←
Da7em11,706 posts
Da7em avatar

Da7em

@Da7_Tech

Tech Expert & Content Creator | AI & Open Source | Ambassador @cognition | DMs open for collabs. 📧 Da7em@Da7tech.com

◷ Joined October 2025▧ 440 Photos and videos
Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·3h

big respect to the OpenCode team but help me see the value why pay $10 for GLM 5.3 and ~300M tokens (maybe less) when $20 gets Claude Opus 5.5, a way stronger model with several times the usage? not a knock on them, just trying to get the logic.

F*CK IT $10 for $500 GLM 5.3 till end of the week

Da7em avatar
Da7em@Da7_Tech·1h

I'm testing GPT-6.2 Astra right now. What do you guys want to know?

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·4h

How’s GPT-6.2 been for you so far?

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·5h

For thousands of years humans have agreed on common sense, on respect, on credit. You point to the person behind the work and you give them credit. With all those options on the table you still refuse and pick the lowest one. I've never seen a post on X and decided to steal it or copy it to claim its success as mine. Not once, not in any form. If you see someone creative, try being creative too. Build your own identity. Have your own touch that sets you apart from everyone else. Stealing other people's work is not how you make it. If you have no identity of your own and steal from others, you're telling everyone you're a nobody not worth following. You have nothing to offer. You're empty inside, so you had to steal from the outside.

I've been running the ship test for months. Every time a new model drops, people are waiting for it. I worked hard on that prompt to make it precise and get the best out of every model. I paid out of my own pocket and subscribed to every plan with every release just to test it. Then just like that some people take the same prompt, try to steal the work and post it on their accounts for views without even giving me credit or sharing the video with proper attribution. Ask them to stop stealing other people's work and they block you. Never seen this level of audacity.

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·5h

I've been running the ship test for months. Every time a new model drops, people are waiting for it. I worked hard on that prompt to make it precise and get the best out of every model. I paid out of my own pocket and subscribed to every plan with every release just to test it. Then just like that some people take the same prompt, try to steal the work and post it on their accounts for views without even giving me credit or sharing the video with proper attribution. Ask them to stop stealing other people's work and they block you. Never seen this level of audacity.

Da7em avatar
Da7em@Da7_Tech·3h

golden chance to fix this I get that you want Dots to win but it's not for me and I can't rely on it. Give us a one-click switch to opt out and share that compute with Codex not unlimited, just a fair chunk of limits. Codex sub gets way more valuable it's that simple @OpenAI

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·Nov 17

Gave the same task to Grok bot and OpenAI's Dot yesterday. Grok finished it in under 2 hours. Dot still has no clue what it's doing, it's completely frozen. If all that compute had gone into the subscription instead of Dot, the subscription would be worth a lot more.

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·21h

Wrote it. Directed it. Rewatched it 10+ times! Yeah, I'm kinda proud of this one.

1:47
Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·14h

Astra 6.2 vs Fable 5.2 Astra's #1 strength is 3D. Magnificent ship details. But the lighting flicker from Astra 6 is still there. Fable and Opus crush it on lighting. Fable's sea is more real too. Ship test aside, with Fable you explain less and get more. I'll always pick it.

0:29

GPT-6.2 Astra. Magnificent!

0:29
Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·15h

GPT-6.2 Astra. Magnificent!

0:29
Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·11h

Astra 6.2 is the only model that's a work of art and deeply flawed at the same time so tomorrow i'll break down how you can bring out its best and actually have fun with it

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·11h

Time for an AMA! 😁 Drop your best questions in the replies.

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·12h

soon i'm leaving tech and AI to become a director and a writer Director Da7em yeah i like the sound of that

Wrote it. Directed it. Rewatched it 10+ times! Yeah, I'm kinda proud of this one.

1:47
Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·13h

This is the dream. A new country to call home.

Your rise will be studied my man, it's been super impressive, I hope you get to a new country asap.

Da7em avatar
⟳ Da7em retweeted
Da7em@Da7_Tech·13h

What’s weird is that it’s 3x through the API, but 6x on the subscription. That seems odd.

Honestly Ultrafast on GPT-6.2 Sol could've been an interesting choice. But burning 6x the normal usage is insane. Nothing justifies it. You take 6x my usage to give me 4.5x the speed? That's a deal where the user loses 100%.