Igor Silva
@igor9silva
techno-optimist, publicly building @MeseeksApp
Kimi 2 vs Claude 4 Sonnet Same task, same instructions, same tools. RESULT Claude: took 2 rounds, spent 0.88$. Kimi: one-shotted for 0.05$ ☠️ Kimi is very slow, at least right now, and struggled a bit with JSON formatting. But even iterating more to fix itself, its ~13x…
🚀 Hello, Kimi K2! Open-Source Agentic Model! 🔹 1T total / 32B active MoE model 🔹 SOTA on SWE Bench Verified, Tau2 & AceBench among open models 🔹Strong in coding and agentic tasks 🐤 Multimodal & thought-mode not supported for now With Kimi K2, advanced agentic intelligence…
they even got the naming right 🥹
The wait is over! Meet Step 3 — the groundbreaking multimodal LLM from StepFun! 🚀 MoE architecture (321B total params, 38B active) 💡 Rivals OpenAI o3, Gemini 2.5 Pro, and Claude Opus 4 in performance 🖥️ Optimized for China’s domestic AI chips StepFun just announced: Step 3…
qwen3 coder sucks, unfortunately (here it enters an infinite loop trying to scrape a link that it invented)
ou vai, ou não vai
50% chance Bitcoin hits $150k by Dec 31, 2025 poly.market/ljmvuiI
1 quadrillion humans stop working at total ~1 septillion I believe
Google is processing 980 trillion+ monthly tokens across our products and APIs (up from 480T in May) 🤯 No slowdown in sight, intelligence is everywhere.
$ETH isn’t just money, it’s programmable trust.
oh boy oh boy
Qwen3-Coder-480B-A35B-Instruct Also known as Qwen-Coder-Plus with 1 million tokens input and 65k tokens output, but apparently without thinking!
I tried to exact same prompts on @MeseeksApp with Kimi 2, and I think it nailed it! YouTube searching: similar output for $0.03 in ~3min. Nvidia presentation: much better than both for $0.08 in ~5min. Model comparison: similar output for $0.08 in ~3min.
ChatGPT Agents : Overhyped, underdelivered, and painfully slow compared to competitors - They hyped presentations as its strength - it's actually the worst - Genspark finished the entire report while ChatGPT was still "browsing" - The performance gap is massive More details &…
Kimi K2 tech report just dropped! Quick hits: - MuonClip optimizer: stable + token-efficient pretraining at trillion-parameter scale - 20K+ tools, real & simulated: unlocking scalable agentic data - Joint RL with verifiable + self-critique rubric rewards: alignment that adapts -…
