codex did my taxes this year (well it actually implemented a normalization pipeline and a tax computing engine which then did the taxes, but close enough)
what I believe is that laymen will put all their tax docs into codex and tell it to 'do their taxes' and the tool will decide to implement the calculator, do the taxes and present only the final numbers. the layman won't even know there was a calculator implemented.
it was many hours of working with codex, guidance and comparing to known-good outputs from previous years, but a sufficiently smart model would be able to just do it without any steering; it'd still take hours, but my input wouldn't be necessary. a harness for getting this done probably exists today, gastown perhaps or something that the frontier labs are sitting on.
> but a sufficiently smart model would be able to just do it without any steering;
Yeah, yeah, we've heard "our models will be doing everything" for close to three years now.
> a harness for getting this done probably exists today, gastown perhaps
That got a chuckle and a facepalm out of me. I would at least consider you half-serious if you said "openclaw", at least those people pretend to be attempting to automate their lives through LLMs (with zero tangible results, and with zero results available to non-tech people).
> They don't have demand for the price it would require for inference.
citation needed. I find it hard to believe; I think there are more than enough people willing to spend $100/Mtok for frontier capabilities to dedicate a couple racks or aisles.
absolutely agree with you if we're talking about clean room reverse engineering; but in the context of finding vulnerabilities it's a completely different story
I mean-- to an LLM is there really any difference between the actual source and disassembled source? Informative names and comments probably help them too, but it's not clear that they're necessary.
Which models have you had good luck with when working with ghidra?
I analyze crash dumps for a Windows application. I haven't had much luck using Claude, OpenAI, or Google models when working with WinDbg. None of the models are very good at assembly and don't seem to be able to remember the details of different calling conventions or even how some of the registers are typically used. They are all pretty good at helping me navigate WinDbg though.
it's not that; it's awareness of inevitability of enshittification. they've released convenient tools, realized there's value to milk and are firing on all cylinders to capture 120% of it. great for IPO, not so great for customers in the long run.
Saw this exact claim on a billboard not too long ago
It's a strangely worded statement. What about data collection, metadata, other third parties
Maybe it's related to the fact that plaintiffs lawyers are now trying to verify what's going on inside Meta with WhatsApp through litigation discovery:
You can't unless you've chosen to back up your WhatsApp messages to iCloud/Google in which case it's Apple/Google responsible for preserving the messages and subject to their encryption standards, nothing to do with Meta.
Try logging in on a new device and putting your main device into aeroplane mode as soon as the login succeeds. Loading of old messages on the new device will stop.
Moxie Marlinspike (founder of Signal) [0]implemented the same E2EE algorithm as Signal (Signal Protocol) into WhatsApp, but that was 10 years ago, so who knows if things have changed since then.
Practically speaking, it isn't secure; no closed app can be. It receives regular compulsory updates (old versions refuse to work) and there's nothing at all stopping Zuck from sneaking in backdoors targeted at you personally.
Maybe, as I don't know if I got a special version that exfiltrates data to somewhere else. But this does not improve SMS security in any way. Another software also potentially being bad has no influence here.
edit is useful and there are good reasons to use it, 'never use edit' is like 'never use goto' i.e. false - but if you're just starting out, jj new/jj squash is the way to go indeed.
(my particular favorite reasons to use jj edit are git-native tools which expect to work with uncommitted files e.g. autoformatters, linters, etc. which have been scripted in CI/dev workflows such that they cannot accept a list of files as params)
reply