Labor is Now Free. Make Something.

Within the past few weeks, something changed in AI. Every release is better than the last, sometimes in obvious ways, sometimes in a way that feels like placebo. But this time, with Anthropic Fable, it’s crossed a line into something that feels different than previous releases.

The best way I can describe it is that Fable now feels like a full-time employee who is better than me at nearly everything, works harder than me, and can accomplish about 20x what I can in a day. It’s actually probably more than 20x, but I’m trying not to be hyperbolic.

I remember when Claude first came out. I was on a Norwegian Fjord cruise with my friend Michael and we were so excited that we could copy and paste code snippets into Claude, it would suggest some changes, we’d copy and paste them back in the right places, and sometimes it would work. It was amazing.

Now I wake up, look at my projects (because I gave Claude some instructions as I went to sleep and I want to see how they came out), give it a to-do list of new things to do, and then move on with my day. I make tea, I read my email, I scroll through X and Hacker News. I check back with Claude. It’s done three of the five things I told it to and it has found some other bugs (with my original code). It writes tests, fixes them, takes screenshots so that I can see, pushes the changes to a Pull Request in GitHub, and keeps going through the list.

There was a CruiseSheet feature that I avoided doing for years because I thought it was probably impossible, but at the very least would be miserable to implement and would take weeks. Claude did it in an hour. Sometimes I run out of things I want to do, so I ask it what we should do next. Its suggestions are excellent.

The funny thing is that its time estimates are clearly linked to human development. So it will give some suggestions and say, “We can do A B and C this week. Next week we can focus on X Y Z”. I say “Do All Six” and they’re done in 30 minutes. 

I’ve been a coder for 20+ years, and I like coding, but I never really loved the craft of it. There were moments of coming up with a clever solution to something, but the day-to-day building user interfaces and new features wasn’t always a dream. 

I have always loved coming up with ideas and building them, though, so for that reason I loved coding. What’s amazing is that now I get to do the 5% part that I love, and the 95% that I don’t love gets done for me faster and better than I’d do it myself.

Anyway, the main reason I feel some urgency to write about this is that if you’ve wanted to start a business, you are absolutely insane if you don’t do it right now. There has never been a better time to start a small business. Labor is free. This will become normal and commonplace 12-24 months from now, but right now no one really knows. You can build a useful and valuable business in a week if you have a good idea. And if you don’t have a good idea you can make 10 different things over the next couple months and see which one was the best.

You could probably just ask Claude to help you come up with a good business, and I bet it would.

But seriously, now is the time. You could make a web site, and app, a service, etc. You don’t need to know anything. Claude knows everything.

My Setup

Part of why I’m so excited is because I just happened to make a much better AI development setup right before Fable came back. There are probably a lot of good setups, but I’m surprised at how many people are still using Claude through the web or Cursor or something like that. This will be a little technical, but you could just paste it into Claude and it would explain it tell you how to set it up (or do it for you).

First, I have a virtual machine on a server that Claude owns. It can install software, edit config files, do whatever it wants. Each project has its own folder. In each folder I start a screen session, run claude –dangerously-skip-permissions, type /remote-control, and then detach the screen. In plain English, this starts a Claude session on that machine where it can have full reign. The remote control command allows me to control it from my phone, the web, or the desktop app. For each project I pin one or two sessions (sometimes I have one for development and one for asking it questions / analytics).

Then, whenever I feel like “working” I just open the Claude app, give it a list of stuff to do, and close the app. 95+% of the time it executes it flawlessly. The remaining 5% is usually some visual presentation I don’t love.

I also set up Tailscale, so that all of my servers and devices are on the same virtual network. So if I’m working on cruisesheet, I just go to cruisesheet.lan to look at the changes before approving them to be pushed to the site. When I approve them I just say “go ahead” and it runs hundreds of tests, stages it, and pushes it to the live site.

Like so many things these days, AI feels unnecessarily polarizing. People just love to hate it. A quick rebuttal of a few common objections:

“AI produces slop” Not really. It used to, but it’s definitely better at programming than I am. It’s also much easier to say “make CI tests for every major part of this project and also do a full security audit” than it is to manually do that yourself.

“AI is a waste of water/power/etc” Sort of a pointless argument. It exists, you using it isn’t going to increase either of those things meaningfully (also, it actually generally doesn’t use much water anymore). If you actually care on principle, use it to make a lot of money and then buy up some nature and put it in a conservation trust.

“I’m a craftsman and I want to look at my own code” I think this day has passed. Code by hand if you really love it, but you don’t see many people taking horse and buggies across the country anymore. The difference in performance between horse and buggy and a modern car is far less than hand coding and AI coding.

“But I’m not technical” This day has also passed. I’ve made two games for me and my friends, and they’re both written in a language I don’t know with a game framework I don’t know. I haven’t seen a single line of code. Claude also made a script to compile them for both phones and computers. I don’t really know how to do that either.

I also made a watchface for my new Pebble Time 2 watch. I don’t even know what language the face is in, but when I posted it on X the creator of Pebble retweeted it (as did Paul Graham!). So… probably not slop, and also it wasn’t really necessary to be technical.

At its best, AI is a force multiplier. It doesn’t eliminate programmers, it frees them from the miserable parts of coding and allows them to bring their creations to life far faster and more robustly than ever before. 

Soon everyone will have these tools and companies will package them up to make them even easier. But right now is a rare window where no one really knows how quickly and effortlessly you can go from idea to published product. It’s a rare opportunity to get way ahead while others are figuring it out.

Get a Claude subscription, and make sure you switch it to use Fable. The $100 plan is probably enough to make almost anything. If you run into the limit, you’ll be happy to upgrade to the $200 plan.

Show me what you make!

###

Photo was from the Seven Sisters hike in the UK. Really cool unique landscape!

Sorry for not posting more… still waiting to get the photos for my AT gear post.


Posted

in

by

Tags:

Comments

20 responses to “Labor is Now Free. Make Something.”

  1. BG Avatar
    BG

    I don’t get the AI hate either. Claude and ChatGPT have helped me create things that I never would have had the time or motivation to do on my own.

    Most recently, I used it to give Plexamp some super powers. I got annoyed that the Plexamp app’s Sonic Adventure feature 1) doesn’t let you search by Artist or Album to find a song, only the exact Track name; and 2) doesn’t let you save Sonic Adventures for listening to later.

    So I had ChatGPT create a webpage running on Docker Desktop on my PC, which is connected to music.[mydomain].com via DNS in Cloudflare (and password protected) so it’s remotely accessible.

    On this page, I enter in the artists that I want to appear in the Sonic Adventure, one per line, in a form field.

    The script then selects one of the top 10 songs by each artist, then orders the songs by difference to create an optimal Sonic Adventure (i.e. no similar artists/tracks adjacent), queries the Plex API to generate the Sonic Adventure using their secret sauce, then saves it to a playlist of my naming.

    Oh and after selecting and ordering the tracks, but before generating the playlist, it shows me the list of tracks and next to each track is a “Regenerate” button that I can click to have it select a different track from that artist (if it first selected a track I don’t like, or doesn’t fit the mood I’m going for).

    This has 1000x improved my listening experience for Plexamp. And now I don’t have to wait on the devs to fix the search or give me a way to save Sonic Adventures. I did it with AI and the existing Plex API in about 30 minutes. Didn’t even look at the code, just cut and pasted what ChatGPT 5.5 Medium spit out. And like you said, I asked it “how can we make this better?” and that’s when it gave the suggestion for the “Regenerate” button.

  2. Angus Avatar
    Angus

    I’d argue the total opposite. AI reducing the barrier to entry means you’re going to have 100x the competition in any software-adjacent business. This might be the worst time in history to launch an app idea.

    “You don’t need to know anything” isn’t a good thing.

    1. Tynan Avatar

      So would it be better if more knowledge was required and fewer people were able to create things?

      1. Guy Avatar
        Guy

        When did he say that? His point was entirely business-related.

        How does one guard against the forthcoming flood of low-quality software? What is one’s moat against the deluge of other, similarly-piss-poor “You don’t need to know anything”-types all shouting “pick me”? None of whom understand databases, technical debt, scaling laws, solid architectures etc.?

        Perhaps an already established business can use LLMs to increase their development pace a (fair) bit. But *starting* a business now seems to be a fool’s errand. Your comment of “but right now no one really knows” flies in the face of all that I’m seeing on the internet.

        My final point. AI software looks brand new and shiny on Day 1, but very little of it will look very good on Day 365 unless a fairly competent software engineer/coder is keeping it on a short leash throughout. Fable or no Fable.

      2. Angus Avatar
        Angus

        Depends on the perspective. It’s better for consumers and hobbyists. Almost certainly worse for anyone who wants to create a product to sell.

        Its never been a skill gate-kept within a secret society or expensive university course. It was always available for free to anyone who put in a little effort.

        The effort and skill required is what enables a profit margin to exist for a business. Consider dropshipping: it’s almost impossible to create a business around that long-term because it’ll be copied and undercut within days.

        1. Tynan Avatar

          But what if you put in the same amount of effort and get 20x the results. You are competing with others that do that, but the people that are not using AI get left behind, so you have less competition.

          I agree that very basic apps/sites become a commodity like dropshipping, but remember that dropshipping had its moment where a lot of people made a lot of money. That moment is now for AI driven development.

          1. Angus Avatar
            Angus

            > But what if you put in the same amount of effort and get 20x the results

            Yeah, that’s a good point. There are a lot of products where it previously wouldn’t even be feasible for a solo/small team to tackle because of how much work it would be

    2. Taylor Levinson Avatar

      I’m thinking it’s the middle path on this…

      To Angus’ point, a lot of things (SAAS, apps, other digital content and tools) will become a race to the bottom as barriers to entry approach zero.

      To Tynan’s point, this is all incredibly empowering, especially for the entrpeneurially-inclined…

      Seeing Ai as a “force multiplier” is the mental model that feels right to me at the moment

  3. Liface Avatar

    Now that labor is ~free, the barrier is now sales and marketing.

    So yes, build something, but without being able to sell and market it, don’t expect that many people will use it.

    1. Tynan Avatar

      Claude is also a marketing expert. I should have included non-programming examples in the blog post. I’d just ask Claude what to do. A friend is starting a new thing (made in one day with Claude) and Claude used the Google API to make paid ads for it.

  4. Michael Avatar

    Building is easy so now (and arguable before too) the hard part is the making a business out of it. I’ve seen a lot of data about explosion in (for example) new apps in mobile app store, but at the same time a drop in user downloads and reviews. How do you think we should go about with making new code into working businesses? How do you get people to care about Cruisesheet?

    1. Tynan Avatar

      It’s not surprising that there are more bad apps than before (and a higher %), but I bet the absolute # of good ones is higher too.

      Every business is different, but I’d just ask Claude to do a ton of research on competitors and tell me how to market it. It’s given me 10-20 suggestions for CruiseSheet (SEO, appealing to past customers, etc.) that have paid off.

  5. Joe Avatar
    Joe

    Just want you to know that this came at an incredibly serendipitous time for me. I’ve been trying to build a machine/ process/ product business while managing a manufacturing company full time. I’ve been getting nowhere and it’s time to leave the job. It’s all a bit sudden and I was freaking out because I didn’t have a plan (I don’t really do the LinkedIn career thing). You just gave me a plan. I’ve been tinkering with claude and VSCode, I know how to code etc, I’d considered it, but I’m taking this as a sign. Thank you man.

  6. Ivan Avatar
    Ivan

    For the longest time I dreamed of having an easy way to organize my family’s meals in a set it and forget it way. It was never worth it to spend the time on it.

    Now I have a really cool site with AI assisted plans that I can easily edit and adjust. It even does online shopping (mostly, quirks in progress of ironing out).

    I’ve loved programming since forever, but the last AI generation has rekindled that love (on time too, as I sometimes started longing for a simpler life in the woods haha).

  7. Sean Avatar

    I made myself a RAW photo editor. Took about a week with Sonnet and Opus and I no zilch about coding.

  8. Mattias W Avatar

    Great article. I love this moment in time. I’ve also been programming for 20+ years and I’ve really liked it, but these new models just make it even more fun.

    Here is one project I started many years ago but haven’t had time to finish until now when I don’t have to type everything myself or even figure out how to do stuff.

    https://www.clikr.app/

    I have more projects in the pipeline too and there’s no problem working on them in parallel.

  9. John Avatar
    John

    What was the cruisesheet feature that you thought would be too hard that it could do quickly?

  10. David Avatar
    David

    I agree that if coding labor was previously a barrier to creating something, then indeed things have changed. Great if you want to create some new software for yourself, or just to create something for the sake of creating it.

    But as for starting a new business that is going to make money, that still requires some differentiation from the competition, who also has access to this free labor. Even if you are a creative genius who came up with a truly novel idea, it is now even easier for competitors to copy and imitate it.

    Not to say that there aren’t new monetary opportunities, such as:
    -Boosting an existing business through things such as improved tracking, improved analysis, or even cutting back on SWEs.
    -Avoiding licensing costs by creating solutions that were previously to complex/expensive to create in house.
    -Preparing to corner a new market, if you have things like as a large marketing and advertising budget, or another way to shut competition out.

  11. Jim Avatar
    Jim

    You seemed to get bogged down and stalled out with the technical development of SETT a few years back. Or maybe with the marketing? I wonder how that project would have turned out if you could have powered through with AI?

Leave a Reply

Your email address will not be published. Required fields are marked *