metalspot

5 ms·

1.
by metalspot 23d ago
LLMs are just one small part of computing. it is fine to learn how they work but they are nothing magical. if you want to build a real working system using LLMs that actually makes money you still have to learn everything else. but i think
2.
by metalspot 2mo ago
llms are 100X for me, but only 10X goes into actual shipping code. the rest is all on doing the dev process the way it actually should be done: incremental POCs, refining specifications, modularization, exhaustive test suites with 95%+ logi
3.
by metalspot 2mo ago
I have used AI almost exclusively for a year, but I was programming by hand for almost 30 years before that. I don't think there is any difference between going from assembly -> C -> Java, TypeScript, C#, Go, etc, and taking the
4.
by metalspot 2mo ago
The problem is that the cost curve for completeness/correctness goes asymptotic at 90-99% so the cost up building complete/correct software is never worth it from a revenue perspective, and only happens if there is significant lia
5.
by metalspot 2mo ago
In the early 90s I had an elementary teacher who was an amiga nerd and had a couple for us to play with. Best class ever.
6.
by metalspot 2mo ago
The concepts of industrial reserve capacity and using dual-use consumer goods to subsidize military production capacity are well known and widely practiced historically in the US. China adopted this strategy from the US, and the US convenie
7.
by metalspot 2mo ago
Interesting. I am not familiar with model internals at this level because I have only been working at the application layer so far, but will definitely research this further. When you get the paper published would appreciate if you can drop
8.
by metalspot 2mo ago
This also comes with significant capability reduction. deepseek-v4-flash is very good in the < 250K range, then degrades between 250-500K, and is practically unusable after 500K. [edit] This is my observation from using it without an spe
9.
by metalspot 2mo ago
Deepseek is still charging cached input at 1/10th the price of any competitor. For an example of my real token usage for a day with DS: Input (Cache hit) 530,949,760, Input (Cache miss) 7,875,004, Output 1,389,685 - it is still 1/
10.
by metalspot 2mo ago
You really have to look at energy/capita and how much energy is embedded in exports. The gross numbers are misleading. The US wasn't building new electrical generation capacity because it didn't need it and there was no marke
11.
by metalspot 2mo ago
I don't disagree but something less bad is better than something more bad if those are the only two options. I only said Duolingo was better than Youtube. Duolingo is nuts. I get my children's spam from them and it is this crazy e
12.
by metalspot 2mo ago
"broad anecdotal consensus" very scientific. i don't disagree, but is it better than youtube? tiktok? instagram? (cocaine? meth? fentanyl?)
13.
by metalspot 2mo ago
Can you point to any serious studies that show the opposite? No. They don't exist. There cannot be any serious studies until there is a generational cohort that can be evaluated for long term impact. But on a personal level I would vas
14.
by metalspot 2mo ago
literacy causes profound brain structure changes and psychological effects, and mass literacy results in profound social transformation. which is why people have be trying to prevent the spread of literacy for thousands of years, to this ve
15.
by metalspot 2mo ago
code review is a dead end. the only way building with AI makes sense is in secure systems that can run unaudited code. code review has always been a liability fig leaf. it is much harder to understand a system from reading the code than wri
16.
by metalspot 2mo ago
Moving module by module is definitely the safer way to do it, but at the module level you do the same thing. You have to have it feature exact so that you can replay test against the same interface and do progressive roll out. The differenc
17.
by metalspot 2mo ago
Rewriting while changing features is the worst idea ever, which almost always leads to failure, and is the reason why "don't do rewrites" is a widespread rule. Whether it is humans or AI, the correct way to do it is always a
18.
by metalspot 2mo ago
An LLM can do this easily if you have a good test harness
19.
by metalspot 2mo ago
How would you know? Are you an expert in the language and culture that its texts were originally written in? Even an expert would qualify their interpretation by saying that it is only an educated guess. Look at how easy it is for humans to
20.
by metalspot 2mo ago
Every single SaaS app will be completely rewritten in the next couple years. Fair warning. If you aren't on that train you won't have a job.
21.
by metalspot 2mo ago
Exactly. AI opens up a massive development frontier of projects that were simply impossible before. It does differentiate though. In the old world most "software engineering" work had nothing to do with software engineering so bei
22.
by metalspot 2mo ago
this is a modernist misconception. ancient religions are philosophical frameworks and the texts are internally coherent arguments for those frameworks. They are just expressed in language that modern audiences without formal training don&#x
23.
by metalspot 3mo ago
priming obviously works. it is inherent in the physical structure of the brain. nerve endings from various all over the body terminate in different physical locations in the brain, and then these are interlinked in an adaptive biological me
24.
by metalspot 3mo ago
No. Google and Apple are already rolling out the infrastructure. FCC is requiring IDs for phones. It will be opt-in for a while but the acceleration of AI risks will inevitably lead to an infrastructure level lock down of the entire interne
25.
by metalspot 3mo ago
The entire internet will be real id verified soon. You won't even be able to send packets without a hardware digital signature linked to a real id. Might as well get used to it.
26.
by metalspot 3mo ago
AI and remote interviewing have definitely exacerbated the problems that existed before though. Leetcode was always weak, but now that it is easy to cheat on it is a negative selector, because the cheaters do best. Leetcode was originally s
27.
by metalspot 3mo ago
The chart at the top of the article makes it clear that the entire thing is pure fantasy
28.
by metalspot 3mo ago
hallucination is good for tasks that have an external oracle like computer programming
29.
by metalspot 3mo ago
The US is primarily powered by oil and natural gas, has massive domestic capacity, and locked in supply all over the Americas. China has a completely different energy mix and they move from a position of competitive disadvantage on ICE cars
30.
by metalspot 3mo ago
> But also a financial one china's advantage is manufacturing capacity. giving away the model is a loss leader for the hardware business. they have a big gap on chip manufacturing. the only way to close that is by developing more ef

More ›