Nick Shilov

you can shitpost or show progress, low views result is around the corner

really enjoyed doing my first indie game in @telegram the main hack is user don’t need to download it

firm size deviates, the fear to keep up with big guys and not to disappear

quite strange, Reddit is a pure diamond of good data for LLM training

me saying the design is good

Start to unslop the design a lil

a massive shift from client side to server side. your future product will be compute, data and delivery

please improve the algorithm @petersellis i still see some political content even if i snoozed off for 24h thank you in advance

let’s call it a tiny progress

Wan 3.0 $0.35 vs Kling 4.0 $1.10 obviously the quality is almost identical

King 4.0 for one line simple prompt, fits perfectly

bruh it’s real, gonna be drama king

me with all vibe coded projects

despite all the fight against distillation, chinese labs keep shipping great models. for anyone who still needed it spelled out, that means they actually know how to do RL and training, not just "stole everything from anthropic" the latest kimi 3.1 is on par on average benchmarks with fable 5 and astra 6, which came out just over a month ago. the gap between open and proprietary models is even shrinking a bit. in video, voice and images, open models are often in first place any alternative and any decentralization is good for the global economy, because it pushes back at least a little on resources and power concentrating in the top five labs. but it's not a given this trend holds as we move toward RSI don't believe in decentralized ownership of compute. the hundreds of billions going into datacenters come from capital markets, and those run on creditworthiness, which is judged by the borrower's scale, credit rating and the guarantee that the assets will generate cash. expect more and more concentration here

Absolutely worths checking easyui.site

yeah boy

Two characters, zero animators, about $1 per episode. Every model, every price and every mistake from building an AI cartoon series, from self-hosted GPUs to pay-per-use APIs.