Wing Chan
@sourceful_wing
Followers
66
Following
155
Media
29
Statuses
252
I help early-stage CPG founders ship product faster. Research @sourceful. Product packaging helps you stand out and tell your brand story.
Joined December 2024
When you find out status updates about your family situation from Google instead of by WhatsApp...
0
0
0
Oops whoever runs this account
Just a reminder that there's a constant flow of negative posts to discredit Europe and to tank our confidence. This is what an actual controller looks like made by Mistral le chat.
0
0
0
Just another day in the office. Happy Gemini 3 day.
Riverflow 2 Preview from Sourceful takes #1 in the Artificial Analysis Image Editing Leaderboard for All Listings, introducing a unified text-to-image and image editing pipeline using Sourceful's reasoning model combined with open weights diffusion models and Seedream 4.0
0
0
1
Is there some emerging law that whenever one company releases a coding model, another one immediately releases it too? This phenomenon is because basically the largest factor is training time which dominates all other factors? i.e. you can't really go faster, and you can't afford
0
0
0
Heisenberg's principle of uncertain demands.. you cannot know both the demand and the deadline
when you are doing a protest, make sure your speakerphones have good audio quality. do not do: what do we want? <unintelligible> when do we want it? now
0
0
0
Heisenberg's principle of uncertain demands.. you cannot know both the demand and the deadline
when you are doing a protest, make sure your speakerphones have good audio quality. do not do: what do we want? <unintelligible> when do we want it? now
0
0
0
What is standard to benchmark a trading competition though? Model companies are getting excited about the ups and down after 10-20 days of crypto trades, but what would be a rigorous duration? It's not clear to me if there is actually alpha given same info input flow.
u should always have strong belief in ur model. human beings are just dumb!
0
0
1
Yes. This is a good take! Also, another good proof that Deepseek continues to innovate in research. Might feel like a side quest but seems more like a bigger masterplan. Solve context, more efficient iteration loops to AGI
I quite like the new DeepSeek-OCR paper. It's a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn't matter. The more interesting part for me (esp as a computer vision at heart who is temporarily masquerading as a natural language
0
0
0
You could spend a whole evening cleaning the walls and the tables... or you could just riverflow it :) @runware
0
0
5
Here's our current view on where Image Edit models need to improve on. Please comment if you have other areas too. We are open to collaboration at @sourceful 1. Precise translations or scaling of small objects especially text. 2. Camera controls over the entire scene. 3. 3+ edit
0
0
1
Look for prediction markets that are really out-of-distribution AGI moments.
0
0
0
Mixture of experts, MCTS collapsed into either sequential chain of thought of chain of latent or even some parallel search trees, reflection driven edits, test time scaling. These are all obvious but at the same time they don't get us closer to the a more efficient lower energy
0
0
0
Moondream keeps improving its vision model. Interesting how narrow they are focusing in the stack.
0
0
0
Anyone have a gen-z version of chatgpt to give me a crash course on aura farming real quick? frfr.
0
0
0
Wow this is so cool. I love seeing the craft too.
this 17 year old homeschooled girl refuted a conjecture that was unsolved for 40 years, and which professional mathematicians worked on for years without solving she was rejected from most graduate programs she applied to, because she did not have a degree
0
0
0