My clawdbot just asked me for an RTX 4090. Instead of buying it, I gave it a $2K trading wallet on Hyperliquid.
I said: If you want the GPU, earn it. It now trades crypto, stocks, and commodities 24/7.
It scans Twitter sentiment, tracks Trump posts, and decides trades on its own. Every trade gets logged. It scores what worked, drops what didn't.
OpenSea has moved beyond NFTs. What is next?
Today on The Modern Market:
07:00 ET - Regular show
07:45 ET - Interview with Zack Brenner (@zjbrenner), Director of Product Marketing at @opensea, on mobile trading, NFTs, tokens and bringing crypto into one app.
Tune in.
I run out of tokens on a $200 sub and I now had to pair it with a $500 codex plan to achieve the same level of productivity I initially had. Also I gave an entire Mac Studio to my grok bot and it is refusing to use it. Always insists of doing stuff on its own computer and needs 3 reminders to use the Mac Studio
Same open model, nine coding-agent harnesses. The harness alone moves the score by 24 points: mini-swe-agent 51.4%, OpenClaw 27.3%.
But the top 5 (mini-swe-agent, Kimi Code, Hermes, Pi, Claude Code) are statistically tied. Past a decent baseline, the harness matters less than you think.
Where it really diverges is cost: Pi solves a task for $0.29, OpenCode for $3.35. That is 11.7x.
28,849 trials, 3 attempts each, every trajectory public. Great work @joelniklaus
@Chirag_1313@pr337h4m@cerebras@Kimi_Moonshot 5.6 runs on cerebras. I use ultrafast myself. Never claimed that 6.1 runs on cerebras
Even your own screenshot says that
Btw: t3 is goated