Immanuel Can wrote: ↑Sat Apr 12, 2025 4:16 am
As for ChatGPT, aren't you tired of trusting a computer program programmed by somebody else to do your thinking for you?
Aren't you tired of using a calculator to do the additions and long divisions for you?
Of course, you still have to sanity check what you see on your calculator screen. For example:
4356.622 times 975.3 is how much?
That's around: 4000 times 1000.
So, the answer shouldn't stray too much from around 4 million.
Apparently, according to google search, it is 4249013.4366 which is an answer that I accept.
Concerning the idea of trusting another programmer, I strive to use reproducible-build compliant free and open-source software:
https://reproducible-builds.org
Reproducible builds are a set of software development practices that create an independently-verifiable path from source to binary code.
It means that for every program I use, I can precisely pinpoint the source code from which it was built. Consequently, a large number of people can independently verify the source code and raise issues. Of course, the more obscure the app, the less likely that this will happen.
This principle is non-negotiable when it is about a lot of money. I would not use a wallet that is not reproducible-build compliant:
https://walletscrutiny.com
Know your wallet like you built it.
Not everyone is a developer. Not everyone has to be.
WalletScrutiny helps everyday bitcoin users verify whether or not their wallet is truly open-source and secure. Enter your wallet's name to get started.
Thousands of wallets reviewed, and counting.
Our wallet analysis process requires running 17 separate tests across various categories. This might sound excessive, but our thinking is that every compromised wallet, whether it’s already popular or just released, has the potential to damage bitcoin’s reputation or discourage future users.
Learn more about our methodology.
The most important piece of software that needs to be reproducible-build compliant is the operating system. I use Debian Linux, which is strongly committed to reproducible-build compliance. Furthermore, new versions are quarantined for two years in "testing" before being released to "stable" in order to give security researchers enough time to scrutinize the source code.
ChatGPT, DeepSeek, and Gemini are a bit new to the game. I currently use these things more as a matter of testing than anything else. It is obvious that sooner or later the training dataset of these programs will also have to be scrutinized. It may require some adjustment to the existing methodology. We may indeed have some catching up to do.
But then again, it is absolutely not desirable to stop innovation over regulatory compliance issues. Regulations always come later.