HN user

yonatan8070

1,028 karma
Posts2
Comments493
View on HN

But is that the fault of the LED? That sounds like the fault of whoever calibrated the LEDs' brightness and directionality.

If the beams were the right brightness and pointed downwards, we wouldn't be having this issue.

Regressive JPEGs 5 days ago

You could probably implement a server that is purposefully slow enough so that the human frame shows up for however long you want. Just need to send the keepalive bytes one-by one.

I don't usually treat LLMs like humans, when I have a more complex question I do take care to use full sentences. Though sometimes I just need something quick so I'd throw ChatGPT a "python print flush= what does"

I do talk to my devices occaiosnally though. Things like "You're coming with me" to my laptop before picking it up and putting it in my backpack

Reading all the comments here about how awful new cars are really reinforces my choice to buy a 2016 Nissan Micra as my first (and currently only) car.

It is capable of beeping, but only does it if I leave the handbrake up or start driving without a seatbelt.

I don't have modern comforts like a backup camera, cruise control, or those fancy motorized seats some cars have. But it does it's job. I push on the go pedal and it goes, I push on the stop pedal and it stops.

I also recently installed an Android Auto capable head unit, so I don't need to illegally interact with my phone to navigate places.

This is also the case for the TP-Link Decos we were using, you can only configure and manage them from the app, no local web UI in sight*

And it's not like a mobile app unlocked some new features that couldn't be implemented in a browser. The app is slow, sluggish, and basic things like adding a DHCP reservation took multiple tries to succeed, each taking an agonizingly long time of watching a spinner.

If you contrast this with my home MikroTik, the UI is less "your grandma could configure it" simple, but it's fast, available over local web, SSH, desktop app, mobile app, and I think also an API, and has every feature I can think of from basic bridging to complex routing and firewalling.

* there was a very basic web UI that I recall had like 1 or 2 settings, don't remember which exactly.

You'd be surprised, at work we recently had a contractor install a new office network (growing startup upgrading from the home grade units we had before), they recommended HP Aruba Instant On, and the guy did all the setup on his phone with their app.

Now that it's my network to manage, I have to say that while it's a huge upgrade from the TP-Link Deco units we had before, the cloud management just makes it worse. The web UI is slow, probably because every click requires a round trip from my laptop, to a datacenter somewhere, to the router/switch/AP 5m away from me, back to the datacenter, and back to my laptop.

ESP32-S31 2 months ago

It's like these for other families too, you've got your STM32, then you can get the sub-models ranging from entry-level STM32C0 to the full Linux chips like the STM32MP2, with lots of options in the middle

It's not X, it's Y = AI pattern.

Yeah, a human has never used this pattern before! Good thing AI always leaves this digital signature which is never wrong, so you always know if the person on the other end has used AI.

Another issue I've noticed is they're sometimes very resourceful. For example when Codex can't directly edit file due to sandboxing restrictions, rather than asking "hey can I apply this diff on the file", it'd ask for permission to run a `cat EOF` command to re-write the whole file, which the UI doesn't surface properly (just shows the first line...).

This sounds similar to what's described in the "Claude deleted my DB post", it decided "I need to do X", then searched for whatever would let it do X, regardless of intended purpose.

I'd say certianly yes. At my company we've set up an experiment: we've forked our monorepo and gave several agents (OpenClaw, Claude code, GH Copilot reviewer, Codex) full R/W access and a Slack channel where we give them tasks and they carry them out. The fork shows 490 commits ahead, 10 commits behind our real repo (we're only a couple people, and use "Squash and merge" on PRs).

They naturally produce bugs at an astonishing rate, and we don't review the code ourselves, but the project is growing faster than their context windows, and I believe we'll drop it soon.

Sounds serious to me

It's highly unlikely that the people behind an attack like this would come out (non-anonimously) and take credit. And it's unlikely they'll be caught. So does it matter to most peoplee if it's Russians, Americans, Iranians, North Koreans, or some other country?

If you're a 3-letter agency, you'd want to know and potentially arrest them, but as a random guy on the internet, or even a maintainer, I really don't think it matters.

I went on the configurator page briefly, like 400$ for 32GB IIRC.

They don't ship to where I am so I didn't stay long