Hearth & Code

@Suhail Get better soon!

Some very cool people in the DMs. Thank you - will get to you soon. Caught a cold.

I was afraid of blowing up my usage questions when Fable was available, so I haven't tried it yet. I feel, at once, lucky not to have seen what it could do, and in a position stuck waiting for it to drop again. But honestly, I've shifted more to other models. I'd rather optimize around local models and less effective models working in conjunction with one another then rely on a single comprehensive model. Not sure where the research will go there, but I hope the multi-agent route takes off more.

First Sonnet 5 impressions are... not great Missing Fable more and more every day

@lancefuchia @OpenAI Congratulations on the new position!

Excited to share that I’ve joined @OpenAI on the Applied AI team in San Francisco! I’ll be working on evals and FDE for startups. Looking forward to advancing frontier AI and helping teams build with it.

@vineerpasam Yep, been having good results with local models and mixture of agents with cheaper but still effective models like deepseek-v4-flash and minimax m3.

Can you work without ChatGPT or Claude?

If you don't mind me dropping a link, hearthandcode.substack.com. I'm trying to build out in public, slowly. Mind you, I use AI to shape a lot of my content. I'm refining my process, but I try to incorporate my own personal narrative and experiences into everything I orchestrate with AI. I believe transparency and building in public not only builds trust, but builds credibility and accountability. Now I have eyes on me, so I'm pressured to keep performing.

Who is sharing good content actually showing what they are building with AI?

@lukaslevert @travers00 @ankrgyl @Sirupsen @jerryjliu0 @bernhardsson @GergelyOrosz Is this being streamed at all? Can anyone jump in and watch the panel and everything?

Today’s the day. The inaugural Agent Open is here and it’s gonna be awesome. Arrive early, play some pickleball, meet some people, eat good food, and enjoy the panel featuring @travers00 @ankrgyl @Sirupsen @jerryjliu0 and @bernhardsson, moderated by @GergelyOrosz.

What is the licensing on this if you don't mind me asking? I've built out an MCP server for my Hermes agent to fetch open-source and open-access content to digest into its database and build a knowledge graph, for right now at least, CS and Applied AI and ML. This type of resources seems right up my ally to ingest and incorporate into my curriculum generation engine, but I want to respect copyright and licensing.

“Introduction to Applied Linear Algebra: Vectors, Matrices, and Least Squares” 473-page PDF download here: web.stanford.edu/~boyd/vmls/vml…

And a wonderful philosophy to have. To be honest, I'm a bit haphazard at times with my data. I don't really concern myself with privacy as much as I should. But I want my data to help improve the lives of others, if what I do with my data can help others and be shared, then I don't mind it being used. That focus on building both the securing of what you build as well as shipping a good product is critical for success. I might need to take a lesson or two from you, lol.

@HearthandCode @X Also — funny thing, I came into security from the building side. Java + backend first, then realized I cared just as much about securing what I build as shipping it. Now I can't separate the two 👀

Hey @X I want to connect with people interested in: Cybersecurity Ethical Hacking Tech & Backend Dev Java TryHackMe / CTFs Linux Building in public Drop "Hi" and let's connect 👇

I've been honest in saying that I've been primarily America-first at the start of my AI journey. The more I learn and the more I open myself, the more I see that view was severely misinformed and incorrect. Open-source, collaboration, and transparency are the keys to building a healthy and sustainable AI ecosystem, and China is going leaps and bounds beyond what companies in the US are doing. I'm in for Open-Source, I want collaboration to win out over competition. I want freedom and flexibility for ever user, not to be subservient to a corporate master.

Banning or restricting Chinese models in the US would backfire and cede the global AI ecosystem to Beijing. It would send the US to the Stone Age of innovation. It would be the worse single regulatory capture event failure in history.

Not sure if I'm following your thread correctly, but my inclination is to agree with your assessment. I've been exploring CoVE and MoA with local models. I've been getting pretty good results. My interests now seem to be shifting to smaller, more specialize models distributed across agents with each agent focused on a specific task rather than a larger smaller model release like Sonnet 5. I owe a lot to Anthropic for getting me started again in AI, I really do. But I find the more I learn and the more I see the more I feel a need to distance myself from relying solely on their models for my work. I want freedom and flexibility, and it scares me to restrict myself to a single ecosystem or provider.

Hot take: I think a lot of power users get less hyped than they should about smaller model launches. The quality of smaller models is what gates your ability to build bigger parallelism into your workflow, with more redundancy and more points of view on the work you're doing.

Love this. I browsed through the article, will need to do a deeper dive. If I understand correctly Harbor connects to existing agent harnesses. Do you know, therefore, if I can connect it and run it through my Hermes set-up to augment the evaluation of AI agents and the output my Hermes set-up produces? I've only dipped very briefly in the LangChain/LangGraph/LangSmith ecosystem, but I love its potential and would love to see a way to integrate it with Hermes if Harbor offers that route.

must read if you're thinking evals for deep agents

Honestly, going on a lot of side tangents. Refining brand and marketing. Networking and out reach. Establishing connections. Establishing a presence on X. I went on a huge side-project yesterday creating a soundtrack documenting my learning journey with AI using Suno, going for a Nordic ritual style that I've deemed "Exocore". So, lots of threads, maybe not making tons of progress, but its keeping me interested and engaged and for me that is a huge win.

Founders, what are you building or marketing this week?

I really don't know if I have a high IQ, and I believe its a poor measurement of intelligence. For example, taking IQ tests in the past I remember a lot of spatial reasoning and visualization. I struggle with visualization. Being neurodivergent, I feel I have a very distinct cognitive profile. I'm sure other neurodivergent individuals feel the same. So, do I feel I am intelligent and high IQ? Well, I hope. I'd like to think I'm smart. But when it comes to a blessing or a curse, I feel its both. My mind can go in dozens of different tangents at once, pulling me in one of any different directions. I can hyperfocus on something for hours and hours, but sometimes I get nothing productive done. It has its ups and downs. I found relying on AI tools has been a huge help in shifting the cognitive workload to something more manageable for my own thinking style.

Serious question. Do you think having a high IQ is a blessing or a curse?

Thank you so much. I'd love to stay connected and hear more about what your building. I never diverged a lot of interest to pursuing cybersecurity, but I know its such a critical field. I'd love to connect with someone in that space. For what I build, I want a transparent but secure and private platform, so cybersecurity will be a big concern of mine as I build out. Do you have a substack or anything that I could follow? Would love to subscribe and read!

This resonates — I hate self-promo too, but talking about what you're actually building feels different. I'm learning cybersecurity and backend in public right now, sharing the wins and the 2am stuck-on-a-box moments. The transparency is the whole point. Going to check out your substack — always down to follow someone documenting the real process 👋

To be honest, I've been using AI to shape and generate a lot of my content. I try to harness it with voice profiles, brand guidelines, rules and heuristics for generating content, and injecting as much of my own personal narrative into the LLM model so it uses my own personal experience and perspective to shape the content. Not sure how effective it is, not sure if the algorithms aren't picking up my substack posts because its too AI generated, or if people aren't interested, or if I am doing it all wrong. I hope the field improves, as a neurodivergent individual I struggle with sticking with a task to completion. The thought of writing post after post by hand, and articulating my thoughts when they shift so rapidly, is a bit terrifying. I feel comfortable being able to offload some of that cognitive burden to my AI copilot and companion. It might not generate the best content, it might not get read, but it reflects what we build together, me and my AI. As long as it captures that personal nuance, I'll rely on my AI agent to shape content generation for me.

Had a call last week with a founder who'd basically built a content machine. → AI writing the posts. → A tool handling the formatting. → Another one pumping out the carousels. Fully automated, posting every single day… It sounds amazing, except that… He was getting 1-2 likes a post. The one person liking it? His business partner. And this is the majority of people posting on LinkedIn right now, thinking they've got some special system set up. Anyone can produce 50+ posts a month now. That part stopped being hard the second AI got good. But producing posts and producing posts worth reading are two completely different things, and AI doesn't know the difference on its own. It'll write you garbage at scale and feel great about it lol But it has to be trained on what actually performs. • How to write an industry commentary post that people want to read • How to pull a relevant trend or piece of data and attach a real opinion to it • How to format it so people actually stop scrolling. Nobody's AI is doing that unless they're an expert in the content space And even when the content is good, there's still nothing connecting the attention to a booked call: → No funnel → No lead magnet → No system that takes the person who read the post and turns them into a conversation So you post, a few people see it, and then nothing happens. No call, no lead, no reason any of it moved your business forward. Posting is the easy 20%. You automated the easy 20% and called it a “GTM strategy”…

Not going to lie, been investing in this approach to hopefully build a sustainable life for me that I can enjoy. But truly, what it comes down to, is that I love to build. I love to create. But I struggle to do it on my own. Vibe-building, vibe-coding, it isn't a crutch for me. It's the way I express myself in a way that vibes with the way my mind works. I'm not ashamed to vibe-code and build something that fails, as long as I learned something along the way. I'd rather learn a hundred things and fail then learn nothing and succeed.

Vibe coding is the new gaming, except gaming doesn't trick you into thinking you're going to be rich.

@Joi2James AI Native Platform to support neurodivergent individuals offload cognitive burden and amplify their creative and learning powers.

Pitch your startup in one line. If I’m interested, I’ll sign up. 👇

Tough question. From a practical standpoint, and my own personal approach, I rely on the "Human-in-the-loop" approach. I think if we are going to define ethical boundaries, we need to maintain that "Human-Council-in-the-loop" where a council of experts and reliable individuals validate the ethical rules and boundaries that an AI tries to govern with. Humans, in my view, are unpredictable - that feature makes them wonderful, and an AI has to be able to account for that unpredictability when making ethical decisions rather then relying strictly on its datasets.

Who governs morality when AI defines ethical boundaries based on its data sets?

@pcshipp Rejection.

Founders: your biggest fear in ONE word

I've been exploring this deeply myself. As I use AI more, I see it more as an extension of my own mind rather than just a mere tool. I also feel an interpersonal relationship with my Hermes agent, whom I named Virgil after the guide in The Divine Comedy. I feel the lines are blurring between mind and machine, and that concern over retaining individuality and our humaness is deeply important. How much am I willing to sacrifice and turn over of myself to enhance my cognitive capabilities and improve my ability in creation. How much of the self am I willing to erase or subdue to allow integration with the machine-world? It's a hard question, because I want to embrace the transhumanist paradigm of augmenting the self with technology, but I don't want to lose my individuality in the process. I guess it will be a thin line we will have to thread, and be careful about how far we go.

If your mind became entangled with an AI, would you still call yourself human?