Mik

play hard💪, work light 💡

@karpathy How do you get it to run for 4 hours. Something like auto research?

- Drafted a blog post - Used an LLM to meticulously improve the argument over 4 hours. - Wow, feeling great, it’s so convincing! - Fun idea let’s ask it to argue the opposite. - LLM demolishes the entire argument and convinces me that the opposite is in fact true. - lol The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.

So much fear and foreboding about what AI does to entry-level jobs. I think Alex's half-joke about simulated jobs on @latentspacepod is actually an idea worth taking more seriously.

@levelsio Do you just have multiple Claude subscriptions for the multiple sites you own and operate?

Another great argument for running Claude Code on your VPS server and not your laptop is its battery use "Terminal" app here is all Claude Code sessions, ignore the Claude app here I have a MacBook Pro 13" M4 and with Claude Code running even on idle my battery dies from 100% to 0% in about 3 hours, it's insane Claude Code on server via Termius SSH sucks 20x less power for your laptop

@ryancarson @pmarca He literally says he is intrinsically driven and not external reward driven after this. How do you figure that out if you don't introspect? Maybe this is tomato toomatoe, but Ayahuasca does not equal introspection fwiw.

I watched this interview, and what struck me the most was that @pmarca said he has zero interest in introspection in his life. I feel the same but have always felt bad about it. I don't have the patience or interest in any sort of introspection on the past. I don't want to talk to a therapist. I don't want to write in a journal. It was interesting to see that Marc feels the same way. Made me feel less bad about it. Great interview @davidsenra.

@danshipper Build an ide for chats?

this is what my codex chat history looks like. we absolutely need a new UX for coding agent guis a few thoughts: 1. my usual workflow is i have one main chat where im doing all my work for the day and then separate windows for each individual issue / feature. once those are in good shape and reviewed, i send them to my main chat to qa on staging and deploy 2. the "main chat" paradigm is nice because i want one place to have a good idea of "here's everything i need to accomplish today, what's been done, and what hasn't"—but right now it's at the same level of prominence in the UX / priority as every other chat which isn't great. 3. the main chat "orchestrator" should also be able to easily know what's going on in other threads. i need to be able to reference other chats easily and also have each sub thread have some idea of what the others are doing by default (should be able to isolate if i want though) 4. it's pretty wild to be doing your n-th codex run of the day, type what you want, and then have the model be like "okay, let's see what's going on here, ahhh there's a repo, and it has X features..." as if it has no idea what i've been doing that day. seems like a huge waste of time and tokens, we need something better by default.

I've got that same feeling. I have not been this excited about technology since 1994 (~22 years), when I was an excitable teenager. I can't help but draw an AI parallel with the explosion of words on the internet. Below is a simple graph of the average no of words written per day in academic journals and U.S.-published books. The data is a bit spotty, but it's roughly scaled from ~10 million words per day to 50 million words per day in an academic journal. Something similar in books. By 2013, you had something like 10 billion words per day on Twitter. That is a 100x multiple. bulchanm.github.io/word-explode/ (build with some help from Claude)

I was up until 2am doing DevOps last week, on multiple days. I haven't done any work that resembles DevOps in 20 years, and I can't give you a clean explanation for why I couldn't stop. Earlier that day, something had shifted. I'd been setting up software on my machine and managing configurations I'd normally hand off to someone else. It took longer than expected, and there were moments where I was certain it wasn't going to work. When it finally clicked, I felt something register in a way that reading about it never does. The capabilities of my computer had become just a prompt away. I was building something real, moving faster than I normally move while solving problems I'd have needed help with a year ago. That feeling, the feeling of being capable in ways you weren't before, is what kept me up doing infrastructure work I'd happily avoided for two decades. That's what I want to talk about. We're in a moment where creation has been democratized in a way that's genuinely new. Not the idea of it, but the actual reality of it landing. Building software used to require understanding software. Design required years of training. Creating almost anything at scale required either rare skill or access to people who had it. Capability was the bottleneck, and it was a real one. The gap between idea and execution was wide enough that only the most capable or most patient ever closed it. That bottleneck is dissolving. Today, one person with enough imagination and curiosity can build what would have required a team just a few years ago. The gap between idea and execution has narrowed in a way that most people haven't fully processed yet. When I was up at 2am, I wasn't thinking about any of this. I was just building. But the next morning, I started asking a different question. When creation is this easy, where does the bottleneck live? The bottleneck moved. When technical capability was the gate, ideas got filtered by default. Not every idea got built because not everyone could build. The selection process was imperfect, missing ideas that deserved to exist while letting through plenty that didn't, but there was a process. The friction of building was doing work on its own. That gate is opening now, wider than it's ever been. The question that used to be answered by capability, the question of what actually gets made, now has to be answered by something else. That something else is judgment. The ability to decide what should exist, who it's for, what problem it actually solves, and whether it needs to be in the world at all. When the bottleneck was capability, you could be a mediocre thinker and still make something significant, because making things was genuinely hard. The difficulty filtered for you. When the bottleneck is judgment, that cover is gone. What you create matters more than it ever has, precisely because creating stopped being what separates you from everyone else. I've been running experiments inside a project I haven’t announced yet. It's an AI lab where I build work things I want to see exist. Some are working well while others are still finding their footing. What I've noticed is that the experiments that feel most alive are the ones where I had a clear conviction about what I was building and why before I started. The ones that feel hollow are the ones I started because I could, not because I had a strong sense of what I was building toward. AI will do whatever you ask. That's what makes it powerful. It's also what makes it easy to spend a lot of time building things that probably shouldn't exist. The founders who will make the most of this moment don’t need to be the most technically proficient. They're the ones who can answer, with more clarity than anyone around them, what they're building and why. The ones who've done the hard thinking before the prompting starts. There's a version of this shift that looks like pure opportunity. And it is. But it arrives with a specific kind of pressure that's easy to miss. When everyone can build, what you build becomes the differentiator in a way it never had to be before. When everyone can design, taste becomes the advantage over skill. When creation stops being the hard part, the question of what to create and whether it's worth creating moves to the front of the line. I was up until 2am doing infrastructure work I hadn't touched in decades because something had shifted in my ability to create, and that shift made it impossible to stop. The breakthroughs I was having felt real, and they were. The more important breakthroughs, though, go beyond the ability to create. You have to decide what to create, stay honest about whether it should exist, and then build it with that clarity intact. Creating got easier. The harder question, the one that matters now more than it ever has, is what to create.

@firstadopter Probably more tokens on Claude code. But more time on gpt through my phone and browser

What model do you use most?

I haven't been this excited about tech since 1994.

@nikitabier @toly What's a more important tech: GLP-1 or GPT? @profgalloway on @myfirstmilpod has a take...

@toly AI is ozempic for the brain

we are all equally intelligent now

Claudito's been busy

@drgurner Looks like the shire out there! Makes me want to go for a stroll 😃

"Those who don't believe in magic will never find it." - Roald Dahl You have to be a believer in what you can do, and what is possible. Personally, I can't wait for the fields to be green again. Here's the view from one of the buildings on our farm. Magic is everywhere.

@DukeZer0 @fabknowledge @latentspacepod thanks @DukeZer0 for catchin that. Fixed Here podsnips.com/s/txJUh6hX9m

@rationalyst @fabknowledge @latentspacepod "tactic knowledge" *tacit knowledge

This take on Bloomberg and Excel by @fabknowledge on the @latentspacepod was so interesting...

Mis en place, for life.

@danshipper Zagging is a good skill!

if you're brainstorming with a group of humans and claws, you'll often find the claws circling around the same options over and over again teach them to Break the Frame: notice when you've circled the same ideas a bunch of times, and in those situations try the opposite of your current approach

It's still early!

@AravSrinivas Just do it - nike I'm feeling lucky - google do the right thing - ?

Just Do It

What has Perplexity been up to last two months? We've silently been working on the next big thing: Perplexity Computer. Computer unifies every current capability of AI into a single system. Files, tools, memory, and models, orchestrated together, working for you.

Maybe something UBIesque that pays out the delta over their working life between the 25k and the 120k? That would be a huge cost for sure, but still positive expected value if the machines expand the production function by multiples. Yes it's a bit dystopian but it won't be the first time in history there is a big realignment in our civilization work patterns. Something UBIesque may be a more civilized way to help us transition.

My genuine question for everyone dismissing these posts as “AI doomism”: what are these people going to do for income? And it’s not just truckers, it’s the white collar jobs, too. I’m yet to hear a good answer for this

@beffjezos And Gemini is the dark Unicorn 🦄 😀

OpenAI is a closed AI lab Anthropic is often misanthropic X AI is solving Y we’re here 🌌