I don't think irresponsible is the right word, but it has drastically reduced Bun's appeal. All the tools we use have a brand to them, and Bun basically changed their brand overnight to "reckless" in my eyes.
I'd guess you're maybe working on more established codebases then? I see this regularly with my clients. I work in sprints and they pay for my time, so they will often deprioritise engineering/robustness work if it's taking up time that could be used some shiny new feature they want.
Also it's rarely "I want new features and I'm happy for them to be buggy or sluggish", it's more "We want to solve this problem and we're not exactly sure what the solution will look like, so lets try stuff and accept it's not going to be perfect"
> The larger point is that while AI coding tools make it easy to pump out lots of code, it’s really (still) the quality of the code that matters
This conclusion is unsupported from the observations. The code makes lots of requests, has too much CSS, and 6 different logo formats. So what? Lots of real, production codebases have just as many warts.
Folks need to stop dealing in absolutes with AI coding. Code quality always mattered, and still does, in certain circumstances. In others, it's less important and speed of iteration has more value. That's still true even with AI code.
I haven't really used redis much so curious to hear your perspective - this seems the opposite to me? A materialised view is just taking the data I already have and rendering it in a different way to speed up my access patterns. It's easy for me to understand where its all coming from, and it's all directly mapped back to the source data so if things change I can easily understand why it might break etc.
For redis, it seems there's no "out of the box" way to take some data from my DB and cache it. It seems it needs to be hand rolled per query you're optimising, you lose any structural link to the source data (redis doesn't know about my table structure), and now I have another service I need to worry about. Or is it much easier nowadays than I am thinking?
I’d push back on this, I think people have a very intrinsic sense of what is valuable and often if you think it’s “perception” of value being rewarded, it’s just that you value something different than that person.
Even in performative scenarios, like say someone gets promoted at work over another person because they are a great “performer” and always make noise, whereas the other actually delivers - they’re being promoted because the promotion is defensible and legible for their superior. That is true value for them, just not to another viewer.
I think GPT 4.5 showed that there is indeed a practical limit we're close too. That was supposedly a high-trillions of parameter model that was deprecated almost immediately because it was slow, insanely expensive, and had questionable benefits over the smaller models. Though apparently the new Mythos and whatever GPT Spud is (if it wasn't 5.5) are back up in the high trillions.
They are fundamentally different, but people desire they be aligned. The public expects the economy to producing higher quality of life for us, otherwise what is it doing? And for whom? But whether it actually does so is a function of other things. That gap seems bigger than usual right now with AI and tech eating the whole economy.
The fundamental difference is that email you host yourself requires ongoing maintenance and expertise to work at a basic level, and people would rather outsource it.
AI inference is different. You get the outcome by passing text through some weights at the time you need it. There's no ongoing work besides training and releasing new models. If I had something that rivalled Opus 4+ I could use locally, I would switch in a heartbeat.
Something else I've been thinking about which makes the economics of AI weird: The more powerful you make AI, the easier you make it for everyone else to make AI. I probably wouldn't bother to train an LLM from scratch, but I'm sure if I spent a few days with Codex/Claude Code I could do it (like GPT-2 level) easily. Obviously the capital moat is massive at the moment, but in like 50 years that probably won't be true.
IMO env files are a bit of an anti-pattern, env vars should be set outside the context of your app so that your app doesn't need to care where they come from, it just uses standard env var APIs to read them. This means developers can choose how they load them and what tools they use to do so.
With that said, the problem still exists just one layer higher. If they are secrets, I use Infisical (https://infisical.com/) which is free and self-hostable, and let's you inject secrets at runtime via their CLI. Very handy for small teams.
On top of that, I use direnv (https://direnv.net/) with a hook to infisical's export command. This means when I 'cd' into a project, infisical runs and injects the secrets for my developer environment.
Everything that isn't secret just lives in .envrc and gets loaded by direnv, and you can just send those files however you want because they aren't sensitive.
Interesting idea and there's clearly something real here, but I am extremely skeptical this played out like the article suggests. The authors are hand-waving some pretty complicated actions... Like how did Mona access supplier pages and place orders? How did she design custom merch? When she held a Google Meet with another agent, what even happened there, how did she connect to Google Meet... does she have the ability to join a persistent session like that with voice?
Interesting experiment but the article feels like it's deliberately withholding those kinds of details to hype the capabilities a little.
I think Anthropic will launch backend hosting off the back of their Bun acquisition very soon. It makes sense to basically run your entire business out of Claude, and share bespoke apps built by Claude code for whatever your software needs are.