You just gotta start. No right time, no right condition, no right place. You just start from where you are now.
I build landing pages and brand websites, still sharpening the craft with every project. I also edit video and design in Figma β still learning, but committed to getting better every day.
This channel is where I'll document that path: the wins, the failures, the slow grind of getting better. Could take months, could take years, who knows. But the climb is the whole point β what's the point of standing on the peak of Everest if a helicopter dropped you there?
Find your αααα΅. Make things. Share them. Let's challenge fate. π
Follow along if you're on the same page.
I build landing pages and brand websites, still sharpening the craft with every project. I also edit video and design in Figma β still learning, but committed to getting better every day.
This channel is where I'll document that path: the wins, the failures, the slow grind of getting better. Could take months, could take years, who knows. But the climb is the whole point β what's the point of standing on the peak of Everest if a helicopter dropped you there?
Find your αααα΅. Make things. Share them. Let's challenge fate. π
Follow along if you're on the same page.
π₯6
Media is too big
VIEW IN TELEGRAM
After a long wait and struggle with mebrat hayl and my phone....this is the final edit......got a lot to improve.Enjoy and put your thoughts in the you know whereπ
π₯3
Forwarded from Onyx Design β‘οΈ
This media is not supported in your browser
VIEW IN TELEGRAM
I watched this kid on TikTok. First thing, heβs lucky to get his passion early.
This generation is tough; they know a lot of things. At my age, I knew α α©α©α and αα£α ααα΅ π
Another thing.. he is privileged to get accessories, and he used them well
Maybe if you donβt know, the camera in his hand is called FX3
- FX3 camera: 555,000β730,000 birr
- Camera gimbal: 130,000β150,000 birr
- The remaining two Amaran spotlights: close to 200,000 birr
So you can call it a mini production company
Some people inherit privilege
Some people build it with pain, patience, and discipline
So if you donβt have the right PC, the right gear, or the right support yet, donβt hate your journey
Just donβt quit
One day, what you are building now will become your privilege
@OnyxDesignx
This generation is tough; they know a lot of things. At my age, I knew α α©α©α and αα£α ααα΅ π
Another thing.. he is privileged to get accessories, and he used them well
Maybe if you donβt know, the camera in his hand is called FX3
- FX3 camera: 555,000β730,000 birr
- Camera gimbal: 130,000β150,000 birr
- The remaining two Amaran spotlights: close to 200,000 birr
So you can call it a mini production company
Some people inherit privilege
Some people build it with pain, patience, and discipline
So if you donβt have the right PC, the right gear, or the right support yet, donβt hate your journey
Just donβt quit
One day, what you are building now will become your privilege
@OnyxDesignx
β€6π1
Hereβs the deal with DeepSeek's new DSpark paper:
The Problem: AI Talks Like a Stutterer
Right now, when you ask an AI a question, it answers by spitting out one word at a time. Before it picks the next word, it has to stop and re-read everything it just wrote.
Imagine you had to write a 10-word text message, but the rule was: "Write one letter, then re-read the entire sentence from the start, then write the next letter." You'd take forever just to say "Hello, how are you?" That's exactly why AI feels slow and "jerky," especially when the servers are crowded.
The Fix: Hire a Speedy "Assistant"
DSpark gives the slow-but-smart main AI a super-fast, slightly dumber "assistant" to help out.
Hereβs how it works:
1. The assistant takes a wild guess: Instead of waiting for the main AI to think of the next word, the assistant quickly blurts out a whole *batch* of words (like 5-10 words) all at once.
2. The boss checks the work: The main AI takes that whole batch, glances at it in one go, keeps the words that are correct, and only rewrites the first one that's wrong. Then the cycle repeats.
Why is DSpark special?
Other companies have tried this "assistant" trick before, but they had two big problems:
- The "Slow" assistant: Wrote a good draft, but wrote it one word at a time. Too slow to help.
- The "Sloppy" assistant: Wrote everything at super-speed, but the words were total nonsense. The main AI had to erase and redo almost everything, so all that speed was wasted.
DSpark's assistant is different. It's trained to write fast AND keep the draft logical.
Plus, DSpark has a built-in "traffic cop." If the server is crowded, the assistant writes shorter drafts (so the main AI isn't overloaded). If the server is empty, it writes longer drafts. It never wastes the main AI's time on garbage guesses.
The Real-World Result
- For the user: The answer just feels way snappier. The paper says it's about 60% to 85% faster. If you used to wait 10 seconds for a reply, now you wait 3 or 4 seconds.
- For the company(Deepseek): When lots of people are using it at once, the system can handle up to 4 times more users without crashing.
The Problem: AI Talks Like a Stutterer
Right now, when you ask an AI a question, it answers by spitting out one word at a time. Before it picks the next word, it has to stop and re-read everything it just wrote.
Imagine you had to write a 10-word text message, but the rule was: "Write one letter, then re-read the entire sentence from the start, then write the next letter." You'd take forever just to say "Hello, how are you?" That's exactly why AI feels slow and "jerky," especially when the servers are crowded.
The Fix: Hire a Speedy "Assistant"
DSpark gives the slow-but-smart main AI a super-fast, slightly dumber "assistant" to help out.
Hereβs how it works:
1. The assistant takes a wild guess: Instead of waiting for the main AI to think of the next word, the assistant quickly blurts out a whole *batch* of words (like 5-10 words) all at once.
2. The boss checks the work: The main AI takes that whole batch, glances at it in one go, keeps the words that are correct, and only rewrites the first one that's wrong. Then the cycle repeats.
Why is DSpark special?
Other companies have tried this "assistant" trick before, but they had two big problems:
- The "Slow" assistant: Wrote a good draft, but wrote it one word at a time. Too slow to help.
- The "Sloppy" assistant: Wrote everything at super-speed, but the words were total nonsense. The main AI had to erase and redo almost everything, so all that speed was wasted.
DSpark's assistant is different. It's trained to write fast AND keep the draft logical.
Plus, DSpark has a built-in "traffic cop." If the server is crowded, the assistant writes shorter drafts (so the main AI isn't overloaded). If the server is empty, it writes longer drafts. It never wastes the main AI's time on garbage guesses.
The Real-World Result
- For the user: The answer just feels way snappier. The paper says it's about 60% to 85% faster. If you used to wait 10 seconds for a reply, now you wait 3 or 4 seconds.
- For the company(Deepseek): When lots of people are using it at once, the system can handle up to 4 times more users without crashing.
Breaking: someone stole a part from the transformer and we won't be having mebrat for nobody knows how long π any new offline hobbies u guys suggest?π Drop it before my phone goes offππ
π3
On second thought, I might re-read these master pieces.... they're tired of gathering dust this whole time
Forwarded from Beka (Beka)
Better Auth is joining Vercel!
Thank you everyone who has been part of this journey in so many different ways π
More to come! Excited for whatβs ahead β€οΈ
Thank you everyone who has been part of this journey in so many different ways π
More to come! Excited for whatβs ahead β€οΈ