New Comments | Hacker News

Looks like matrix is in this niche now.
https://www.cnbc.com/make-it/2026/03/06/e-shaped-economy-rep...
https://www.axios.com/2026/01/22/jobs-stock-market-rich-amer...
If it is not written down, the agents can’t know it, so the rule is to write everything down. File an issue before starting work, update with a note at every step. One task per issue. You know… the stuff we would do for each other, if we all didn’t hate Jira and company so very very much.
It's easier to setup and maintain schedules, routines, or more complex functionality.
It's certainly not critical functionality, but it's nice. Same with my laundry machine and dryer sending a push notification when the cycle ends. Unnecessary, but nice.
Yes, and it's mentioned in TFA.
> the slowest phone you could spend money
https://puri.sm/posts/the-danger-of-focusing-on-specs/
> ANCIENT Snapdragon 4s Gen 2 you can get in a $20 flip phone from Walmart Right Now has performance cores
And you will not control your main computing device, https://news.ycombinator.com/item?id=47367568
Just next to this word, there is "Lean verification", which is not about arithmetic but about the logic.
But the problem is not even there. LLM are not able to do math not because they fail to do 12,483,324,281/188,553.37463286 * (243+2-983), but they are not able to do math because they don't _understand_ 12,483,324,281/188,553.37463286 * (243+2-983).
I cannot do 12,483,324,281/188,553.37463286 * (243+2-983) myself just like that, yet I can do advanced math. Because I understand this expression. I know what it means. When the LLM was not able to produce the correct number, it was not because it knew what this expression meant and was "distracted" and forgot to report a 1 when doing the mental calculation. It was because the LLM was seeing the expression and was trying to guess the answer based on the answers of other texts that had similar expressions. It did not _understand_ the expression itself: it had no idea what it meant, it was just guessing the answer by copying the output it has seen in the wild.
And this is why it was a good element to demonstrate that LLM cannot do math: if they don't understand the expression 12,483,324,281/188,553.37463286 * (243+2-983), then they don't understand the expression f(t)=\sum_{k=-\infty}^{\infty} c_k e^{j k\omega_0 t}, \qquad \omega_0=\frac{2\pi}{T}.
And that is why the AI framework to do proper math and help mathematicians needs harness with tools like Lean. The LLM part does not understand what the expression means, it needs a tool to discard, internally, all the LLM outputs that are just non-sense.
This is just one problem I have with his claim, there are many others.
They have some duty to the developers who paid them for distribution rights whether they're making AI slop, manual slop or something worthwhile, and their processes need to scale appropriately if only to make that judgement and reject the apps they don't want in a more timely fashion.
If we linearly interpret the last year of progress from GPT-5 to Opus 5.5, times 5x, assuming no plateau, could such an intelligence realistically achieve a task like that?
Maybe also not, perhaps some form of RSI would need to deliver accelerating progress.
There are a lot of assumptions and speculation here. Still, at the end of the day, it does not seem that unlikely that they succeed in creating AI that can reason over so many parameters as to remove all intellectual bottlenecks from the process, leaving mainly the physical limitations.
And what is the limit there to building a good model of the human organism? I won't pretend that I know, but simply pointing to clinical trials, as some here do, instead of making an argument based on how such a digital model cannot be created seems unserious.
It's okay for me to declare a macro with `__attribute__((warn_unused_result))` for GCC and Clang, and `[[nodiscard]]` when the C code is later compiled with C++ (or nothing).
He might also feel he needed to do it for many reasons, some more self-serving than others.