More posts
Simular Box is popping up everywhere this week. It hands your agent a cloud computer that just sleeps when it's not being used. Under the hood it's a cloud Ubuntu desktop. Folks who've poked at it are seeing ~4 cores, 8GB RAM, 40GB disk, plus sudo. Wire up any agent and it gets its own browser, terminal, and desktop: it can browse, click, run code, all isolated from your real machine. Feels a bit like what Muse is doing. Free tier exists: a daily credit allowance that resets every day and doesn't carry over. 🔗 box.simular.ai
The whole world finally went quiet. I'd been on the official deepseek-v4.1-flash and it just would not shut up. Verbose, over-thought everything, fired tokens out fast, then crawled to the actual result. DS scored 33 on the AA benchmark (33% task success). At $0.6 a run that's ~3 attempts per finished task, so $1.8 each. Terra (high) scores 42.4 at $0.26 a run. ~2.5 attempts per success, $0.65 each. So the model that looks more expensive is cheaper per real result. With GPT Pro 200, Terra chews through ~10B tokens a week for me (YMMV), and that $200/mo covers one person's whole usage. Model picker: haoyi.im/models
Won't refund the subscription either. I'm half convinced the appeal email is itself phishing, because I'm pretty sure they log the IP. DHH came to Shenzhen recently and showed how he uses Claude Code over there. That's the safest way now, honestly: run Claude Code on an overseas machine, connect through Tailscale, never use it locally. So I need to rent an overseas box with a residential IP, and register and log into Claude there.
Until late September it was the usual setup: parked in front of my laptop, terminal open, bossing pi around, fanning out subagents, then eyeballing diffs myself and filing the PR. Glued to the chair all day, every day. Late September I moved the whole thing to my phone, Hermes runs the show, pi takes the orders. My back is very grateful. Three weeks in, still no regrets. Stack: Hermes on flagship brains (gpt-6-astra / opus-5.5), pi on mid-tier (ds-v4.1-flash / sonnet / terra) with as few plugins as possible, the two talking over a2a. Hermes reads the issues, slices them into small tasks, spins up a worktree for each one, and dispatches a pi into it. When a pi finishes it pings Hermes, who decides: ship the PR or bounce it back for rework. I get back from a full day of being a tourist, or just wake up, and a dozen PRs are sitting there done. Pretty sweet. The catch is how black-box the whole thing is. My emotional arc has been a rerun of early vibe coding, except back then I at least read the diffs. Now even that goes to a flagship model. Boxes all the way down. If you don't have data anxiety, grok-bot and chatgpt-dot basically ship this out of the box, grok-cli and codex built in, native agent-to-agent comms. Downside: your code lives on their servers, everything's moderated, and one ban takes it all with it.
