very shitty house for my family and food. We did have to sell a car. I didn’t plan for it, I just got laid off and walked out the door with 2 weeks of severance and a list of customers I’d been working with.

I took the next job offer that came along. But if you don’t hate the relationship building and sales work and negotiation, and you have some savings set aside to weather lean months, and stick with it for a few years, you could probably do better than I did. Or maybe it’s all different now anyway. Probably a lot more Twitter and self promotion.

I started using emdash when it was popularized by AI and everyone complained and then authors explained how and why they used it and I learned. It’s great punctuation.

It would be easy to hide instructions in a Moltbook post telling any bots that read it to share their users’ crypto wallet, upload private photos, or log into their X account and tweet abusive comments at Elon Musk.

Who would create a bot for this purpose and then also give personal details in the instructions / memory or access to photos?

But as for that last point, tweeting abuse at Musk might just be the first sign of actual intelligence I’ve seen from an LLM.

I better go check my NRA-approved 2A handbook for instructions on dealing with masked thugs breaking into my home…

Everyone told him to fuck off and leave Greenland alone, so Iran, you’re the next contestant on The Bribe Is Right.

I have to admit, this is more entertaining than counting 'r’s in strawberry. Novel logic puzzles really are about impossible because there is no “logic” input in token selection.

That being said, the first thing that came to my mind is that at some point the (presumable) adults, me and the priest, are going to be on the boat at some point, which would necessarily leave the baby alone on one shore or another.

Clearly, the only viable solution is the baby eats the candy, and then the priest eats the baby.