Disagree on the second sentence - that attitude is how I'm seeing production systems break all over the place these days. "I don't know, so the AI output must be correct."
Even aside from the hardware aspect of it, or the ick factor of biometrics to use the web, or possibly launching it and getting full adoption, or the fact that it absolutely could have the data spoofed, or that new attacks would be developed to MITM someone's pulse and mirror it on their bot's connection...
Where is this validation happening? Not on every transaction to every web server, surely. How could a HTTP protocol possibly actually achieve this kind of validation?
> find communities where the problem already lives and share your story genuinely.
Yes, That is the lesson that I wish more creators would learn and follow. Good marketers already know this and do it. That is why they pay attention to what channels are working, and which are not. They pay attention not just to who their audience is, but where those people pay attention. Hint: It is not always on the web.
> 17 signups. $0 revenue. No product built yet
This is where it goes off-track. You have zero signups. You have 17 leads for the future. They are not the same thing.
I'm not saying those are the greatest mice in the world, but that is a history museum. It shows old stuff, with all the warts. Dismissing it as a "piece of shit" because your wrist hurts is missing the point of why it is on display, and more an indicator that OP has a problem with their wrist than anything else.
Only in some industries, and only on modern stacks. Those of us who work on legacy platforms in enterprise environments don't need it at all. On the contrary, the younger folks who use it can't get good info out and are trashing systems when they try.
I do use basic LLM assistance, at a chatbot level. It is close enough and quick enough to give me a good head start when writing something new, and its problems are fairly quick to see and fix. But the fully baked tools are overkill for the value they offer, at least where I work.
I'd say that you need to know your environment, know what AI tools are available, and know which ones work best in your particular slice of the industry. Because if I ever go back to modern stacks, I know the AI tolls will have far more value.
So who has done the math on the price point where offshore coders are cheaper than LLMs? They both need good specs, guidance, and validation of their work. So it seems to me that when the effort to manage people vs. the effort to orchestrate LLMs is about the same, and the price is about the same... what are we actually gaining here?
Show HNs have been getting slammed/spammed with low-effort projects, often from new accounts, often more about trying to do marketing than engage with the community. Because of that, there was definitely discussion of putting some limits on who can post them. I have no direct knowledge of what change they may have put in, but from your description, it looks like they went ahead and added some limits.
Most of the prior projects I've seen that used public github data for anything along these lines suffer from the same flaw - Many coders' work is private. You can't see it to include in the system, so what you are really ranking is people's public code. And for many devs, that is their experiments, not their best work.
this is true. we could request private commit permissions but I was afraid that would scare people away from signing up. If there is demand for it we could enable it and it would certainly paint a more complete picture - with the obvious caveat that we'd need a large number of users to sign up and grant those permissions.
reply