Modern LLMs, just like everyone reading this, will instead reach for a calculator to perform such tasks. I can't do that in my head either, but a python script can so that's what any tool-using LLM will (and should) do.
Long multiplication is a trivial form of reasoning that is taught at elementary level. Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper. That was Opus with extended reasoning, it had all the opportunity to get it right, but didn't. There are people who can quickly multiply such numbers in their head (I am not one of them).
I tried this with Claude - it has to be explicitly instructed to not make an external tool call, and it can get the right answer if asked to show its work long-form.
Mathematics is not the only kind of reasoning, so your conclusion is false. The human brain also has compartments for different types of activities. Why shouldn't an AI be able to use tools to augment its intelligence?
Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper
LOL, talk about special pleading. Whatever it takes to reshape the argument into one you can win, I guess...
LLMs don't reason.
Let's see you do that multiplication in your head. Then, when you fail, we'll conclude you don't reason. Sound fair?
bunnie your book "Hacking the XBox" taught me how to get started on reversing electronics, took the fear out of the process, and replaced it with fun. Thanks for the multi-decades long effort you've made to make these tools available and accessible and approachable, your contributions to the hacker community are immeasurable and I cannot say thank you enough.
If code becomes essentially free (ignoring for a moment the environmental cost or the long term cost of allowing code generation to be tollboothed by AI megacorps) the value of code must lie in its track record.
The 5-day-old code in chardet has little to no value. The battle-tested years-old code that was casually flushed away to make room for it had value.
What you describe is essentially what happened, the AI result working from specs and tests was more performant than the original. The real AI you describe just rewrote chardet without looking at the source, only better.
Is there any visibility or accountability to record exactly what it did and not look at? I doubt it. So we're left with a kind of Rorschach test: some people think LLMs follow rules like law-abiding citizens, and some people distrust commercial LLMs because they understand that commercial LLMs were never designed for visibility and accountability.
There should exist a .jsonl file somewhere with exactly that information in it - might be worth Dan preserving that, it should be in a ~/.claude/projects folder.
I've never understood the concept of an app wrapper for a link aggregator (HN, reddit, etc). The whole goal is to provide links to external sources, and now I'm browsing the web in a limited web browser without all my extensions etc.
Am I missing some core concept here? Why would I want to browse the web in this app as opposed to a web browser?
Sometimes I like to save the links and comments I find particularly interesting with the "favorite" button, though lately I've debated saving them somewhere else too with a more complicated setup that could also archive both the links and the comments.
As someone who used to use native RSS readers a ton back in the day, the limited web browser usually isn't a problem for just reading a few articles.
I like native apps for things, even link aggregators, because my I want to use my OS's native window management and app management instead of just shoving everything into a browser tab, of which I already have too many. Because then it's just CMD+Tab to Chrome, and then figure out which of the 20+ tabs I'm trying to get to instead of CMD+Tab directly to that specific app.
Anyway, just a bit of old man yelling at cloud but I've always disliked the proliferation of "web app all the things." Might as well not even use a desktop OS at this point and just have a full screen browser window and call it a day.
I'm trying to understand your position here. An app with it's own way to manage multiple browser windows is better, because you have too many tabs open in your browser. If you have multiple links open, the tab management is now a problem in your desktop app instead of the browser. If you don't, then you don't have to manage tabs anyway. What does this solve that a separate browser window doesn't, except not having any way to add extensions like ad blockers or tampermonkey scripts etc?
I guess you could make a web app or app clip but I think this is a cool project. would be good to have a theme engine.
Look at NetNewsWire how good a native app of this kind can be. NNW in particular has great shortcuts, like or opening links in the native browser, and read/unread functionality
I usually don't have multiple HN articles open at a time, but I can see how that would just be replacing one problem (too many browser tabs) for a worse problem (too many, now limited, browser tabs).
It's just nice to have HN as it's own app instead of just another tab in a single app. Same reason I use mail.app vs. webmail, native music app vs the web player, etc.
PWAs also solve the problem, more or less, but it is nice to have something native.
Some people love giving up as much customization and control over their software as possible. iOS over Android. MacOS over Linux. Chrome over Firefox. App stores over installing programs yourself. Apps over websites.
There are various arguments for it (better compatibility/cohesiveness, minimalism, less debugging) but it overall seems like the opposite of the "hacker" mindset which makes how much market share MacOS has in the space very strange.
You can do this for much cheaper - all four of your tires are broadcasting a unique ID to report tire pressure, the radio to pick it up is cheap (because cars), and TPMS has no facility to randomize or otherwise secure this.
It’s actually even easier, your car has a plate on the front with a unique ID that a camera scans, often to automatically track your park time for ticketing.
I can’t really care about obscure Bluetooth tracking when every business has CCTV doing facial recognition.
I can assure you this is not the case with all Toyota models or even most. It's often integrated into the radio instead of a separate module, or simply not on a dedicated fuse, but sometimes it is. Disabling it can also lose other features of the car such as navigation, remote start, the Bluetooth mic, or the mandatory eCall feature if you're in the EU.
and similar things have happened about once a year ever since. Now in the news article I linked to a huge part of the problem was that the police didn't follow it up correctly, went to where the accident had been reported rather than where it had occurred, didn't see anything, and then gave up.
But if the car had rung from where it had actually crashed then the incident would have EISEC[1] data tagged to it, which would have given them actual co-ordinates to hit.
Even easier, electromagnetic radiation can be used to detect the presence and exact location and movements of not just automobiles, but also people! Many people have detectors for these things that can literally see through transparent material that makes up large sections of the walls of many houses and apartments.
A camera has quite a few failure modes (bad lighting, fog, dirty lens, obscured by plant growth, privacy laws, etc.) which a TPMS receiver & directional antenna don't.
The plate is pretty trivial to fake though. For one thing you can just remove it, but it's trivial to alter with just spray paint. Or using an outdated plate, or someone else's plate, etc. it's identifying sort of how an phone number is supposed to be identifying: nominal, but not secure and trivially abused for fraud
It's trivial if you're concealing your conceal your identity when committing a crime, but a huge pain in the ass and a crime itself if you just want to protect yourself from creeps tracking you.
Sure it is, but people can't realistically think to randomize their plate numbers to avoid tracking... IANAL but is it probably a criminal offense to do so.
In Europe and the US all new vehicles now have a visible ID under their front window glass, it’s called a VIN. It’s even standardized where it must be.
I'm pretty sure it should be possible if one really wants to do it. Think of a high-power IR flash and a high-res camera synchronized with the flash, with fixed focus on where the VIN would be passing. If the flash pulse is short but strong enough, it should be possible to read the VIN. Maybe some polarizing filters to remove glass reflections are needed.
Only reason I know is because I wondered if I could walk to the booth and press the button for a new parking ticket and pay for 5 minutes instead of 4 hours..
I believe that every morning someone in the tech industry wakes up and devises a new place to cram some sort of radio. And it's appealing enough the the unwashed masses such that it becomes widely adopted and then unavoidable. I don't want TPMS in my tires. It's not as if checking tire pressure is difficult. No one will consider moving away from TPMS. You'll only hear technologists say "yes, but we could improve the standard! Perhaps encrypt it." They only know how to solve technological problems with more technology.
Not all cars have active TPMS. my Volvo xc90 had them but in later models they switched back to passive ones. So it is not even a given for higher end models.
That's not as far as BLE or TPMS can work at, but it's not exactly like the NFC arrangement in a credit card, either. 5 meters is enough for a motivated attacker to do some undetected bulk data collection.
> The cannonball crashed into the church and went through a first wall. It then ended on the altar of the Chapel of the Virgin. [...] The cannon ball was walled into the left wall of the Chapel and a commemorative epigraph was added to it.
It wasn't showing the wall where it had crashed through, it's showing where the ball has been mounted into the wall for display.
Really looking forward to NY funding upgrades to the computers connected to CNC machines which tend to be pre-2000 vintage running software that's even older.
The entire concept is absurd on about 10 different levels.
If your goal is to buy the cheapest machine you can find in the world, chances are good everything you buy is going to come from China. That Prusa Mk3 you bought ages ago can be upgraded to the latest model, which means you have the option of turning that device into a lifetime machine, something ONLY Prusa offers.
Yes, the initial purchase price is higher, the lifetime price might not be.
Last time I looked, the MK3->MK4 upgrade kit is basically the same price as a complete MK4 kit (very little can be reused. New electronics, motors, extruder)
The upgrade kits are definitely a good thing, going from MK3 to MK3S to MK3.5S was a worthwhile upgrade path and has prolonged the useful life of the printer. But they have their limits.
(And with 3D printing going more mainstream, there's a large segment of the market that has no interest in building printers from kits or stripping down printer to install upgrades - even though some of us find that quite enjoyable)
reply