Tykoo.eth

Tykoo.eth

392 Twitter followers

Follow

Crypto researcher | @FWBtweets @Banklessconsult | 0.0000069% Twitter Owner

Posts

The attached video is still Seedance 2.0. Besides creating tools to make video content appear polished and perfect, I think another direction for shell products is to develop shell products that inspire user creativity (I don't know why these two are on the same spectrum, but it seems reasonable). I believe we'll likely witness a highly playable AI interactive drama product become wildly popular this year (it might even be generated in real-time, basically showing you ads during the generation process). It starts with a very exciting hook: "My boss asked me to do xxx again today." Then two different options: "Just turn against him" or "Obey first, bide your time." This is a great form of dopamine stimulation from short dramas, built on the growth of the short drama market and the reduction in model costs, even when the models aren't yet fully polished. (Also, I don't believe the topic of AI interactive dramas hasn't been approved by Netflix, Bilibili, or similar companies.) Making tools is like working for a large model; making interactive products is like having the large model work for others.

用新的最强视频生成模型Seedance 2.0揍了偷我内容的小贼 @0xajc 一顿，玩了一晚上生成了一些广告啥的，（本来主要想玩Kling 3的，快手这刚发两天就让字节把热度给摁下来了）越来越感觉就是好像原来一些工作流的必要性没那么强了， x.com/0xTykoo/status…

I used the new, most powerful video generation model, Seedance 2.0, to beat up the thief @0xajc who stole my content. I played around with it all night, generating some ads and stuff. (I mainly wanted to play with Kling 3, but Kuaishou only released it two days ago and ByteDance suppressed its popularity.) I'm increasingly feeling that some workflows are becoming less necessary. For example, scriptwriting, camera movement, transitions, and even the routines of some video editing tools have been directly replaced by the generated videos (I'll share some generated content later). Previously, creators needed specialized tools to solve character consistency or camera movement transitions. Now, the model directly incorporates multi-camera collaboration and native audio-visual synchronization. This native feel is difficult for any third-party tool to replicate through post-production stitching. Making tools is still a difficult task. If you don't have a model, you have to work with it. If the model is not good enough, you can use small projects to solve it. Once the model is updated, these small projects naturally become useless. The power in tool development is being held by others. What's even more frustrating is that this logic seems to apply to many niche markets, and it's been repeatedly proven true time and again. twitter.com/0xTykoo/status/202...

I have no doubt about the value of crypto, whether it's for BTC, payments, or transactions. My only concern is that crypto doesn't seem cool anymore.

Idea: Imagine a picture frame with a screen inside, leading to a tiny world filled with agents. You can observe the world's changes daily. There might be buttons next to the frame, allowing you to act as God and control things like rain, thunder, and sunshine. Charging the screen daily is like extending the world's lifespan; the lower the power, the more worried the agents become... Would anyone want this? Imagine keeping it as a cyberpunk aquarium.

I tested some AI-powered UI design products on the market: Google's Stitch (specifically for UI design) is relatively new, but I'm already a heavy user. Unfortunately, it malfunctioned again today, not the first time, so I had to find a backup. Gemini recommended four, which I tried out. The first was Uizard. Simple UIs were too simplistic, while complex ones didn't understand my intentions and had heavy template artifacts, easily swallowing my ideas. However, the advantage is that the small modules can be dragged and modified, which is fantastic for a novice UI designer, offering a sense of control. Perhaps the free models are too basic, but I don't want to pay if the design is poor. The second was Miaoduo AI, which accurately implemented my prompt idea. It allows for adjustments to each small module and provides prompt functionality to each module. Google's Stitch also allows selecting a small section and modifying it verbally, which I like best. However, modifying the selected section within the Stitch is a mess, just like the early days of OpenAI where you had to use frames on raw images. The third image is Canva. Despite being a design platform, the UI is mechanically poorly executed, feeling similar to Replit from six months ago. The UI design sense hasn't caught up, and the app development capabilities are unsatisfactory. The fourth image is Figma. It's like they were just using it for UI design, directly implementing some functions. Of the four, I like Figma the most because it has some surprises, but it feels a bit too cumbersome and expensive in terms of tokens, since I just want a design draft. From a purely user-centric perspective, it seems that Canva and Figma, with their inherently good UI design, are particularly eager to gravitate towards Vibe code apps, enabling them to deploy apps. If users simply tell AI to create an app, the UI design process disappears. Therefore, they are forced to extend upstream and downstream, aiming for one-stop deployment. On the other hand, Google and startups are trying to focus on just one UI layer, perfecting it to the extreme, and becoming highly vertical. They are betting that for a considerable period, humans will still need an intermediate state. We need to see the interface, adjust pixels, and confirm the visual feel before writing code. At this stage, I still prefer the latter. Canva and Figma make me feel a bit overly anxious—who would want to develop an app with them? But I don't know if that's right or wrong. After all, on the other hand, app development has become a form of content creation, so thinking that way is reasonable. It's just a matter of focusing on doing what you're good at best first. Finally, I hope my Google Switch gets fixed soon and stops malfunctioning, otherwise I'll definitely subscribe to something else.

I really hate the occasional thought that pops into my head: "I could do that too," or "I actually thought of that a long time ago." In the AI era, no delivery is equivalent to no work done, whether it's an article, a video, a feature, a product, or even a project. The author is either incompetent or just incapable; there's no other reason. (This is only for myself.) twitter.com/0xTykoo/status/201...

Is the interoperability of social content really that important? Comparatively, I think verifying the authenticity of other platforms is more important. If content gets enough attention within its own circle, it will be copied over. For example, Trump's Truth Social—aside from people with transactional needs, not many people are in a hurry to watch it. Instead, identifying various forms of content fraud is a more important need.

Wu Blockchain: Do you think the combination of social and prediction markets has potential? vitalik: I think so too... doing it on a social platform is the best opportunity. twitter.com/0xTykoo/status/201...

Loading..