Morgan (@morganlinton)

Okay, the @VulcanBench results for Qwen3.8-Max are in, and it is not what I expected. First, for anyone new to VulcanBench, here's a quick TL;DR on the eval suite: 23 frontier-hard software engineering tasks taken from real merged OSS PRs, run in a Docker sandbox, 3 runs per task across all three of its effort levels. No puzzles, no random abstract stuff, all real things engineering teams would do with these models. It looks like Qwen3.8-Max has a major overthinking problem, it uses a LOT of tokens and is very slow, period, no other way to see it. My cost to run this benchmark was $126.25, to run the exact same eval suite with DeepSeek V4-Flash was only $13.60. This makes Qwen3.8-Max an insanely expensive model. The tasks Qwen genuinely can't solve fail at every effort level, extra reasoning didn't help. The regression is almost all in work it already handles: six tasks that low solves every single time account for 83% of the 26-point drop, three of them collapsing to zero. It's not losing the hard problems. It's losing the ones it already knows how to do. Since Qwen3.8-Max hit a lot of wall clock budget caps, I thought I'd share more about this. - VulcanBench caps both steps (50–200) and wall clock (5–60 min), each scaled by repo size. - This is aligned with how comparable harnesses bound agents, DeepSWE caps rollouts at 100 environment steps, sitting right inside my step range; Terminal-Bench enforces a per-task wall clock; SWE-bench Verified scaffolds typically allow 20–60 min per instance with 250–350 step limits. - Every model on my chart gets the identical budget, and Qwen is the slowest model I've tested at 20–25 min/task. Soooo... Alibaba positions Qwen3.8-Max as trailing only Claude Fable 5. But on the kind of real coding work engineering teams would actually throw at it, under a fixed budget, its best setting lands mid-pack and its default lands last, so common. If you want to optimize for accuracy, Grok 4.5 is the move. If you want accuracy per dollar, DeepSeek V4-Flash is hard to beat, heck it's 10× cheaper than Qwen and you get higher accuracy. Qwen just isn't in the game at this point, this is not a model I could see engineering teams using for daily coding work.

I've been using Grok Build since the first 24-hours it was made available. It is still my favorite agentic coding cli, and there are so many little features I just love. So I thought I'd put together a little thread on what makes Grok Build so special, and some features you should try if you haven't yet. Let's kick it off 🧵

Very cool - Grok Build is clearly getting better by the day. Two nights ago I ran an overnight build and it failed. Last night, success. I really like the multi-agent orchestration behavior, it does a really good job of spinning up agents to build vs. review, and an orchestration agent that makes sure everyone follows the plan. I'm impressed, the @xai team is clearly making really meaningful progress here, very quickly.

Phew, Grok Build is really thorough, pretty incredible. Relay feature for PasteLocal is done, pushing to Github now. And if you want to know what this adds, here's a bit more about it, this is a feature I really wanted in there, and honestly didn't really expect to have done this week. • Per-peer E2E encryption: clipboard data is encrypted individually for each paired device using X25519 + HKDF + AES-GCM (no plaintext ever leaves the client) • Durable persistence: relay state is now stored in an atomic state.json with proper TTL handling and compaction. Pending clips survive relay-server restarts • Safe compaction: fixed unsafe map mutation during expiration/compaction that could cause data loss or undefined behavior • Full bidirectional CLI: new commands: pastelocal relay send, inbox, fetch, and status, plus pastelocal-remote --relay --peer --send • Auto-sync: when watch.enabled + are on, meaningful clipboard changes are automatically pushed to paired devices • Improved DX: doctor checks are now only shown when relay is enabled, and the TUI shows basic relay status • Multiple review cycles: went through full implement → review → fix → re-review (effort 4), plus a final targeted regression fix round Try it out and please send any and all feedback!

@steipete Thanks Peter, I’ve been super impressed with how much OpenClaw has improved over the last month. As someone that’s been using it since the early days, it’s kinda like a whole new product now, so much polish 🤌