Caleb Andersen

Building @nativshark_ to make learning Japanese simple✨

[Caleb Andersen @CalebAndersen]

@dannypostma @jessethanley @marckohlbrugge @grok hehe. you can do it ^^

when you're on linux you can just use your local agent and say "i want to do thing" and it'll code your computer to do it pretty much.

ive used weylus (pen input needed), moonlight (just stream) to do it, wayvnc would also work probably.

[Caleb Andersen @CalebAndersen]

@jessethanley @dannypostma @marckohlbrugge @grok i mean is you might as well right? get 64gb+ ram tho ofc

[Caleb Andersen @CalebAndersen]

@jessethanley @dannypostma @marckohlbrugge @grok i cant actually remember since its been so long since i benchmarked it.. maybe it was on par w/ the M4. anyway, fits in my handbag and its fun to use:) pair it w/ a small split keyboard if you want. dock it wherever.

[Caleb Andersen @CalebAndersen]

@jessethanley @dannypostma @marckohlbrugge @grok Ye, i do all my dev work on it haha.

I run my full app, postgres, opencode harness w/ several instances, subagents with worktrees, nvim, zen browser w/ bunch of tabs, chrome for playwright e2e, discord, spotify.. in terms of power its like somewhere between the M1 Max and M2

[Caleb Andersen @CalebAndersen]

@kentcdodds example from this situation? 👀

[Caleb Andersen @CalebAndersen]

@thdxr how are we feeling about setting a "compact at %" in OpenCode? being configurable per model would be wild. right now in v1 you can reserve tokens, in v2 we have a buffer, but no beautiful % option😌

[Caleb Andersen @CalebAndersen]

@dejiolowe @tylertringas unlikely.

[Caleb Andersen @CalebAndersen]

@Suhail are they prompts you're re-using? homie write some commands / skills. idk what harness you're using but all of them have some sort of "disable-model-invocation: true" frontmatter capability in their skill files.

[lauren @poteto]

if your definition of coding is just the act of typing code, then yes, it's solved. agents can type code faster than anyone can.

but i think most of us think of coding as the act of solving problems with code. in that definition, coding is definitely not solved. agents can generate code quickly and build whatever you want, but without the right constraints, they end up causing more problems than they solve.

i think of this like my gardening analogy. agents are like magic beans that grow in any condition, but in wild and unpredictable ways. if you put the right scaffolding around them however, they can grow into whatever shape you want.

you still want humans thoughtfully and intentionally steering this outcome. we have a name for them: engineers. agents have greatly lowered the barrier to entry to doing engineering. what was once the domain of those who understood syntax and logic is now accessible to anyone who can explain their ideas. english is a new programming language for engineering outcomes with agents.

now, even ordinary people can build impressive software. and the good ones will learn engineering through first principles. as someone who is a self-taught programmer (i first learned how to code through Excel macros!), this is incredibly exciting to me, and i can't wait to see what everyone learns and builds in this new world.

let a thousand gardens bloom!

[Caleb Andersen @CalebAndersen]

@blako87 @jessethanley @OmarchyLinux @opencode there's nothing different about it other than the config, or in other words, everything is different. linux is always just linux, thus the config is everything. omarchy is just a nice tasteful well maintained wrapper on arch, but its still arch, which is "still linux".

[Caleb Andersen @CalebAndersen]

@jessethanley @bossriceshark @OmarchyLinux @opencode i been using everyday as my main work driver since i bought it last year:) it's so good. the portability is undefeated, literally fits in my handbag.

[Matt Pocock @mattpocockuk]

"I do see some parts of strategic coding being automated"

I agree partially. Lots of folks are building automated pipelines like:

- Bug report -> repro -> fix
- Feature request -> prototype
- Cron -> codebase architecture RFC -> refactor

But managing these pipelines is very much a strategic concern, one that a human needs to apply their judgment to.

So this is not proof that agents can think strategically. It's a way of arranging their tactical abilities more sensibly.

Another point is that it's really, really difficult to get these pipelines right.

So the phrase "coding is solved" feels incorrect on two counts:

- Strategic programming is nowhere close to being solved
- Efforts to make agents good at strategic programming are really tactics in disguise

[Caleb Andersen @CalebAndersen]

@thdxr This is the take.

[Caleb Andersen @CalebAndersen]

@ow @dmcgco this is the way

[Caleb Andersen @CalebAndersen]

@dmcgco @ow set up filters / folders in the one inbox and its still fine tbh.

[Caleb Andersen @CalebAndersen]

@DThompsonDev just markdown files, then open in your markdown viewer of choice. Obsidian is nice. Very customizable if you want, or simple if you don't. But just markdown, so you own / control your notes fully.

[Caleb Andersen @CalebAndersen]

@jessethanley https://t.co/KT66jPe39K

[Caleb Andersen @CalebAndersen]

@thdxr @fjcero @boristane i like the plan agent as it is now, but if it got any heavier it would be annoying. ask would be great tho.
ask, plan, review, build are the only agents needed imo.

[Caleb Andersen @CalebAndersen]

@jessethanley @PlanetScale @crunchydata @Redisinc db to planetscale or crunchydata, self host the rest on vps(s), ideally in at least two different DC, use coolify/dokploy for a decent UX managing your apps / services / containers, save a lot of money and learn a bunch in the process.

[Caleb Andersen @CalebAndersen]

@jessethanley @thdxr still not gonna make me use it