joshgachnang

How are you automating all the parts? OpenClaw/Hermes?

As low as $69/month through the compounding pharmacies.

I didn't read this as "pair program every line".

I write a bunch of features. Most are pretty boring. A junior isn't going to learn much by pairing. But occasionally, I do big architecture changes. Those ones are perfect for early collaborative design, pairing, and discussion. The whole team benefits from understanding the architecture better and juniors get to see how seniors think about it. Then you can pair with juniors on the prompting or, more likely, the implementation plan to hand to the agents. That's maybe once a week.

Hey, I love the idea and the fully integrated approach! The feedback loop sounds great, and overall far better than the pretty generic companies I've dealt with for GLP before.

I've led building an EHR totally from scratch at one company and currently advising another as they build their EHR, both totally integrated approaches in industries where you usually have lots of disconnected care (adolescent serious mental illness and ABA/autism care). The state-by-state regulatory challenges are very real! If you want to chat, my email is in my profile!

And James Earl Jones!

The best class I had in high school was American Dream: a mixture of literature and history/social studies. You'd learn about a time and then read a book for that time, like Grapes of Wrath or The Jungle or All Quiet on the Western Front. It was a double period, had two dynamic teachers (one old hippie and one young fun guy), and most of the grades were from papers about how the book and history related. More schools should have something like that.

BugBot in Cursor has been great, I rarely find myself looking at it and disagreeing, especially with medium/high impact. I also have a custom architecture reviewer that just runs Claude, that's been great for finding duplicate code, boundary breaking, etc. We still have Github Copilot reviewing because it will (very occasionally) catch something BugBot missed, but the false positive rate is wayyyy higher, to the point I've considered disabling it a few times.

This assumes the police are always upstanding and never have ulterior motives. Which has proven not to be true. [0] and tons of others

[0] https://www.kcra.com/article/new-bedford-police-officer-on-l...

I've been working on the same thing for my ReMarkable, but I get it uploaded as a newspaper doc every morning since you can't install apps on it. Overall it's been working reasonably well!

This looks pretty awesome, like almost exactly what I've been hoping for. Is it possible to run the sessions on a remote server to keep tasks running? I mostly work from my laptop and it's nice to fire off tasks to a remote server (mostly been Cursor Cloud Agents lately) and close my laptop, maybe check in from the app later.

They escaped using the package proxy cache. I'm not seeing how a VM would have prevented that.

How would a VM have stopped any of this?

I've run into a bunch of segfaults in Bun whenever I went slightly off the beaten path, probably from the "horrifying" Zig code. A huge test suite + a clean rewrite with a path towards more safety is how I would go about it too, personally. Now whether I would do that rewrite in Zig or Rust, I'm not sure. But I have high hopes it's going to work a lot better, especially since Anthropic uses Bun heavily.

Are LLMs better at writing Rust than Zig? I'd assume so, there's way more Rust open source code than Zig. And if so, that's a very solid reason to switch IMO.

I used Stelo from Dexcom with my Oura Ring, no prescription. I didn't really learn much, other than occasionally realizing I was cranky because of low blood sugar.

And yet the rest of the party falls in behind him.

Just setup my team with Cursor. They actually got back to me on an enterprise plan (Claude keeps ignoring me). Cloud Agents have been great for keeping multiple streams going at the same time. Adding in computer use has been great for actually testing out features and showing they work for PRs. Bugbot so far has been the best AI reviewer I've tested. Composer 2.5 is great, though still using Opus for planning.

I can do most of this with Claude Code, but there's definitely a cost in maintaining it for the whole team.

Car payment, insurance, parking, gas/electricity? Going to be over $400/mo in almost all cases in any of the cities Waymo is in.

This overall is pretty close to how I've set up my implementation skill. One thing I'm curious about is how well the analogies like "We don't make dinner in a dirty kitchen." work vs something a lot more straightforward. Any input OP?

Nice, this looks exactly like what I've been looking for. I tried Fly.io Sprites and it _almost_ got me there, but I got annoyed logging into my CC every new feature. Unfortunately I wound up going all in on Cursor Cloud Agents, which overall has been decent.

Amazing, this is exactly what I've been looking for a while now, especially the remote dev box part. I'll give it a try, thank you!

You could put a chainsaw in the front and get a robot Pochita! [0]

[0]: https://chainsaw-man.fandom.com/wiki/Pochita

I'm constantly annoyed how much trouble I have using a .health domain (looking at you T-Mobile). I can't imagine using this many subdomains off a .us.

Amusing that it's on a Tuesday again. TACO (Trump Always Chickens Out) Tuesday.

Not sarcastic at all! This is a novel way to handle something that's been in the back of my mind (managing a lot of agents successfully).

This looks absolutely wonderful. Is it possible to run against Claude remotely (e.g. on a VM?). Or should I ask Claude to add that?

A lot of us use software written by other people we have no reason to trust and we haven't reviewed - most of open source libraries.

My non-techie friends either barely notice Liquid Glass or go "ooo this is nice!". It has annoyed me on occasion, but I barely notice it any more. Much ado about nothing.

While Wisconsin was debating this, they also closed a bunch of DMVs and limited hours for other ones.

The WI constitution enshrines the ability to vote. So you may think it's silly and for 99% of people it may be silly, but if anyone is prevented from voting because there's not a reasonable way for them to get a license, their rights are being infringed.

As a reader without a dog in this fight, I appreciate the comment and having the options. It's pretty common on HN.

This is a really cool design, pretty similar to what I've built for implementation planning. I like how iterative it is and that the whole system lives just in markdown. The verify step is a great idea I hadn't made a command yet, thank you!

This seems like it'd be great for solo projects but starts to fall apart for a team with a lot more PRs and distributed state. Heck, I run almost everything in a worktree, so even there the state is distributed. Maybe moving some of the state/plans/etc to Linear et al solves that though.