raihansaputra

i know they're not making any promises but i'm also very excited for it

how are you doing the etl from mssql to duckdb? are you doing naive snapshots? or ducklake?

throwing in my 2 cents: It just replaced pandas for me. It's just so much easier to write sql against csv/json/whatever format data in jupyter/marimo notebooks through duckdb rather than reasoning through pandas. SQL is far more natural for me, and agents also work through it easily.

So this is using deepseek-v4-pro / flash and using claude code as the harness right? That's really good that it works. I'm pretty impressed with how v4-pro is doing, sadly there's no subscription packages so I'm not using it too much. (Wafer used to offer it but now they don't).

What was the media for updates? Send them a CD or a flashdisk and they plug it in? I assume the PVC backing etc they handle on their own?

If you’re counting by the framebuffer than the only 2x is the hidpi-forced-into-lowdown like the link suggest. The ratios I’m referring to is the UI scaling against the physical display pixel dimensions.

Apple basically settles into 220 dpi physical displays (250 moc MacBook Pro), rendering UI at a 2x scale as a default, except for MacBook Airs.

my experience with RustDesk + Tailscale is really good. when i only needed windows remote devices, the Windows Remote Desktop over Tailscale is great. I used RustDesk for access to macOS and Android devices (all over Tailscale) and works great too. No complaints really. Much less hassle to setup than Parsec (latency isn't as good though), and way snappier than AnyDesk through their network. Great tool enabled by Tailscale.

i wish that's for their logging/alert. i definitely gauge model's performance by how much those words i type when i'm frustrated in driving claude code.

It’s actually around 1.5x for the default resolution out of the box and 1.3x for “more space” setting on m1/m2 MacBook Air. 1.1x supersampling on Macs makes it worse because down sampling to pixel alignment becomes a hot mess.

goal? yeah. but in reality just timing it right (starting a session at 7-8am, to get 2 sessions in a workday, or even 3 if you can schedule something at 5am), i rarely hit limits.

if i hit the limit usually i'm not using it well and hunting around. if i'm using it right i'm basically gassed out trying to hit the limit to the max.

ok it's a bit late but i think a big part is the non-verbal thing you're putting out.

my story is me and my wife moved to another country a few years ago for my study. after 4 months moving there, she already know and conversed with the people working the apartment and some neighbors. while i mostly just exchanged cursories and nods and glances. then one day we just walked out together and the same people i passed earlier just says hello and converse and stuff with my wife and me. yes she's very much an extrovert but i can see people are way 'more open' and my wife has that too. me on the other hand do have 'i don't want to bother you so please don't bother me' vibe.

man this is cool. this is what i want since i read about fly.io's sprites.

slightly related to this, and i'm not familiar with linux sandboxing/containerization in depth, but any similar turn key solution for linux desktop/server? ideally i have something like sprites/shuru but on my own linux/ubuntu server instead of in the cloud.

i agree that is annoying but seems like anthropic's stance is that the task/agent should be provided an environment to write the file in the output you provide or provided a skill.md description on how to do that specific task.

personally it's a blurry line. most times i'm interacting with an agent where outputting to a file makes sense but it makes it less reliable when treating the model call as a deterministic function call.

huh i never noticed it. interesting to see, especially because everytime esp32 is brought up the opinion is that they're not fir for production scale units/quantities.

so it's the one managing the ubiquiti specific features, and controlling the switch chip?

actually the web search/image needs the higher tier subscription with their mcp. k gladly paid for it, i basically don't care about usage anymore. it's great at execution. i still find it not so great at planning though. i wish i can use api sonnet/codex for planning and glm on claude code for executing easily, with codex/claude code. best thing right now is kilo code with the orchestrator but haven't tried it.