mbriggs

@juristr I just got one too, great machine :)

Might be slightly distracted today 🤩🤩

@Govindtwtt Everything gets cheaper due to lower overheads -> this creates new categories of jobs we haven't had before

Everyone says “AI will take all the jobs.” If that happens… how does this future actually work? No jobs → no income → no spending. So who buys things? Who pays rent? Who keeps the economy moving? What am I missing here?

@aryanlabde I would go further Vibe coding is far harder, and requires much more skill then traditional coding.

Hot take: Vibe coding only works well if you already know how to code.

@unclebobmartin Fair and I totally agree, have you seen stuff like this going around? jamon.dev/night-shift It feels right to me, but its also really making me think of waterfall :D I have also seen the term "Light Waterfall" get bandied around a lot

I work in iteration, like any good agililista. A bit of planning, a bit of executing and testing, rinse and repeat. Plans are important. Overplanning is suicide. And, besides, AIs can't follow a plan any better than a human. So the old adage remains true: No battle plan survives contact with the enemy.

@unclebobmartin I would be curious as to your thoughts on how it feels like what most of us are landing on with agentic work is basically waterfall reborn -- Get your specs in a great spot, whiteboard, guardrails, etc, build after you have all of that in place. It feels weird to me, I can't imagine what it must feel like to one of the agile manifesto people.

@Steve_Yegge I first saw your ai adoption level scale last year, I was somewhere around L5, and I was pretty skeptical. Then I went to L6, L7, and im currently putting some polish on my orchestrator as I type this. Just want to say you were right, even if i still think gastown is a bit nuts and beads is overworked :-)

I honestly think every dev should probably build an orchestrator. I think it may be like building a lightsaber for a jedi. If you don't know the nuances of what you want or how to do it, you shouldn't use one. If you do know, you shouldn't use someone elses.

@GeoffreyHuntley would you put your name on <thing>? think hard, this is important.

this is my favourite prompt of all time: “how could this be better?” reply with yours and why it rocks!

@unclebobmartin I think its a pretty common thing where you dont move faster, you do more with AI. You don't delay refactoring, you don't leave restructuring for later, you do everything you should be doing as a professional at the same time as the implementation. That has compounding value.

The name of the my project is empire-2025. That should tell you that I've been working on this project for three months. And that work has been intense. Not quite full time but probably 25-30 hours per week. To complicate matters I've used this project to experiment intensely with all kinds of ideas. From little refactorings, to major architectural changes. From crap analysis, to heavy duty mutation testing. From acceptance test parsing, to heavy duty headless monitoring runs. All that experimenting takes time. A lot of time. A lot of _computer_ time, and a lot of AI time. I often spent hours simply babysitting the progress of these experiments. And so the question arises. Given three months, would I have achieved an equivalent game by simply coding it myself. Probably. That's about how long it took me to write spacewar, and the two games are roughly equal in complexity. But I wouldn't have learned much. And that was the point of this whole exercise. I wanted to learn. And I'm still learning. What if I hadn't been experimenting? In that case I would have declared victory two months ago. I had this system essentially running back in late January. So, using a back of the napkin estimate, and subtracting out all the experimenting I did, and assuming that it would have taken me the full three months to code by hand. I'd say the AI tripled my productivity. That's a very crude guess. What's more my latest experiments suggest that productivity declines as complexity rises. The bigger the project the lower the overall productivity will be.

@juristr I think of L6/L7 as the pain zone, L8 stuff starts getting sustainable again :)

We're cooked. 90% of ppl responding are > L6 😜

@trikcode You are missing the vc reinforcement cycle here. Try to raise without using aws gcp or azure for infra, even at an order of magnitude higher cost with marginal benefit to many use cases.

The tech stack of most startups: - Frontend: React (because everyone uses React) - Backend: Whatever the CTO knew from his last job - Database: Postgres (or Mongo if someone watched a YouTube tutorial) - Auth: Copied from a blog post - Payments: Stripe (obviously) - Deployment: "It works on my machine" - Documentation: LOL - Tests: "We'll add those later" Valued at $10 million.

@nalinrajput23 stability. I ran gentoo in my 20s, arch in my 30s, and debian stable in my 40s. Debian _never_ breaks.

Linus Torvalds still runs Fedora with GNOME, not Arch. Now I’m genuinely curious what makes him stick with that choice?

I spent enough years using TUIs professionally that I can can say with certainty, the usecase for things like tmux is for things like persisting remote session, and thats it. I switched years ago to kitty (and then ghostty) for multiplexing and said good bye to an era of pain.

I have come to the conclusion that either the machine can't autonomously design well, or I have a skill issue. I have had amazing design conversations with it, and i can reference ideas like afferent / efferent and general / specific without explaining. But for the life of me I can't get it to come to good conclusions without direct collaboration, or even be able to flag structural issues without listing every single thing i want it to analyze.

Juniors entering the field will still need to understand what code is. But they won't need most of the philosophy that we've been used to. The emphasis will all be on pragmatics and engineering. So they won't need to know OOP, but they will definitely need to know dependency inversion. They won't need to know functional programming, but they'll definitely need to understand purity and the costs of mutability. They won't need to know about structured programming, but they will need to understand modularity.

For all the hype and hullabaloo about AI and vibe coding; this is not a layman's domain. The engineering bar is going way up, and getting over that bar will not be easy.

@FlyaKiet @superset_sh its more an ergonomic thing. if you split a lot usually its fine, sometimes its less fine based on some kind of output. or maybe its fine on desktop, but not on laptop, etc. Not a critical thing, but its a thing from tmux i miss when its not there :)

@mbriggs_dev @superset_sh Zooming is very interesting, have you tried righ click to move into new tab? It’s not ideal but could be what you’re looking for

@superset_sh been using it and really dig it, three requests 1. I would like to be able to have a bind to "zoom" a pane. basically make that take up the full tab with a visual indicator by the pane title bar saying its zoomed, press again, unzoom back to normal 2. ability to target that pane title bar with themes specifically 3. ability to remove previously imported themes

SWE Maturity Model 1. Write code / build software 2. Get agents to get good results for a particular aspect SDLC 3. Write a skill to reduce the effort and increase repeatability of a particular aspect of SDLC 4. Automate (or leverage a harness automation) to leverage a skill to get good results for a particular aspect of SDLC