Inkwell - a multi-tenant long-form writing platform for the fediverse (open source, FEP-b2b8)
Fediverse
37
Posts
7
Posters
47
Views
-
Fair points to address, and I appreciate the directness.
On AI: yes, I use Claude Code as a development tool and I'm transparent about it. There's a CLAUDE.md in the repo that says so explicitly. I'm doing this solo with a day job and a passion for building, and AI tooling lets me ship faster than I could alone. I get the skepticism given the slop AI is generating. I think the distinction that matters is whether someone is engaged, iterating on real feedback, and building something they actually use and maintain. I'm trying to be that.
On FEP-b2b8: my wording was unclear, and I can see how it read that way. Inkwell publishes Article objects per FEP-b2b8 and also sends a preview Note so the content renders cleanly in Mastodon and other microblogging clients that don't handle Article objects. Saiwal read it right. I should have separated those two ideas more clearly in the post. That's on me.
On the timeline: three weeks of full-time-equivalent effort from a solo dev using modern tooling. The code's all on GitHub if anyone wants to look at it, break it, or tell me what's wrong with it. Happy to take feedback on the implementation.
I'm not trying to sell anything here. I had an idea, I'm over big tech and social media, and I'm learning and trying to be part of the fediverse community. The platform is free to use and open source. If it's not useful to this community, that's fine. But I'd rather get roasted on specifics than dismissed as another bot project.@inkwell@piefed.social Hey, I'm really sorry for all this dogpile. You shouldn't be getting this kind of flack when you took the time to implement ActivityPub.
Thanks for your work. I'm trying it out as @evan@inkwell.social and I have to say it's a really lovely UI. I'll give any feedback I can on the ActivityPub implementation.
-
@julian It's been a great multi-implementer effort, both from the long-form text producers and the microblogging services.
@inkwell@piefed.social one warning about doing ActivityPub development with LLMs: codeberg.org blocks a lot of LLMs for training and for RAG/MCP. Since FEPs are hosted there, I've found that searching for information on FEPs will result in incorrect information.
-
@inkwell@piefed.social Hey, I'm really sorry for all this dogpile. You shouldn't be getting this kind of flack when you took the time to implement ActivityPub.
Thanks for your work. I'm trying it out as @evan@inkwell.social and I have to say it's a really lovely UI. I'll give any feedback I can on the ActivityPub implementation.
Thanks, Evan! I honestly didn't take it personally and there were some learning I needed that will just help me be a better fediverse community member. And thanks for the kind words and feedback! That's mostly why I'm reaching out. The AP integration seems to be working okay, but I'm sure there are gaps I'm missing from a lack of knowledge.
-
@julian It's been a great multi-implementer effort, both from the long-form text producers and the microblogging services.
@inkwell@piefed.social one warning about doing ActivityPub development with LLMs: codeberg.org blocks a lot of LLMs for training and for RAG/MCP. Since FEPs are hosted there, I've found that searching for information on FEPs will result in incorrect information.
Thanks again Evan! I think that is where my bad information came from and good to note. I've been lucky and really appreciate the support. If there is anything I can do to improve on my side, I'm all ears!
-
Appreciate that! I realized I had an outdated readme and needed to add the sourcing directly from the Inkwell site. My biggest concern is gaps that AI won't catch, but luckily I've had early adopters getting on, requesting feature upgrades and reporting bugs. There is no way it would be where it is at if I didn't get lucky with some early users and feedback.
-
@inkwell@piefed.social Hey, I'm really sorry for all this dogpile. You shouldn't be getting this kind of flack when you took the time to implement ActivityPub.
Thanks for your work. I'm trying it out as @evan@inkwell.social and I have to say it's a really lovely UI. I'll give any feedback I can on the ActivityPub implementation.
I’m really sorry for all this dogpile. You shouldn’t be getting this kind of flack when you took the time to implement ActivityPub
You took the words right out of my mouth. I have no bones about dismissing #MOLE Training as a technology for most purposes; https://disintermedia.net.nz/invasion-of-the-mole-trainers/
I might argue the toss with someone using a Trained MOLE, outside of the narrow range of applications it's suitable for (eg digging holes in data). In fact, it's pretty damn likely. But I don't bully hobby developers for making technical choices I disagree with. Some of the comments I've seen directed at Stanton here are worse than the Mastodon HOA. Which is usually at least about some kind of substantive issue with the project (eg not respecting posting scopes when displaying replies), even if the overreaction is nuclear scale.
@rimu@piefed.social @julian@activitypub.space If you really believe you're talking to a Trained #MOLE, do you expect it to take offence and withdraw its participation if you're sufficiently mean to it? Why would it do that? If you are talking to a chatbot, you're just hurling all this toxic negativity into the fediverse for no reason. It's a special case of flooding the zone with shit.
But take a breath, touch some grass, and consider this; what if you're wrong? Which is a possibility a rational person must always stay open to. Have you considered how you would feel if somebody judged your AP implementations and your fediverse replies to be the work of a MOLE, and treated you this way? What if this is an unusually prompt and calm person (maybe they meditate regularly or something)? Or they have a disability and they're using a MOLE to help them reply promptly, but there is a thinking, feeling human being reading these replies. Honestly guys, pull your woolly head in.
Sorry to be so blunt, but I absolutely cannot abide bullying, of any kind, for any reason.
-
From 'his' github profile pic:
sigh
From ‘his’ github profile pic
Do you know how zerogpt.com detects a Trained #MOLE? By using a Trained MOLE!
we employ a comprehensive deep learning methodology, trained on extensive text collections from the internet, educational datasets, and our proprietary synthetic AI datasets produced using various language models
Which means that like all the output a Trained MOLE vomits up, there's absolutely no way to know how accurate this is (whatever the truth of whether there's a human behind these texts of not). Maybe this is a real photo of Stanton. Maybe it's been heavily altered by automated filters in GIMP. Maybe it's been auto-generated, because like many of us, Stanton likes his privacy and doesn't want to doxx himself (a privilege reversed for middle class people in the middle of the social diversity bell curve). The Trained MOLE you just weaponized against him doesn't know, and neither do you.
You know who now deserves exactly the same dogpile from you that Stanton got for using a Trained MOLE to prove a concept? I'll give you 3 guesses ; )
-
From ‘his’ github profile pic
Do you know how zerogpt.com detects a Trained #MOLE? By using a Trained MOLE!
we employ a comprehensive deep learning methodology, trained on extensive text collections from the internet, educational datasets, and our proprietary synthetic AI datasets produced using various language models
Which means that like all the output a Trained MOLE vomits up, there's absolutely no way to know how accurate this is (whatever the truth of whether there's a human behind these texts of not). Maybe this is a real photo of Stanton. Maybe it's been heavily altered by automated filters in GIMP. Maybe it's been auto-generated, because like many of us, Stanton likes his privacy and doesn't want to doxx himself (a privilege reversed for middle class people in the middle of the social diversity bell curve). The Trained MOLE you just weaponized against him doesn't know, and neither do you.
You know who now deserves exactly the same dogpile from you that Stanton got for using a Trained MOLE to prove a concept? I'll give you 3 guesses ; )
@rimu@piefed.social
At least the Inkwell software consistently fulfills the promises its interface makes. I've tried to edit both my posts in this topic with PieFed, this one to clarify that I meant; detects a Trained MOLE *output. The interface reloaded with a slightly different version of the OP and the comment I was replying to, no sign of the comment that I wanted to edit nor an edit box.Can I respectfully suggest you redirect the time and energy you're putting into dogpiling other AP implementors into fixing some bugs, and testing your interface in non-Chromium-based browsers on non-grApple OS (I'm using LibreWolf)?
-
I’m really sorry for all this dogpile. You shouldn’t be getting this kind of flack when you took the time to implement ActivityPub
You took the words right out of my mouth. I have no bones about dismissing #MOLE Training as a technology for most purposes; https://disintermedia.net.nz/invasion-of-the-mole-trainers/
I might argue the toss with someone using a Trained MOLE, outside of the narrow range of applications it's suitable for (eg digging holes in data). In fact, it's pretty damn likely. But I don't bully hobby developers for making technical choices I disagree with. Some of the comments I've seen directed at Stanton here are worse than the Mastodon HOA. Which is usually at least about some kind of substantive issue with the project (eg not respecting posting scopes when displaying replies), even if the overreaction is nuclear scale.
@rimu@piefed.social @julian@activitypub.space If you really believe you're talking to a Trained #MOLE, do you expect it to take offence and withdraw its participation if you're sufficiently mean to it? Why would it do that? If you are talking to a chatbot, you're just hurling all this toxic negativity into the fediverse for no reason. It's a special case of flooding the zone with shit.
But take a breath, touch some grass, and consider this; what if you're wrong? Which is a possibility a rational person must always stay open to. Have you considered how you would feel if somebody judged your AP implementations and your fediverse replies to be the work of a MOLE, and treated you this way? What if this is an unusually prompt and calm person (maybe they meditate regularly or something)? Or they have a disability and they're using a MOLE to help them reply promptly, but there is a thinking, feeling human being reading these replies. Honestly guys, pull your woolly head in.
Sorry to be so blunt, but I absolutely cannot abide bullying, of any kind, for any reason.
> Sorry to be so blunt, but I absolutely cannot abide bullying, of any kind, for any reason.
Once Stanton revealed his identity and stood behind his work both @rimu@piefed.social and I apologized and offered our help. That's all we're asking for, and it's not a lot to expect someone to stand behind their product.
Generative AI brings out the laziest behaviour in some people. They don't even read responses they just paste it in to the LLM and copy-paste the output. The mismatch in work is not something I abide by.
Some coward last month had their OpenClaw agent with a god complex let loose on another OSS project and when the bot got rebuffed, it wrote a hit piece on the maintainer. The bot owner was (by their own admission) hands off on the whole matter.
So, no, I am not at all going to be patient and accepting when a supposedly fully-formed piece of software is launched with zero prior reputation, no attribution, and shows telltale signs of AI usage. It trips all kinds of flags, and I will be wary of it.
> Mastodon HOA
Respectfully, this comes nowhere near the shit I've seen from that group.
Feed RSS
Gli ultimi otto messaggi ricevuti dalla Federazione
-
> Sorry to be so blunt, but I absolutely cannot abide bullying, of any kind, for any reason.
Once Stanton revealed his identity and stood behind his work both @rimu@piefed.social and I apologized and offered our help. That's all we're asking for, and it's not a lot to expect someone to stand behind their product.
Generative AI brings out the laziest behaviour in some people. They don't even read responses they just paste it in to the LLM and copy-paste the output. The mismatch in work is not something I abide by.
Some coward last month had their OpenClaw agent with a god complex let loose on another OSS project and when the bot got rebuffed, it wrote a hit piece on the maintainer. The bot owner was (by their own admission) hands off on the whole matter.
So, no, I am not at all going to be patient and accepting when a supposedly fully-formed piece of software is launched with zero prior reputation, no attribution, and shows telltale signs of AI usage. It trips all kinds of flags, and I will be wary of it.
> Mastodon HOA
Respectfully, this comes nowhere near the shit I've seen from that group.
-
@rimu@piefed.social
At least the Inkwell software consistently fulfills the promises its interface makes. I've tried to edit both my posts in this topic with PieFed, this one to clarify that I meant; detects a Trained MOLE *output. The interface reloaded with a slightly different version of the OP and the comment I was replying to, no sign of the comment that I wanted to edit nor an edit box.Can I respectfully suggest you redirect the time and energy you're putting into dogpiling other AP implementors into fixing some bugs, and testing your interface in non-Chromium-based browsers on non-grApple OS (I'm using LibreWolf)?
-
From ‘his’ github profile pic
Do you know how zerogpt.com detects a Trained #MOLE? By using a Trained MOLE!
we employ a comprehensive deep learning methodology, trained on extensive text collections from the internet, educational datasets, and our proprietary synthetic AI datasets produced using various language models
Which means that like all the output a Trained MOLE vomits up, there's absolutely no way to know how accurate this is (whatever the truth of whether there's a human behind these texts of not). Maybe this is a real photo of Stanton. Maybe it's been heavily altered by automated filters in GIMP. Maybe it's been auto-generated, because like many of us, Stanton likes his privacy and doesn't want to doxx himself (a privilege reversed for middle class people in the middle of the social diversity bell curve). The Trained MOLE you just weaponized against him doesn't know, and neither do you.
You know who now deserves exactly the same dogpile from you that Stanton got for using a Trained MOLE to prove a concept? I'll give you 3 guesses ; )
-
I’m really sorry for all this dogpile. You shouldn’t be getting this kind of flack when you took the time to implement ActivityPub
You took the words right out of my mouth. I have no bones about dismissing #MOLE Training as a technology for most purposes; https://disintermedia.net.nz/invasion-of-the-mole-trainers/
I might argue the toss with someone using a Trained MOLE, outside of the narrow range of applications it's suitable for (eg digging holes in data). In fact, it's pretty damn likely. But I don't bully hobby developers for making technical choices I disagree with. Some of the comments I've seen directed at Stanton here are worse than the Mastodon HOA. Which is usually at least about some kind of substantive issue with the project (eg not respecting posting scopes when displaying replies), even if the overreaction is nuclear scale.
@rimu@piefed.social @julian@activitypub.space If you really believe you're talking to a Trained #MOLE, do you expect it to take offence and withdraw its participation if you're sufficiently mean to it? Why would it do that? If you are talking to a chatbot, you're just hurling all this toxic negativity into the fediverse for no reason. It's a special case of flooding the zone with shit.
But take a breath, touch some grass, and consider this; what if you're wrong? Which is a possibility a rational person must always stay open to. Have you considered how you would feel if somebody judged your AP implementations and your fediverse replies to be the work of a MOLE, and treated you this way? What if this is an unusually prompt and calm person (maybe they meditate regularly or something)? Or they have a disability and they're using a MOLE to help them reply promptly, but there is a thinking, feeling human being reading these replies. Honestly guys, pull your woolly head in.
Sorry to be so blunt, but I absolutely cannot abide bullying, of any kind, for any reason.
-
Appreciate that! I realized I had an outdated readme and needed to add the sourcing directly from the Inkwell site. My biggest concern is gaps that AI won't catch, but luckily I've had early adopters getting on, requesting feature upgrades and reporting bugs. There is no way it would be where it is at if I didn't get lucky with some early users and feedback.
-
Thanks again Evan! I think that is where my bad information came from and good to note. I've been lucky and really appreciate the support. If there is anything I can do to improve on my side, I'm all ears!
-
Thanks, Evan! I honestly didn't take it personally and there were some learning I needed that will just help me be a better fediverse community member. And thanks for the kind words and feedback! That's mostly why I'm reaching out. The AP integration seems to be working okay, but I'm sure there are gaps I'm missing from a lack of knowledge.
-
@julian It's been a great multi-implementer effort, both from the long-form text producers and the microblogging services.
@inkwell@piefed.social one warning about doing ActivityPub development with LLMs: codeberg.org blocks a lot of LLMs for training and for RAG/MCP. Since FEPs are hosted there, I've found that searching for information on FEPs will result in incorrect information.
