Why depend on one API provider?
In principle, there can be an OpenRouter-like product for any API.
For example, a screenshot API router, an IP-to-location API router.
And then there could be a SuperRouter, a router for any API. Balances pricing, reliability, and quality for you.
It is likely less about generosity or company philosophy, I suspect OpenAI is clearly winning in marketing now by just counter-positioning to Anthropic and capitalizing on the contrast.
OpenAI would probably do the same as Anthropic if they were short on cash. If so, it might be a war of attrition on capital.
> Anthropic reduces limits, OpenAI increases them.
> Claude subscribers can no longer use their subscription with OpenClaw, but OpenAI acquires and encourages its usage.
And so on...
I enjoy business games. Damn, I enjoy them! Especially observing how companies compete.
VCs and CEOs who tried Claude Code/Codex to build a prototype for two weeks, didn't deploy it to production, abandoned it, and never faced the consequences of the technical debt produced:
"Software is solved and infinitely available. AGI is just not evenly distributed yet."
If LLMs were invented around 2000, would they still be generating websites in Perl or would we have reached TypeScript and React much faster?
I wonder if LLMs just lock us into whatever stack they are trained on.
I wouldn’t have been able to do that in the past without AI. That is really something.
My biggest productivity gain with AI is not at my peak performance. It is actually the opposite.
At my peaks, AI somewhat distracts me.
But at my lowest, when I don't sleep well, am busy, or lack energy, I still manage to fix issues, ship improvements, and close support tickets almost every day.
My close friend: “You should hire and delegate.”
Me: “Why?”
Him: “To have fun. Or if you can’t, just take that offer and sell your product for $1,000,000+.”
Me: “And what’s next?”
Him: “Have a lot of fun and you can build whatever you want.”
Me: “But what do I do now?”
😆
And I fail to answer that in every imaginable position. As an entrepreneur, as a developer, as an investor, as a customer…
Who cares?
I am open to new ideas, but I ask myself: “What would make me care about the number of lines of code I generate?”
I care about my customers success. About error rate, uptime, onboarding, marketing…
What worldview should I have had for the lines of code to be important?
It is counterintuitive now, but marketing might get easier.
A product that just works, doesn't have critical issues, and has decent performance will already stand out.
I am fine to default on tech debt for one-off scripts and marketing/internal tools. I ship them based mostly on vibes.
But no token seller affiliate will convince me to vibecode critical paths.
For production with paying customers, I reread and often refactor every line.
If everything can be automated now, automate what you find boring.
Keep what you enjoy doing.
That’s your edge. Why outsource it?
Reading Jung with ChatGPT and how memory is injected into our conversation is something:
"If you want, I can help you articulate a non-cringe, non-marketing-bullshit 'myth layer' for ScreenshotOne that actually aligns with how you think."
And you know what? Yes, I want 🤣
1 dimension is having a job and living alone.
Entrepreneurship is 2 dimensional space. You see what the job was about and a few more nuances.
Having kids is living 4 dimensions. Those who understand will understand.
Enlightenment is living outside dimensions at all.
I wish I could outsource my thinking to LLMs.
If only they actually thought.
We watched a movie yesterday, and I realized how fulfilling it should feel to create something completely finished.
I never have this feeling when building software. There is no cut line. My backlog is infinite. Even for side projects.
That frustrates me a lot. Often.
5. Do not expose any environment variables or critical data access in the code that runs browsers.
6. Keep the browser version always up to date.
puppeteer.guide/posts/sandbox/
1. Always run browsers with the sandbox.
2. Do not disable site and process isolation.
3. Do not run browsers under a root user.
4. Use security profiles with minimum allowed operations.
And:
Wanted to write a short note about security best practices when running Chrome/Chromium but ended up with a long-form blog post, as usual 🤦♂️
If you don't have time to read but run headless/full browsers for automation or as part of your product:
I am not afraid. But I am likely losing the upside of risking too much.
However, I just want to keep doing what I love for decades. Money becomes secondary after a certain point.
If you have found what you care about, you are already lucky. Make sure not to lose it because you were unprepared.
I assume inflation, grocery prices to rise.
Wars may last longer than we expect.
AI-related revenues for my product may decline, with increased AI expenses.
I prepare.
I pay for my domains years ahead.
I keep extra cash for servers.
I invest in ETFs and bonds but also hold cash.
You won’t find a person more optimistic than me.
Peace, AGI, and prosperity will come, and humans will figure out everything eventually.
But optimism doesn’t mean being careless.
The beauty of engineering is that you can’t fool yourself forever; at some moment, reality will kick in.
I feel there is a short window of opportunity until companies start to default on technical debt, AI psychosis fades, and the limitations of LLMs become obvious.
If I wanted to bet my money on skilled software engineers being in higher demand in the future, not lower, how would I do it? What stocks would I buy?
There is a huge difference in sentiment and language between people who understand how LLMs work, their fundamentals, and attention parrots who want to cash out fast on the current hype wave.
Pessimism is irrational.
If now is a great time, enjoy it.
If you think you are in bad times, it can always be worse. You are fine.
But if now is the worst time, then we are good, because we are touching the bottom, and only up from there.
Everything will be fine.