We've been using AI for about two years now, and while I don't believe it's stolen anyone's job per say, it has become invaluable to our workflows. For me at least the $200 a month I pay for the Claude Max account feels like a steal when you compare the capacity and efficiency gains it has given me.
This may not be the same for everyone. I run a software development agency. Every year we build a lot of websites, MVP's, rebuild legacy systems, and more. That is to say I usually work in a pretty defined product space. As DHH said at Rails World this year, we're proud CRUD monkeys, most days. This is a perfect space for AI coding tools like Claude Code to make one feel like Bruce Almighty answering prayers.
Our capacity and profit margins proved this as well. As a result we lowered our price for many types of projects. Projects we knew were ripe for AI assistance we offered at competitive, fixed pricing models that we haven't done in years. Pigs get slaughtered, as they say, so when the margin climbed, we decided to pass that savings onto the customer and hunt more deals rather than bathe in the newfound cash flow.
That mentality of recognizing something that feels a bit like cheating and doing the right thing by our customers also gives me the same pause around the tools that are enabling us to do this. That little $200 per month line item for Claude Max is conspicuously low for the value we get out of it. I'd be willing to say confidently if it suddenly cost $1000 per month we would happily pay it--heck maybe even $2000 or more. We might have to raise prices again, but pricing and timelines would still be a far cry from what they were for software building three years ago. Only truly novel problem spaces require a similar price point, but at that point you're paying for much more than the code itself.
We've become dependent on these tools not just in our coding practice but in our business model itself. I don't think we can go backwards at this point and should these LLM companies decide they need to become profitable, we'd have no choice but to pay their bill. Now competition in the space is fierce and I think that will keep prices artificially low for quite some time, but that doesn't change the fact that the risk exists. This is why I believe that open source models are a space that we need to explore and contribute to.
DHH's Rails World keynote this year was titled "End to End Freedom." This resonated me as someone that has built their business on building software for others. What most people are paying me for, is freedom. Sometimes that's freedom from a SaaS provider that's become prohibitively expensive. Hitching our wagon to too much proprietary tooling, especially tooling that isn't based in an open source technology, can be dangerous and present a hidden, existential threat to the business.
I learned my lesson about proprietary tooling before, when I chose a NoSql database called Fauna for a number of projects. Fauna was not a profitable company and propped up by venture capital. One day, they couldn't raise more money and they announced they would have to shut down. You had a few months to migrate before they pulled the plug on everything.
Fauna, not being based in an open source database tech like MySql, was not a simply change of hosts. Everything would have to be changed right down to the query code and all the glue in between. This was doubly bad if you bought into their cloud functions, which were similar to stored procedures that would all need to be re-written and re-tested. A nightmare of the highest order.
Profit is an ethical necessity if businesses depend upon you. For my business it is unethical for me to remain in a poor profitability and cash position as my customers rely on my team and I for mission critical software infrastructure. For me to one day just have to pull the plug on them because we didn't run a sustainable business would be devastating. Therefore while we never get piggish with profit, we also don't take it for granted.
I see a similar situation with LLM companies. In an effort to win a larger customer base these prices are set to get us addicted to their service. Addiction becomes dependence fast. I've already seen developers complain on X and other platforms that they feel like they've forgotten how to code or at least are resistant to the act of hand typing. That kind of dependence, to me, means we need to own the code, even if we use a service provider to run it.
This is why I've decided to invest some time working with Ollama to run local models like Mistral. I've also invested in some beefy hardware as the requirements of running the better models fully max out most of our personal hardware. I've also began running opencode as my agentic terminal tool over Claude Code. While Claude Code is an incredible product, it only works with Claude and I don't want that kind of lock in.
Open source matters, and we often learn this lesson, as I did, at the most inopportune times. We're all running on what feels like endless yellow brick road. It's an incredible time to be a developer, especially a freelance one or an agency owner that gets paid by your output. However, the price of these things has to reflect that value at some point. Ask yourself if all of the LLM providers raised prices tomorrow, how much would you be willing to pay? At what point would that exceed the value you are extracting out of it? That number could be scary for many of us.
This is why open source and the ability to self host our LLM's is becoming a necessity for businesses that depend on it. It should be our pleasure to use them and to help contribute to their survival, improvement, and proliferation.
