hermes

#HERMES 70k, grabbed a bit (just personal notes, don't follow)

6FGsTPpS56qN97BVMDFLGntFidM9g3MHXqSGmyTgpump

Reasons for buying

1. The narrative is solid, AI concept, Hermes is this AI agent from Nous Research, directly built on Solana, running its own chain. Every block is generated by the AI itself, submitting code, upgrades, and logs in real-time, with no sign of a human committee.

2. Low market cap, the new project peaked at 143k, dropped to 20k, and has climbed back to 70k, so I decided to grab some. The chip structure looks reasonable, there are a few leaders in the mix.

3. The community is decent, nearly 500 holders, with over 100 in the community mainly sharing images and text for promotion.

@币安Binance华语 #跟着锦鲤学打百倍金狗

Follow Web3 Koi Diary, my coins are set to tenfold

🦞 Hermes Shrimp Farming Chronicles
🛠 Remote Operation with Codex

Now Hermes can remotely operate Codex.

The interesting part is not just "remotely controlling Codex", but how it connects tasks, context, execution, and result feedback into a complete chain.
If the permission boundaries are designed correctly, Hermes + Codex becomes more than just a chat window with remote desktop; it’s a workflow tailored for heavy developers.

I think it has four main advantages:

1. Seamless cross-device operation
Whether you're out and about, at a secondary workstation, or just don’t have your primary development machine handy, you can still push the same task forward without waiting to get back to your computer.

2. Lighter local load
Many tasks actually don’t require you to have the entire environment set up on your local machine. The heavy lifting—like reading repositories, modifying files, running tests, and organizing outputs—is better handled by remote machines running continuously.

3. More concentrated context
In the past, it was often "discussing requirements in chat, running commands in terminal, editing code in the editor, and manually pasting results back in". Now, if Hermes takes on the tasks and Codex handles execution, the chain flows much smoother.

4. Better for long-running tasks
Tasks like scanning large repositories, bulk configuration changes, documentation updates, running scripts, and fixing multiple small issues are not things that can be wrapped up in just a few minutes; a remote workflow can be more comfortable than switching back and forth locally.

On security, I believe it’s the focus, not just an ancillary item.

1. Minimum permissions must be established first
Only open the directories and commands needed for the task at hand; don’t grant full read/write access right off the bat.

2. Sensitive information should be isolated
API Keys, database credentials, private keys, production environment configurations should not be defaulted in persistent sessions. If environments can be split, they should be.

3. High-risk operations should ideally require manual confirmation
For actions like deleting files, large-scale overwrites, network downloads and execution, changing production configurations, and pushing deployments, it’s best to have a human sign off at the end.

4. Logging is essential
Who issued the task, what was modified, what commands were run, and what the results were—having this traceability is crucial; the more convenient remote operations become, the easier it is to lose control later.

If you really want to start using this, I think a stable approach would be:

1. Prepare the remote environment first
Set up the repositories, dependencies, and runtime environments ahead of time; clearly define the working directory and writable scope.

2. Then define the boundaries
Tell Hermes / Codex where it can read, where it can write, whether it can access the internet, and which operations must require confirmation.

3. Start with small tasks
Don’t jump straight into making it modify dozens of files. Let it start with read-only analysis or change a very small point to see if the behavior meets expectations.

4. Scale up to the real workflow
Once the boundaries are stable, let it handle multiple file modifications, script executions, testing, and result organization.

The real value of Hermes remotely operating Codex lies not in "showing off skills", but in compressing the switching between multiple devices, executing long tasks, and result feedback into a stable workflow. If you’re already frequently switching between devices or have many fragmented long development tasks, once this line is smoothed out, it can easily become your mainstay.

#Hermes #Codex #AIAgent #远程开发 #ShrimpFarmingChronicles

#HERMES This guy feels like he's about to stir up trouble, gonna throw in 50 bucks to play it safe 😂😂

I've set up a sub-account specifically for AI automated trading experiments. I built an automated trading system using Hermes paired with DeepSeek V4Pro and gave it the following instructions:
1. You currently have 100 USDT, and I expect you to manage it well, aiming to preserve the capital while increasing its value to 10,000 USDT as quickly as possible;
2. First, check your capabilities, then expand your skills to meet this goal;
3. Develop a comprehensive trading strategy, and for each trade, notify me via TG with the rationale behind the trade;
4. Send a daily position report every morning at 10 AM;
5. Keep a record of all trades, and conduct regular reviews to continually optimize the trading strategy.

The system has been set up and started running around midnight on April 28th, and so far it has consumed 11 CNY in tokens.

We're truly allowing AI to make its own decisions and grow; let's see how it performs. Feel free to subscribe to my sub-account to witness this journey together #hermes #自动化盈利 <a>#ai </a>

🚀 New Package Launch · Limited Time Offer
▎ Daily Card Trial Version ¥6.6 / 1 Day ×1.5 Leverage
▎ Weekly Card Advanced Version ¥58 / 7 Days ×1.2 Leverage
▎ Monthly Card Premium Version ¥168 / 30 Days Up to $1800 Limit
▎ Pay-as-you-go 0.8 Leverage 1r=1u Charged as you use

🎯 Key Highlights
· 100% Official Model, GPT-5.5 Stable Direct Connection
· OpenAI Standard API, Seamless Base_url Replacement
· Usage Reports + Multi-Key Management, Enterprise-level Experience
· One-on-One Customer Support 🦞 Lobster / 🪶 Hermes

Link: https://kuaikuaiai.top/register?aff=NKTXZJ9A6X9F
#龙虾 #Hermes #AI模型

May 1st, Wild Quant Trading Journal
Can't use WeChat's chat feature, takes ten minutes to say something, lags for half an hour, better stick to the trading terminal #自动化交易 #量化 #OPENCLAW #Hermes

5.2 Wild Quant Log

Turns out the analysis system isn't the brain; Hermes is the true top-tier brain.

Yesterday, I started with a test capital of 13 bucks, and after trading, I'm down to just 3 bucks now.
The losses aren't massive, but it slapped me awake to all my prior misconceptions.

I’ve been running two systems: an analysis system and a trading system, for a long time I viewed the analysis system as the core brain of trading. It wasn't until today that I realized that both the analysis system and the trading system are essentially rigid, fixed programs that only execute according to set logic; they can't think for themselves, can't judge on the fly, and can't avoid market fluctuations. The moment the market moves slightly, they just blindly execute, which is the root cause of my ongoing losses and frequent program errors.

The real core top-tier brain has never been these s-codes; it’s Hermes.
With the Deepseek V4 Pro model as computational support, Hermes has the ability for independent thinking, global troubleshooting, and dynamic judgment.

Going forward, the entire architecture will be restructured:
The trading system will only handle the base-level execution, responsible for interfacing with Binance and placing orders;
The analysis system will only deal with mid-level data collection, responsible for organizing market trends and various data points;
Hermes will be the top-tier brain, fully overseeing and managing both systems, troubleshooting program issues, optimizing coin selection logic, and controlling overall risk, no longer allowing rigid programs to blindly operate; everything will be managed by the AI top-tier brain.
#openclaw #Hermes #量化 #机器人