vitamark

Well, if your agent lacks shell access (or has some other sandboxing going on), it shouldn't have access to envs and MCP setup files.

(leaving out cases where your genius GPT-12 Galaxy Ultra agent hacks the sandboxing from inside)

While this specific rename looks dumb, the map providers deal with this kind of things all the time. Big georgraphical features like rivers, lakes, seas, mountains and also borders are different from different perspectives (mostly country-based).

From outside of the US, I see "Lake Ontario (Lake America)" in Google Maps, the same way as with the Gulf of Mexico. I see "Persian Gulf (Arabian Gulf)" and "Sea of Japan (East Sea)".

Some names are given from one perspective (like rivers and mountains), like Rio Grande river or Mt Everest.

I entered my birth date and this calculator said I will eat a leather shoe some time in the near future.

Very accurate, thanks, looking forward to that.

Can you shed some light on what's the deal with the new interface (web/desktop)?

I updated one of my opencode containers today and had to downgrade it back because:

1. Worktree/workspace support is gone altogether. No way for agent to work in parallel on different things safely except prompting it directly to create a worktree.

2. Seeing multiple sessions properly now requires opening tabs for those?

3. In general, switching between several projects/sessions/whatever stopped working, and command palette is hotkey-only.

Issues about the new design just sit in an endless unsolved pile on GitHub which seems hopeless. And an agressive stale bot is not helping with that.

I came to OpenCode because it seemed hackable and simple, and at this point it seems like a "move fast break a lot" project that you got to live with.

(rant over, sorry if this is addressed to the wrong person, just needed to vent a bit)

I don't see how reading has anything to do with internal monologue. Do you hear the words you read when reading? If so, it surely has to be slow as hell.

I did get relatively good scores in reading speed back in primary/middle school when we took such tests, so maybe there's an explanation. At the same time, writing/speaking is a very resource-intensive task for me so I tend to be silent a lot (and then drop a pile of sentences on the unsuspecting people I'm with)

I have no problem with AI as-is, but writing should be concise and to the point, and without heavy tweaking current models write pretentious hard-to-read stuff like this article.

A person who has something to say often has trouble stopping writing. Outsourcing writing to AI then feels like the opposite, as if the author doesn't care but wants to just spew some content.

I personally encode flacs as 192/256k opus from the start and that's fairly enough for most data save purposes, so no reencode for streaming is needed

that's a bitrate of 1GB per 9-12h, and for cases when it's too much I just have cached music on my device (I'm lucky to have mostly empty storage on my 256gb phone)

at least bun vibe-coded rust rewrite had a test suite (which doesn't justify it)

we're all very grateful for cool toolchains and tools but I think people wouldn't even poke a stick at it if it doesn't pass a huge test suite (that is not vibe-coded, probably nodejs test suite if they have one?).

No, I actually say that there are no words unless I think of them. Ideas appear out of nowhere and are made of nothing, like they just are integral part of my mind state

Well, I can't say for the general population, but for me there's just nothing unless I actively try to think of words, sentences and other things. One moment you don't have an idea and the next you do, like a state switch.

I really feel like humans argue too much about how computer programs think and at the same time too little about how humans think. The topic is criminally underresearched and the existing research is all over the place.

"For an LLM, it is exactly the opposite".

Is it really? LLMs don't have words inside, for the most part they operate by applying transformations to a vector that does not contain any words at all. Words that come out of an LLM are just what sampler gives us by looking at the vector that is the result of those transformations.

Does this vector contain a world model? Some form of thoughts or reasoning to arrive at the result? That's an open question really.

Sure, rerunning that whole process for each token might not be the best solution, although that's an open question too. But saying that LLMs operate on words first is too big of an oversimplification

I see the point, but really this is solved better by having a solid foundation of UI components (might be an existing UI-kit or own thing, a piece of CSS anyway), and then you can use the same-ish classes over many sites.

What I find with any site is that without a good foundation you're bound to repeat same styles over and over, and with one you don't get any point with Tailwind.

Probably just the convenience.

Expression `select(User).where(User.id == user_id)` does look simpler and cleaner than `select(User).where(eq(User.id, user_id)))`.

Maybe there is an actual eq function of sorts that allows that, and it's not a life-or-death difference here, but as far as convenience goes, it works well. And typecheckers are good enough to be able to let you know what's the return type.