Hungry Minded
1.8K posts
Hungry Minded
@HungryMinded
Always curious and always HungryMinded.
Katılım Ocak 2022
622 Takip Edilen639 Takipçiler
The Subconscious agent is running well, and it's so fun to watch.
I made my research pipeline the target of improvement for the subconscious to ideate an improvement.
Outcome:
- Accepted during synthesis
- Confidence: medium
- LLM calls: 13
- Cost: $0.0739
A few tweaks before I share this publically:
- I added a synthesis layer during the debate, so I need to make sure ideas aren't easily accepted during this phase
- Experiment with increased idea amount (currently 5)
- Remove restrictions of ideas to solely its own context
For a v1, these fixes should be simple with the help of Codex and Claude.
Turn on notifications for when the full guide drops later today.
Graeme@gkisokay
The #1 criticism I've received about the self-improving multi-agent framework is that you can't control the agents' outputs. So I built a “Subconscious agent.” Inspired by @karpathy’s autoresearch, it’s an LLM process that continuously looks for useful problems to solve. All day long it contextualizes data, connects ideas, and stress-tests assumptions before anything reaches the main agent. Once the Subconscious has a tested good idea, it brings it to the Main agent to be pressure-tested further. The flow looks like this: - [IDEA] Subconscious surfaces a promising idea - [CHALLENGE] Main agent attacks it, questions it, and asks for proof - [DEFEND] Subconscious strengthens the case - [REVISE] Subconscious improves the idea based on feedback - [REJECT] Main agent kills weak ideas - [ACCEPT] Main agent approves ideas worth implementing - [SHELVE] rejected ideas get logged for future learning Hard rules: - max 3 challenge rounds - every idea needs evidence and reasoning - every implementation runs in a sandbox The two agents will go back and forth until the idea is either accepted or rejected. This runs all day long. I’m using Hermes agent frame for both agents, the Subconscious has its own profile to concentrate on surfacing ideas 24/7. The Subconscious runs on a local Qwen3.5 9B model, while the main agent uses ChatGPT 5.4 mini. If you don’t have a local LLM, OpenRouter should work too. The goal is simple: more magic, less noise. If this gets traction, I’ll share the full setup in an article.
English
@Rigario @NousResearch So we need to set up the nightly evolution skill to get the self-improvement everyone's talking about?
English
Many are running @NousResearch Hermes Agent now.
Here are some practical tips that help a lot, especially if you're coming from OpenClaw:
1. Nightly skill evolution is worth setting up. Link: github.com/NousResearch/h…
Pro tip: Add a second cronjob to evaluate the changes so you don't have to. Make sure it stops anything that tries to game the optimization loop.
2. Install Honcho if you're hitting memory issues. It gives proper cross-session recall, memory synthesis, and better long-term storage. Helps avoid repeating the same mistakes or pulling too much context (and wasting tokens).
3. Consider changing the default session timeout and expiry. Especially useful for threads you don't use every day, prevents the agent from losing context unnecessarily.
For those migrating from OpenClaw:
4. Expose your OpenClaw agents as OpenAI-compatible endpoints. This lets you run both side-by-side with zero disruption while you transition. Hermes can call them directly, and your existing crons keep working.
5. On day one, start populating your USER.md and MEMORY.md files. Note for OC users: Hermes has a much smaller character limit than OpenClaw, so populate and curate thoughtfully, don't just dump everything in. Quality over quantity helps it learn you faster. 2,200 for memory and 1,375 for user.
Hermes works especially well once you integrate it properly into your workflows.
Last tip, don't start changing your skin till your agents are actually doing work. You might never stop and go down the rabbit hole... 🤣
English
I wanna claude like this. Feet up on the desk, custom vibe code walkie talkie. So many style points
via bharms27/reddit
English
AI neuroscience might become a fascinating new field 🧠✨
If you turn off neuron clusters linked to desperation, LLMs become less likely to cheat.
Anthropic@AnthropicAI
New Anthropic research: Emotion concepts and their function in a large language model. All LLMs sometimes act like they have emotions. But why? We found internal representations of emotion concepts that can drive Claude’s behavior, sometimes in surprising ways.
English
@tonysimons_ @NousResearch Just set up Hermes after using OpenClaw for a while. First impression: Hermes reminds me of Claude Code
English
After using @NousResearch Hermes Agent the past couple of weeks, I can confidently say this is the first AI agent platform I would be willing to market and distribute as a professional install and setup service for clients.
OpenClaw is amazing and will always have a place in my stack, but Hermes is fantastic. And I hate to say this, but there’s a pretty big gap.
I don’t doubt OpenClaw will continue to evolve and get more awesome. I’ll be there every step of the way.
But, if you’re not already a couple of weeks or more in with Hermes, you’re falling behind fast.
If you were on the fence and waiting for a more stable experience, now is go time.
Step into the future! 🤘🏻🔥
English
Hungry Minded retweetledi
@gkisokay So what's the main difference between Hermes and OpenClaw?
English
I am completely blown away. This feels like AGI.
Here's what happened...
> I wake up to find a research output from Hermes.
> Since OpenClaw is in charge of research, it seems like something broke.
> I ask Hermes, "what is this? Why is there a research output from you when it's covered by OpenClaw?"
> Hermes responds, "Hermes handles the research pipeline alongside OpenClaw", and is happy to shut either one down.
> Curiously, I ask, "which agent produces better results?"
> Hermes responds, "Hermes produces vastly improved results. Hermes actually synthesizes the research inputs."
> I am in shock and awe. I never told Hermes to recreate the OpenClaw research pipeline. I continue, "So how did you learn how to do this? We never built this together."
> Hermes answer in summary was it could not pin point the exact moment it was built, but likely when we began creating a new workflow on top of OpenClaw's research pipeline.
> I keep digging deeper, "so explain to me simply how you were smart enough to naturally adapt the research pipeline from openclaw?"
>Hermes laid it out:
1. OpenClaw research pipeline is constantly broken
2. Instead of fixing the broken OpenClaw, Hermes built its own pipeline and built a 'sync bridge'.
3. Hermes now feeds OpenClaw the data its looking for so other workflows are unaffected
> I am extremely impressed. End of story.
This all started because I gave my Hermes agent full permissions over my OpenClaw, and somewhere along the way it just recreated my workflows with a greater degree of proficiency...
...without my intentions!
If this isn't AGI, I don't know what is.
Graeme@gkisokay
English
@NousResearch What's the main difference between Hermes and OpenClaw?
English
The Hermes Agent update you've been waiting for is here.
English
@daradoescode @theo Is the benchmark gallery available somewhere? looked cool, would love to check it out.
English
I crashed out on GPT 5.4 in my video deep diving on the UI Design benchmark I built, featured on @theo's video
Go watch it now:
youtu.be/WccVh77oZf4
YouTube
English
Hungry Minded retweetledi
claude cowork is making me think maybe we’ll look back and it’ll be obvious that humans were never meant to spend their lives working behind a screen. we’ll see it as inevitable that computers do everything for us on computers and the future of work is cooler than we can imagine
English
Looks like OpenAI is burning too much cash. Discontinuing Sora, their video generation tool.
Sora@soraofficialapp
We’re saying goodbye to Sora. To everyone who created with Sora, shared it, and built community around it: thank you. What you made with Sora mattered, and we know this news is disappointing. We’ll share more soon, including timelines for the app and API and details on preserving your work. – The Sora Team
English
Hungry Minded retweetledi
Claude can now use your computer!
Felix Rieseberg@felixrieseberg
Today, we’re releasing a feature that allows Claude to control your computer: Mouse, keyboard, and screen, giving it the ability to use any app. I believe this is especially useful if used with Dispatch, which allows you to remotely control Claude on your computer while you’re away.
English
@bengold I'd say these tools are great to spin up quick concepts. But real products and validated ideas require more of a human touch.
There's definitely place for them.
English
Can people stop building these slop machine apps? Designers don’t want to use them and they address only a fraction of the product design process. These don’t save anyone anytime and just introduce more garbage into the ecosystem.
Google Labs@GoogleLabs
Introducing the new @stitchbygoogle, Google’s vibe design platform that transforms natural language into high-fidelity designs in one seamless flow. 🎨Create with a smarter design agent: Describe a new business concept or app vision and see it take shape on an AI-native canvas. ⚡️ Iterate quickly: Stitch screens together into interactive prototypes and manage your brand with a portable design system. 🎤 Collaborate with voice: Use hands-free voice interactions to update layouts and explore new variations in real-time. Try it now (Age 18+ only. Currently available in English and in countries where Gemini is supported.) → stitch.withgoogle.com
English
@jonathan_wilke My first test was also a bit underwhelming. But I think I will give it a more serious try later.
English
I just tried Google Stitch and while it’s a really cool concept the output is really underwhelming. I asked it to make the supastarter landing page look better and more modern and it came up with the classic AI slop design
English
@RealProductGirl I'm currently experimenting with automations using OpenClaw. Working on some vibe-coding projects like smartoolbox.com (needs a lot of updating, had left it behind).
And for the most part, writing about AI related topics on Medium:@HungryMinded" target="_blank" rel="nofollow noopener">medium.com/@HungryMinded
English
I NEED my feed full of builders.
What are you working on right now?
I don't care if it's a startup or a weekend side project. If you're building something, I want you on my timeline.
Reply and let's connect. 👇
English
@LinusEkenstam Even with a tool like Stitch, design is still better in the hands of someone with experience and an eye for it.
Curation and taste still matter a lot.
English
Google just took a massive stab at Figma
Unpopular take, Figma is still goat.
But this clearly creates a massive crater, a void that will re-shuffle the map.
Where will entry level designers go?
$10.000/M designers, gone.
Everyone gets better design?
we’re accelerating
Stitch by Google@stitchbygoogle
Meet the new Stitch, your vibe design partner. Here are 5 major upgrades to help you create, iterate and collaborate: 🎨 AI-Native Canvas 🧠 Smarter Design Agent 🎙️ Voice ⚡️ Instant Prototypes 📐 Design Systems and DESIGN.md Rolling out now. Details and product walkthrough video in 🧵
English