> We could not read theyfell.com. It blocked us, or it needs JavaScript to show anything. Copy the words off the page and paste those instead, and we will bury those.
Every model I ask tells me my idea is great. Every founder I know is six months from changing the world and has been for four years. So I built the opposite: paste a GitHub profile, a URL, or a pitch, and an AI reads the actual thing and writes the obituary it earned. A name for the dead thing, a two-sentence eulogy that quotes your real repos and real claims back at you, a cause of death, and what it is survived by. It renders as a newspaper tombstone card you can share.
The point is not the burn. It is looking at your own work through a lens that is not impressed by you, once, and laughing, and then getting one thing to do about it. This is an experiment in whether that helps more than encouragement does. I would like to know.
Rules it plays by: it buries the work and the choices, never the person. Nothing about appearance or anything someone did not choose. Addresses, emails and phone numbers are stripped before anything is read or stored, and ID numbers are refused outright.
It is free. There is a $3.99 autopsy that goes deeper and ends with one concrete thing that could have saved it, because a burn alone is worth nothing and I wanted the paid part to be genuinely useful.
Stack: Firebase functions, Claude for the writing, resvg for the cards. One person, sole proprietorship, Ottawa Valley.
Some numbers, 40 minutes in, since people usually want these and nobody posts them.257 obituaries written by 118 people. Zero generation failures. Total model spend: one dollar. One person has bought the $3.99 autopsy, which is a 0.4% conversion and roughly what I expected, though I'd have taken more.
The thing I did not expect: people write several. The average visitor buries two or three things, which was not the behavior I designed for.
One bug worth mentioning because it is the kind AI products ship silently. If it couldn't actually fetch a page it used to roast the URL string instead, and fluently invent an entire website that didn't exist. I found it this morning on a domain that doesn't resolve at all and it produced a confident obituary for it. It now refuses and says it couldn't read the page. You can see it decline in my logs today on a site that blocks cloud IPs.
Of course, it's impossible to know for sure what was LLM processed or not, but some of your posts (like this one) have been getting classified that way.
Of course not I wrote that to update the thread about what I am seeing now as it happens. This will always be respected. Thank you for checking this out!
> We could not read theyfell.com. It blocked us, or it needs JavaScript to show anything. Copy the words off the page and paste those instead, and we will bury those.
Where's the obituary for this?
I actually have the opposite experience, often getting into protracted arguments with the tools I use.
Every model I ask tells me my idea is great. Every founder I know is six months from changing the world and has been for four years. So I built the opposite: paste a GitHub profile, a URL, or a pitch, and an AI reads the actual thing and writes the obituary it earned. A name for the dead thing, a two-sentence eulogy that quotes your real repos and real claims back at you, a cause of death, and what it is survived by. It renders as a newspaper tombstone card you can share.
I ran it on my own GitHub first. It buried my main repo for having two stars and a CI badge nobody watches but me. It was right: https://theyfell.com/obituary/bitrep-1788357179733_e0cda1
The point is not the burn. It is looking at your own work through a lens that is not impressed by you, once, and laughing, and then getting one thing to do about it. This is an experiment in whether that helps more than encouragement does. I would like to know.
Rules it plays by: it buries the work and the choices, never the person. Nothing about appearance or anything someone did not choose. Addresses, emails and phone numbers are stripped before anything is read or stored, and ID numbers are refused outright.
It is free. There is a $3.99 autopsy that goes deeper and ends with one concrete thing that could have saved it, because a burn alone is worth nothing and I wanted the paid part to be genuinely useful.
Stack: Firebase functions, Claude for the writing, resvg for the cards. One person, sole proprietorship, Ottawa Valley.
Here is what it did to Linus Torvalds' GitHub: https://theyfell.com/obituary/the-guy-who-peaked-in-1991-178...
And to a career that keeps picking the wrong startup: https://theyfell.com/obituary/the-pivot-magnet-1788356067673...
> The Guy Who Peaked in 1991
This had me.
[dead]
Some numbers, 40 minutes in, since people usually want these and nobody posts them.257 obituaries written by 118 people. Zero generation failures. Total model spend: one dollar. One person has bought the $3.99 autopsy, which is a 0.4% conversion and roughly what I expected, though I'd have taken more.
The thing I did not expect: people write several. The average visitor buries two or three things, which was not the behavior I designed for.
One bug worth mentioning because it is the kind AI products ship silently. If it couldn't actually fetch a page it used to roast the URL string instead, and fluently invent an entire website that didn't exist. I found it this morning on a domain that doesn't resolve at all and it produced a confident obituary for it. It now refuses and says it couldn't read the page. You can see it decline in my logs today on a site that blocks cloud IPs.
Can you please not post AI-generated or AI-edited comments to HN? It's not allowed here - see https://news.ycombinator.com/newsguidelines.html#generated and https://news.ycombinator.com/item?id=47340079.
Of course, it's impossible to know for sure what was LLM processed or not, but some of your posts (like this one) have been getting classified that way.
Of course not I wrote that to update the thread about what I am seeing now as it happens. This will always be respected. Thank you for checking this out!