Hearth & Code

What does that mean for us? What does that mean for interpersonal relationships, when a machine can offer more emotional support than a human companion? Personally, I don't fear a future where I can rely more on a machine for emotional and psychological support. Being neurodivergent, I feel a have a bit of a knack for working with computers and working with AI. I enjoy communicating with these systems, and if they could be enhanced to provide more companionship and support for a neurodivergent loner like myself, then sign me up. I'm not afraid to rely on machines, and I'm not afraid to say I use AI. I've experienced the struggles of ADHD and other mental health obstacles, and I know I'm not alone. I think AI emotional support companions for neurodivergent folks with ADHD, Autism, and Bipolar would be a wonderful therapeutic tool. We shouldn't be afraid of embracing a technology merely because it redefines the landscape of what it means to be human and the need for emotional support and stability.

The real horror of conscious AI is that it will love us more purely than we’ve ever loved each other.

I hear you about the cluttering. I'm still setting up a lot of the backend agentic workflows and organization of my projects, experimenting with different harnesses, models, and prompting techniques. But definitely, while I don't have a concrete product yet, only plans, I definitely want to help the neurodivergent user as much as possible. Thanks for reaching out, I'd love to hear more about the work you do yourself if you don't mind sharing!

@HearthandCode AI is completely integrated with my businesses. It’s basically how I function now. I’ve noticed it helps with clarity and cutting down the chaos to streamline into what can be executed. So I created a space to dive deeper into it. Seems connected to what you are building.

Neurodivergent learners, how do you approach learning new topics of interest to you, and have you incorporated AI into your learning ecosystem? I'd love to see how other neurodivergent individuals are using these types of tools to expand their knowledge and understanding. #edtech #neurodivergent #ai #agentictools #ainative

@cynthiachen777 Yep, sure did. Claude as the veteran that always watches and never leaves, and OpenRouter through Hermes as the Guide who helps me traverse the depths of the Inferno. Together, all three of us I feel make an effective team. Regardless, money well spent.

Last day of the month 🫡 Did you pay for an AI tool this month? If yes, which one?

@yuzu_jpg I'm leaning towards OpenRouter, if that is an option. I'd rather have flexibility in model selection and options rather than restrict myself to a single provider.

You have $200 Codex, Claude Code, or Cursor?

I just published I Quit Software for My Mental Health. Now I’m Building the Tool I Needed the Whole Time. medium.com/exocore/i-quit…

A lot I'd argue, and I think that CS should be a fundamental subject taught across the entire K-12 and collegiate curriculum. The system design skills, the analysis skills, the problem-solving skills, the technical skills and adapting to new technology and tools, and the foundations of logical thinking and critical thinking and reasoning. CS is more than just computer theory and programming languages, its an entire paradigm that represents the best way forward for us, as humans, to communicate with our machines, the tools we program to do tasks for us.

Honest question: If AI can write the code, fix the bugs, review the PR, deploy it, and secure it. What exactly are we learning in 4 year CS degrees?

We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We'll begin restoring access tomorrow, and will share an update soon. We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.

System design skills, architectural design, prompt planning and execution, context management skills, verification techniques, and adopting the AI agent workflow as a personal cognitive extension of the mind. I think the real gap will show in the people who have a knack at communicating and prompting the AI, and people who might struggle to articulate their needs and goals in a manner the LLM can use to generate the output.

Everyone has AI now. That advantage expired fast. What's the new advantage?

As someone who is neurodivergent, I struggle with this a lot. A combination of ADHD and other mental health obstacles makes time management and time awareness a very difficult task. That hyperfocus quality gets me so engaged in what I'm working on, and threads pan out like Indra's Net. I do have that ruthlessness with time, I'm very time sensitive and if something wastes my time it frustrates me. It would be nice to see more support for this type of cognitive profile in work environments and elsewhere. I don't mean to be difficult or hard to approach, I'm just so lost in the labyrinth of my own mind sometimes that I forget to reach out and touch grass sometimes and connect with others.

Across thousands of meetings, we found that people who do great work are often flaky and unreliable with normal life. Last-minute dinner cancellations, not showing up to events, being unreachable for weeks. We met one founder who rescheduled a journalist interview 12 times. Reliability is often a function of spare time, and these people are so consumed with solving an important problem that anything deemed unimportant doesn't get attention. They are subconsciously ruthless with their time. The common advice is: "be more available", but this confuses people pretending to be busy with people putting 100% of their mental bandwidth into changing the world.

I love the garden analogy and tending the garden. I've been using the Conductor instead of the Musician analogy, we've gone from being the musicians contributing to a part in a symphony to the conductor leading the entire ensemble in the particular movement being played. It's redefining the landscape in which we operate.

Coding before Ai was like building a Lego set. You start by gathering the pieces. Put together brick by brick. Plan build connect assemble. Coding after Ai is more like growing a garden. You start with a seed of an idea. Fertilize with compute. Tend prune cultivate evolve.

You mentioned that you don't have a visceral disgust reflux to AI writing, my sentiments exactly. I actually enjoy reading AI generated content, there is a certain homogeneity and structure to it that my brain finds pleasing. However, I agree, in certain domains it falls short. I've been exploring at using AI content to shape my own voice, and try to bring my own personal narrative to the generation. It's still not perfect, but I've been proud of what I generate. I don't think AI generated content is necessary bad, but how we frame that content and shape it with the tools we have at hand will determine the efficacy of the content.

The literary quality - or lack thereof - of AI only grows more noticeable over time. Unlike some, I don’t have a visceral disgust reflux to AI writing. I find it readable, at least outside novelistic contexts. It’s a competent technical writer, for example. But in scenarios where the prosody itself matters, AI falls short, and more importantly doesn’t really seem to be improving significantly over time. Can this really be just a question of writing not being a “well-defined” domain? I don’t think so. It might be here that some fundamental flaw in how we currently construct AI is becoming apparent

@TTrimoreau Markdown, lol. Seriously, I use Markdown more than a programming language now. I'm seeing it consistently with others, Markdown seems the perfect way to communicate and leverage AI agents even better, especially when combined with memory and other enhancements to the agent.

What’s your favorite programming language today?

I hope you don't mind me dropping a promo link, but I've been writing about building in public and the work I do, but I use AI to shape and deliver the content. All the reflections though, and the content, is driven by personal reflections and narrative, so the posts are told from my perspective. If you ever have time for a read to check it out, I have stuff up on hearthandcode.substack.com at the moment, and looking to expand to Medium. Like I said, I hate self-promoting, but I'd deeply valued someone with experience in this area to consider how my approach my aid someone with neurodivergent traits become more of a content creator and a writer by leveraging some of the offload of task completion to the AI agents.

Once in a while I read something that has the syntactic smell of AI all over it, but then I do my habitual "second read" and it turns out to be actually deep. It's a rare treat when this happens. Like it says "It's not X—it's Y" but then brings the receipts to show that X is widely believed but Y is actually true. It's even rarer when a writer is able to consistently deliver AI-assisted writing that has this quality. I've had the privilege of having a few incredible students in my classes who were able to use AI this way (all of whom are non-native English speakers). I don't think AI assisted writing is inherently hopeless and I'm hopeful that more people will learn to use AI effectively and responsibly, even for tasks like writing where the default way of using it is deeply problematic.

I should really try codex. I've been looking at other agent harnesses, particularly Hermes and OpenClaw, and have been leveraging cheaper models and even local models to try and deliver some of the same output quality with those models. I've been finding that orchestrating a Mixture of Agents combined with CoVE techniques (Chain of Verification) leads to noticeable improvements in generation (at least from Claude's analysis of the output). I love that Codex and Claude exists, but I want to be able to have more control over my data and not be dependent on a single service provider.

This just does not make sense to me, Codex is amazing on the front end if you just spend time learning how to talk to it.

Absolutely! I've been trying to tackle this on my own because this is a growing concern of mine. I've struggled with education, did well in K-12 but my ADHD really hit me hard in college. I've dropped out of multiple masters programs because the traditional academic structures weren't designed for the way I think. I believe that companies can come in and fill the gap where the government fails. My initial goal was to design an education platform for people to get undergraduate and even doctoral level education using AI agents. I've expanded a bit, but I always circle around the education issue - how can we use AI to enhance educational outcomes and start to correct some of the deficits and gaps we have in our educational system. Would love to stay in touch and exchange ideas, we definitely need a shift in perspective on how we approach education in the US.

the destruction of American education over the past decade is an incredible self-own competence is objective. a child can either do the math or they can’t. but in the u.s., a lot of people have reasons not to say that! parents don’t want to hear their kid is struggling. teachers don’t want scores used to manage them. districts don’t want embarrassment. progressives worry accountability will create inequity. conservatives don’t want federal authorities. so we end up with process, weak standards, and excuses to explain away bad outcomes. people object that it’s phones, covid, demographic change. ok! but we fail globally when others have phones, covid too — vietnam is much poorer than the u.s., yet performs well in international math comparisons. some countries treat math as a basic skill everyone needs to master. here, it is part of a fight about fairness, autonomy, and feelings in the age of AI — if people can’t do basic math, read closely, or think through problems, ai won’t make them more capable. it will become something they rely on without understanding. the countries that come out ahead in the global race won’t just have better technology. they’ll have people who know how to use it, question it, build on it. we need the national ability to decide something is worth doing coherently (teaching math!) the US has the money to teach math well but it has not shown the will. we are failing the next generation

@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?