Founder @holoweave. I mostly post about agents and a bit of investing. Built product orgs at CoinMarketCap, Binance, Chainlink and Zendesk.
Joined April 2012
- Tweets589
- Following226
- Followers343
- Likes141
Pinned Tweet
After testing Grok Bot, Muse, Hermes Agent, Hark, Instinct, and OpenAI Dots on real personal assistant work, shopping, and agent coordination through Holoweave Weave...
Hermes on its own computer/VM is S-tier once you customize it. For the best balance of capability, always-on access, and actually finishing jobs, Grok Bot is my top pick right now.
I said this before the quota was restored and have seen so many people say otherwise.
I think people are very quick to jump on the hate bandwagon these days.
Grok Bot is back! It reverted exactly to where my usage was before the issue happened.
Seems like my generous interpretations were true.
Based on the amount of quota that Grok Build has and how fast Grok Bot exhausted even on the highest plan, the most charitable interpretation here is that Elon doesn't know that there is an actual bug causing issues.
That's the issue with internal users having unlimited quota.
The sad part is the fact that it's taking so long to resolve whatever usage billing dashboard issue.
If this would not be acceptable in all of my old product orgs and I know there are some of my ex-colleagues there right now. What's up?
Grok Bot as a product is already great, now if we can just get a great model, it's really going to take off.
The new @NousResearch subscription is pretty good, using a small model results ins quite a bit of usage at the moment. "Small" could be DSV4.1 Flash which is pretty good.
Did some ops, troubleshooting and research work and it only used 3% of daily and 1% of the week quota.
This is quite insane. With all the competition that is around, who even really uses Instinct?
Pass this to your agents and ask them to check all your devices and mitigate if needed.
‼️ UPDATE - Tailscale confirms the Tailworm SSH flaws are real.
Its new investigation page says most of what it found is privilege escalation and calls the issues “serious.” Fix targeted within 24 hours — no CVE or patched version out yet.
Mitigate now:
→ Set TS_SSH_DISABLE_FORWARDING=1 in the tailscaled service and restart
→ Tighten SSH access rules to the users/devices that need them
Read: thehackernews.com/2026/10/un…
Really cool bundle, I'm kinda subscribing just to support the project, will swap over to see what its like.
We've revamped Hermes Agent subscriptions.
Every new plan now includes a cloud agent out of the box, and usage has daily and weekly limits instead of one monthly block of credits.
Subscribers can switch to the new plans any time at portal.nousresearch.com/bill…
1.5Tb needed to run. This model is truly only open weights for commercial clouds to run.
Releasing the model weights and technical report of Kimi K3.
Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights: huggingface.co/moonshotai/Ki…
Tech report: github.com/MoonshotAI/Kimi-K…
Tech blog: kimi.com/blog/kimi-k3
Look like the Chinese open source part of the cycle is here. Looks like its about a GLM 5.3 class model but within 250GB.
Thats the HBM of the GB300 or dual spark territory.
MiniMax M3.1 Flash weights are now public. 428B parameters total, 23B activated per token.
Agentic performance: A generational leap in agentic tasks, coding, and cybersecurity — 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam, and 82.2 on CyberGym.
Efficient by design: MiniMax Sparse Attention and DSpark speculative decoding improve inference efficiency.
Hugging Face:
huggingface.co/MiniMaxAI/Min…
ModelScope:
modelscope.cn/models/MiniMax…
API: platform.minimax.io
M Plan: minimax.io/m-planMiniMax
MCode for desktop: agent.minimax.io/download
Quota and cost: Muse is the most generous in a way that stays usable.
Dots has basically unlimited main-bot chat, but that doesn't matter when it doesn't complete the work. Hermes sits in its own category (you control the stack). Grok Bot is tighter, which tracks if they have product-market fit and know the pricing they want.
Bottom line: start with Grok Bot for the practical combination of power + access, then layer Hermes for the heavy lifting once you need maximum customization. The always-on VM + multi-profile setup is the closest thing to an actual teammate so far.
What are you running these for right now? Let me know if you'd like me to share anything about my setup.