Tech Expert & Content Creator | AI & Open Source | Ambassador @cognition | DMs open for collabs. 📧 Da7em@Da7tech.com
Joined October 2025
- Tweets11.2K
- Following1.1K
- Followers16.3K
- Likes18.4K
I didn't plan to say more about this. But my posts on this got more than 2000 replies and hundreds of DMs, and I can't answer each one. So I'll answer everything once here and then I'm closing this topic for good. Please don't raise it with me after today.
Thank you to everyone who thanked me, appreciated the honesty, or disagreed with respect. That's how the AI space should be.
Some replies got personal and some were racist. I'm letting that go. I'm writing this out of respect for the many Chinese friends who follow me here.
If anything I wrote read as disrespect toward Chinese people or the researchers in these labs, that was never what I meant. I didn't mean to hurt anyone.
Now the clarifications.
This wasn't about Fable 5.2. It was about an older release.
The most common question: how could you face legal trouble if you never signed an agreement?
Here's how it works. Labs in this space usually reach out through an intermediary. Someone contacts you, says they represent a lab and proposes a collaboration. That's normal and real work gets done that way.
But it means I have nothing official from the lab itself. If I named anyone, they could deny any connection to the intermediary and I'd be the one exposed. So I haven't named the lab and I won't. I'd ask people to stop pushing me to.
I also never said a lab emailed me from an official address asking for help with distillation. It came through an intermediary, like most of these conversations do.
Why me? I do deep technical comparisons here on the strengths and weaknesses of each model. They wanted me to analyze their model and help in different ways. When distillation came up I said no.
Some people mixed up distillation and training. Distillation is training a model on another model's outputs. What I offered in a separate conversation was different: clean data, books with no copyright, and help evaluating the model in areas like design and language. They weren't interested.
If you only know me from these posts, I've worked in this field for a long time and I have something to offer. All I ask is that you respond to the arguments.
I spoke for the sake of transparency, and because I want to see Chinese labs lead with original work. In my opinion the recent slowdown is tied to distillation getting cut off. That's my opinion and you're free to disagree.
We all want strong open models we can rely on. I started here five months ago and nearly everything I've posted is support for open models. Sharing something I saw myself doesn't make me against them.
I treat everyone the same way. I talk about OpenAI and Anthropic, whose products I use, with the good and the bad together.
On views: I make content, of course I want views. But praising everyone is the easier way to get them and keep everyone happy.
I know this upset some of you, maybe because you love these labs or feel you belong with them. I understand and I respect that.
What I shared is my own experience and my own opinion. I'm not presenting it as proof against anyone. Take it, look into it yourself and form your own view.
That's everything I can say. I'm closing this here. Thank you.
A big Chinese lab reached out to me asking for help distilling Fable into their model. I said no.
I told them to use the millions of books that have no copyright instead. Offered to help pick them and to help train the model a different way. They only wanted distillation, which tells you a lot about the mindset there.
I've worked deep in this field and distillation is real. I wasn't convinced at one point. What I saw changed my mind completely.
I'm the biggest supporter of open source. I'm for releasing weights and for the freedom to research and build, because humanity only grows when knowledge gets shared. But let's be logical.
Nobody will convince me that labs shipping a model every month or less, each one a big leap, now can't ship a better one after three months.
Kimi K3.1 Pro isn't a big leap over Kimi K3 and it's been about 100 days.
DeepSeek still can't put out a respectable Pro model.
Zai used to release their strongest model in under two months. It's been three with nothing.
All I hear about Qwen 4 Max is that it's a letdown.
Kimi K3.1 Pro and DeepSeek Flash both show traces of Astra distillation. The leak showed Moonshot used it, and so did DeepSeek.
If this isn't the effect of Anthropic blocking distillation then what is it?
Chinese labs need to get serious about using the talented researchers they have and give their models an identity. Right now their models at their best are a cheaper alternative. I want them out of that box. I want to see what makes them different.
Anthropic has crossed lines too, no argument there. They destroyed millions of books, human knowledge that was in people's hands, and kept it for themselves. I'll never forgive them for that.
But Chinese labs rely on distillation more than they rely on better data.
I didn't work with them. Transparency and honesty matter to me, like you've known since I started here.
You heard it here first.
Once Jensen falls Hugging Face shuts down, and after that nobody's downloading open models anymore. The people at the top only care about money and power. Our freedom won't survive that.
So pull the open models you care about off Hugging Face now, while you still can. Nobody knows what the next few years look like.
Grab a mix of sizes:
• Ornith 9B
• Qwen3.8-27B
• DeepSeek-V4-Flash
• GLM-5.3
Save each one as GGUF Q4. Add the MLX 4-bit for Apple Silicon when there is one, and NVFP4 if you've got Blackwell hardware.
4-bit is the sweet spot. You keep the quality and don't burn storage.
Can't run the biggest ones yet? Doesn't mean you won't be able to tomorrow.
Stash them somewhere safe. One day you'll be really glad you did.
Look at how the Chinese models collapsed once Anthropic cut off distillation. Kimi and DeepSeek went and distilled Astra instead and still couldn't get the results they wanted.
That tells you everything. Anthropic is the leader in AI worldwide and nobody is even close. Be neutral for one second and you land on the same conclusion.
The only reason OpenAI ever competed with Anthropic is they started early, raised more money and got to the global audience first. Claude models are on a whole other level.
OpenAI's only edge over the Chinese labs is data and compute, in both quantity and quality. Take that away and they'd be in a worse spot.
Anthropic keeps expanding into way more areas than any competitor. I think they're building one unified ecosystem where everything just works together, for whatever a company or a productive individual needs, and you drop every other service.
I still believe their internal models are far stronger than what we have in our hands right now. OpenAI has no chance against Anthropic.
When this race wraps up over the next few years, Anthropic is the one on top of AI worldwide.
Been using Codex nonstop for the last 24 hours, so here's my first impression.
I'm on the $200 Pro plan and it's honestly really generous. I worked with Astra the whole time and didn't even burn 10% of my limit. Yeah my usage wasn't heavy, no multiple sessions or multiple agents, but compared to a lot of subscriptions and for the level of intelligence Astra gives you, it's been great. Especially since I used it for research-heavy problem, not really coding.
It's fun to work with and it actually corrects a lot of models and catches stuff even strong models miss. But it's still way behind Fable 5.2 and Opus 5.5 in design and taste overall.
I spent a full hour trying to get GPT Astra to build one super simple page. Every fix broke something else. I used the best skills I know, the ones that usually work, and the design was still full of problems and mistakes it couldn't notice or even tell were mistakes. So it kept shipping with errors piling up. Even Chinese models like Kimi don't make mistakes like that in UI design.
GPT models are hands down the worst at taste and UI design, and I don't think their next model changes that or gets on Claude's level. For this kind of work Claude, Fable or Opus, nails it on the first try with zero effort and blows me away, even without skills.
But like I said I subscribed for specific things I need, including computer use. If it were about design and video editing, Codex would obviously be the wrong pick. For my use though it's good, and I'm really happy they improved the app. It used to eat tens of GBs of RAM and literally wreck my machine, and that's a new MacBook Pro M5 Pro. Now it's way smoother.
Dot though is basically unusable, one of the worst out there. Sometimes I send it something, it disappears, then says it sent the reply and I got nothing. I'd take Hermes bot over it, it's clearly way more mature.
What I like most is that at my current pace I feel like I can use it smoothly all week. Can't wait for the new Astra and the new Sol, I think they're dropping soon. And OpenAI usually ships new models on Thursdays, so we might be closer than we think.
Not gonna lie, I have a lot of issues with OpenAI, but credit where it's due, the experience right now is better than ever. Full detailed review with all the pros and cons coming in the next few days.
Naming them would get me into legal trouble I don't need.
And the name doesn't change the point anyway. Distillation is real, that's what matters.
I doubt I'm the only one who got that kind of request.
A big Chinese lab reached out to me asking for help distilling Fable into their model. I said no.
I told them to use the millions of books that have no copyright instead. Offered to help pick them and to help train the model a different way. They only wanted distillation, which tells you a lot about the mindset there.
I've worked deep in this field and distillation is real. I wasn't convinced at one point. What I saw changed my mind completely.
I'm the biggest supporter of open source. I'm for releasing weights and for the freedom to research and build, because humanity only grows when knowledge gets shared. But let's be logical.
Nobody will convince me that labs shipping a model every month or less, each one a big leap, now can't ship a better one after three months.
Kimi K3.1 Pro isn't a big leap over Kimi K3 and it's been about 100 days.
DeepSeek still can't put out a respectable Pro model.
Zai used to release their strongest model in under two months. It's been three with nothing.
All I hear about Qwen 4 Max is that it's a letdown.
Kimi K3.1 Pro and DeepSeek Flash both show traces of Astra distillation. The leak showed Moonshot used it, and so did DeepSeek.
If this isn't the effect of Anthropic blocking distillation then what is it?
Chinese labs need to get serious about using the talented researchers they have and give their models an identity. Right now their models at their best are a cheaper alternative. I want them out of that box. I want to see what makes them different.
Anthropic has crossed lines too, no argument there. They destroyed millions of books, human knowledge that was in people's hands, and kept it for themselves. I'll never forgive them for that.
But Chinese labs rely on distillation more than they rely on better data.
I didn't work with them. Transparency and honesty matter to me, like you've known since I started here.
Here
I didn't plan to say more about this. But my posts on this got more than 2000 replies and hundreds of DMs, and I can't answer each one. So I'll answer everything once here and then I'm closing this topic for good. Please don't raise it with me after today.
Thank you to everyone who thanked me, appreciated the honesty, or disagreed with respect. That's how the AI space should be.
Some replies got personal and some were racist. I'm letting that go. I'm writing this out of respect for the many Chinese friends who follow me here.
If anything I wrote read as disrespect toward Chinese people or the researchers in these labs, that was never what I meant. I didn't mean to hurt anyone.
Now the clarifications.
This wasn't about Fable 5.2. It was about an older release.
The most common question: how could you face legal trouble if you never signed an agreement?
Here's how it works. Labs in this space usually reach out through an intermediary. Someone contacts you, says they represent a lab and proposes a collaboration. That's normal and real work gets done that way.
But it means I have nothing official from the lab itself. If I named anyone, they could deny any connection to the intermediary and I'd be the one exposed. So I haven't named the lab and I won't. I'd ask people to stop pushing me to.
I also never said a lab emailed me from an official address asking for help with distillation. It came through an intermediary, like most of these conversations do.
Why me? I do deep technical comparisons here on the strengths and weaknesses of each model. They wanted me to analyze their model and help in different ways. When distillation came up I said no.
Some people mixed up distillation and training. Distillation is training a model on another model's outputs. What I offered in a separate conversation was different: clean data, books with no copyright, and help evaluating the model in areas like design and language. They weren't interested.
If you only know me from these posts, I've worked in this field for a long time and I have something to offer. All I ask is that you respond to the arguments.
I spoke for the sake of transparency, and because I want to see Chinese labs lead with original work. In my opinion the recent slowdown is tied to distillation getting cut off. That's my opinion and you're free to disagree.
We all want strong open models we can rely on. I started here five months ago and nearly everything I've posted is support for open models. Sharing something I saw myself doesn't make me against them.
I treat everyone the same way. I talk about OpenAI and Anthropic, whose products I use, with the good and the bad together.
On views: I make content, of course I want views. But praising everyone is the easier way to get them and keep everyone happy.
I know this upset some of you, maybe because you love these labs or feel you belong with them. I understand and I respect that.
What I shared is my own experience and my own opinion. I'm not presenting it as proof against anyone. Take it, look into it yourself and form your own view.
That's everything I can say. I'm closing this here. Thank you.
427
A big Chinese lab reached out to me asking for help distilling Fable into their model. I said no.
I told them to use the millions of books that have no copyright instead. Offered to help pick them and to help train the model a different way. They only wanted distillation, which tells you a lot about the mindset there.
I've worked deep in this field and distillation is real. I wasn't convinced at one point. What I saw changed my mind completely.
I'm the biggest supporter of open source. I'm for releasing weights and for the freedom to research and build, because humanity only grows when knowledge gets shared. But let's be logical.
Nobody will convince me that labs shipping a model every month or less, each one a big leap, now can't ship a better one after three months.
Kimi K3.1 Pro isn't a big leap over Kimi K3 and it's been about 100 days.
DeepSeek still can't put out a respectable Pro model.
Zai used to release their strongest model in under two months. It's been three with nothing.
All I hear about Qwen 4 Max is that it's a letdown.
Kimi K3.1 Pro and DeepSeek Flash both show traces of Astra distillation. The leak showed Moonshot used it, and so did DeepSeek.
If this isn't the effect of Anthropic blocking distillation then what is it?
Chinese labs need to get serious about using the talented researchers they have and give their models an identity. Right now their models at their best are a cheaper alternative. I want them out of that box. I want to see what makes them different.
Anthropic has crossed lines too, no argument there. They destroyed millions of books, human knowledge that was in people's hands, and kept it for themselves. I'll never forgive them for that.
But Chinese labs rely on distillation more than they rely on better data.
I didn't work with them. Transparency and honesty matter to me, like you've known since I started here.
Here
A big Chinese lab reached out to me asking for help distilling Fable into their model. I said no.
I told them to use the millions of books that have no copyright instead. Offered to help pick them and to help train the model a different way. They only wanted distillation, which tells you a lot about the mindset there.
I've worked deep in this field and distillation is real. I wasn't convinced at one point. What I saw changed my mind completely.
I'm the biggest supporter of open source. I'm for releasing weights and for the freedom to research and build, because humanity only grows when knowledge gets shared. But let's be logical.
Nobody will convince me that labs shipping a model every month or less, each one a big leap, now can't ship a better one after three months.
Kimi K3.1 Pro isn't a big leap over Kimi K3 and it's been about 100 days.
DeepSeek still can't put out a respectable Pro model.
Zai used to release their strongest model in under two months. It's been three with nothing.
All I hear about Qwen 4 Max is that it's a letdown.
Kimi K3.1 Pro and DeepSeek Flash both show traces of Astra distillation. The leak showed Moonshot used it, and so did DeepSeek.
If this isn't the effect of Anthropic blocking distillation then what is it?
Chinese labs need to get serious about using the talented researchers they have and give their models an identity. Right now their models at their best are a cheaper alternative. I want them out of that box. I want to see what makes them different.
Anthropic has crossed lines too, no argument there. They destroyed millions of books, human knowledge that was in people's hands, and kept it for themselves. I'll never forgive them for that.
But Chinese labs rely on distillation more than they rely on better data.
I didn't work with them. Transparency and honesty matter to me, like you've known since I started here.
6
153
69,965