I've found what works best here is just switching to every 2 weeks or every 4 weeks. If you have little to talk about in a 1:1, feel free to end early, and then double the length of time until the next one.
Yep, time box it, so you know you have time, but allow it to e shorter!!!! Or reschedule. Mostly I had 20mins, every week with most. Some became 45mins or more, as we rambled on about tech or some other topic. And one requested it once every 2 weeks. Fine with me. If that makes them feel better, please.
While I agree with basically all of this, and find the FSD on my Tesla to be quite useful, a question pops into my mind.
Why can't Waymo ALSO develop the same smarts and just also solve the sensor fusion issue such that they can use the right set of sensors in the right environmental conditions, and then leapfrog Tesla's capabilities?
Because they don't have a fleet of millions of people labeling the data for them and paying for the privilege of doing so. Waymo has about 3700 vehicles. Tesla has millions. Waymo only operates in known environments and collects a very limited range of data. Tesla collects data everywhere that people drive their cars.
I thought about this and I think it boils to how the model is trained.
Tesla trains it models from actual drivers purely based on (input) Vision and (output) actuators - Brake, Steering, Accelerators.
Human output is based on what they and the camera sees. So, it's a 1:1 match.
If Waymo were to do that, it'll muddle the training set. The Lidar input may override camera input.
I always struggled when Musk mentioned Lidar will make it ambiguous. It didn't make any sense to me why having a secondary failback sensor messes things. But, if you put it in the training data context, it absolutely makes sense.
This is an interesting viewpoint, but isn't it also solveable?
Just because the human in the scenario only took vision as input, why does that matter to the training data and the model? The actions are the same.
To put it another way, what about all the cultural context the human had, or the sounds, smells, past experiences at the same intersection, etc? Even Tesla can't record this, but I'm not sure that matters.
E.g If the driver brakes because they saw a pothole, and Lidar captures someone biking 200m away on their own path, it may mistakenly put more weight on brake causation to the 200m away object (because large moving object) vs the pothole.
I'm exaggerating, but I hope you get the point. It isn't even conflicting sensor signals about the pothole, but conflicting information about the causation. With vision only there is no conflict for the training data. This was my Aha moment. Multiple Sensors are absolutely important for fallback and extra safety, but screws up training that are based on Human Drivers
I think Elon himself doesn't understand this and hence can't articulate it, while just repeating whatever his ML engineer has said.
That is vastly preferable to slamming into the back of an emergency vehicle because the cameras are dazzled by the strobes, or slamming into tractor trailers because the cameras were blinded by sunlight. Or slamming on the brakes because the car thinks a shadow in the road is a physical object...
> such that they can use the right set of sensors in the right environmental conditions
Because this part is really hard, and that's why Tesla abandoned the fusion approach. You cannot possibly foresee all the conditions in which LIDAR or any active sensor will malfunction/return wrong data/return data that's only slightly off for that ONE specific time. And even if it doesn't, you need to trust it to not return noise. And when it does return noise, how do you classify it as noise?
Cameras are passive sensors - they get whatever light comes in and turn it into an image. Camera is capturing shapes that make sense to the neural nets: it's working. See all black/white/red/cannot see any shapes? Camera is not working, exclude it from the currently used set of sensors or weigh it less when applying decisions, because it's returning no signal (and yes, neural nets have their own set of problems).
EDIT: cameras also provide more continuous context: if 1 pixel is off, is clearly bright red in a mostly-green scene where no poles can be identified, the neural net will average it out and discard it as noise. If 1 pixel says "object" in LIDAR, do you trust it to be correct? Perhaps the ray just hit a bird or a fly, but you only see a point, it's a lossy summary of the information you need.
They could in theory. If they put at least as much emphasis on the AI side as Tesla does. Or if someone else cracked vehicle AI wide open and left it open for them to copy, and then they did exactly that, and found a way to bolt on their extra sensors in a useful fashion while at it.
As is, Waymo's playing it smarter than Cruise did, but they're not all in on AI yet. So I don't expect them to "leapfrog Tesla" in that dimension - and it's the key dimension to self-driving.
I don't know. Cost might have been part of it but I also recall hearing that he thought since humans can drive with two eyes and no LIDAR then the car should be able to do the same thing.
I got downvoted for saying this last time the topic came up but constraints focus a project. It’s best to start work with as few variables as possible, and only add new ones when absolutely necessary.
I'm working on a similar problem in computer vision and we're quickly approaching the point where our pure vision work is better than our Lidar supported track because we've had to deal with the constraints instead of having a crutch to lean on.
I agree, but these are also the exact constraints that lead to an early leader getting overtaken by a longer term, yet better set of plans. Not saying that's the case here, but given how much success Waymo has had so far, over really everything Tesla has produced, says quite a bit about the likelihood of the approach, even if it's not yet there.
You're missing two things from the whole picture:
1. Cloud mode works without local network access, so their server is involved in the transit of the data to the printer. This is pretty minor, but still within their rights to preserve.
2. For printing from the app, they actually run the computationally expensive slicing algorithm on their servers, so this is totally reasonable to protect.
> No, we aren’t being blocked. Turn on LAN mode, pair regular Orca slicer, ignore Bambu for the rest of eternity. Plenty of people have done it.
You're just saying that Bambu users feel the need to purposely circumvent Bambu's artificial restrictions to be able to continue to use Bambu hardware they bought and paid for.
It's a toggle you set in the printer directly, nothing is circumvented. Only the access through their cloud service is impacted, but the printer works locally like any other.
Pretty sure you can still print locally either via LAN or just SD card. At least I can on my A1.
The current monetization that they are using is that you can charge for a print on their platform and they take a cut of the sale. If you don’t charge for the design, then it is still free hosting and delivery.
I see where the worry is, but at the moment it seems like people are imagining a worse case scenario.
If you turn on LAN mode, it acts exactly like every other printer. You can print directly to it from any slicer over your LAN, or dump gcode on the SD card directly.
People are saying the LAN mode lacks access to the webcam and possibly some other things. That is what this whole controversy is about. It's re-enabling some cloud features as local only and Bambu is calling it privacy or fraud.
They probably want to establish a commercial-use license. If you have a big print farm, you likely need all of those remote capabilities so you're going to need to pay for the license. The schmucks at home will likely continue to get it for free. Locking them into the cloud API by dangling convenient features just ensures most people won't stray into the local-only mode.
> 2. For printing from the app, they actually run the computationally expensive slicing algorithm on their servers, so this is totally reasonable to protect.
That's an artificial vendor tie-in, and arguably a feature that only involves their client app and their backend. It's understandable if access to their backend is restricted to a subset of their users if that's the business model they wish. Preventing paying customers from using the hardware they bought and paid for by imposing artificial restrictions is not cool.
They've bought a machine that executes gcode and that it does (at least to my understanding) regardless of where that gcode comes from.
If you want special secret sauce gcode from the bambu cloud, you need to use the bambu cloud.
Those are not the same thing, so IMO it is legit what they do there, because it's such a clear-cut split.
You own the physical thing but not the ecosystem around it.
___
I would of course personally never buy a bambu lab printer, because they're cloud-tied nonsense that was going to behave exactly like that (the split between hw and ecosystem), but other people knew that too and still bought it, because "what a nice ecosystem".
idk.
I just don't think that "right to repair" should mean "right to be saved from the consequences of my own bad actions".
Those bad actions continuing to have no real painful consequences (and with that no real learnings + behavioral correction) after all is why the state of tech has become as bleak as it is right now.
And, honestly, if you can afford a bambu premium machine, there's a 97% chance that you could easily shoulder a total write-off.
There's also a 97% chance that your ego can't, but that's the main thing causing all the bad things in the world and should've died a long time ago. Approximately post-highschool.
I feel you on every point, but I'd point out that they sell VERY reasonably priced replacement parts for basically everything on the A1 Mini that I have that is worth repairing.
I wouldn't buy an alternative to a P1S, because only the P1S is the best at being the P1S. (Whatever that might entail)
Instead, I'd look at things from the perspective of "what do I want?" and not "What does the market offer? Okay, I want that thing. But no, I want an alternative to it that is that thing but without downside"
Letting a brand set your frame of reference is the first step into total dependence.
Thanks for your reply. Only used PLA so far. But later I'll need "engineering parts", Nylon/PA12 or something like this. Strong, water and UV resistant, outdoor.
It shouldn't be too complicated and not too expensive. E.g. while the Prusa Core One+ seemed nice (from a superficial look) it costs more than I wanted to spend. P1S came out as the best (barely) adequate printer for what I thought I would need when I looked at it. But it's difficult to say if you are a beginner and basically have no idea...
Yes but what does "equally good job of printing" mean, I wonder.
That's what I meant with "the P1S is the best at being the P1S when measured by the P1S".
I am pretty sure that if you for example do functional PLA parts, there will be many, many more options that tick exactly that box.
I do of course understand that people want to have the mental peace of buying one thing and being told that it can do everything, but, as said, you pay for that emotional labor with lock-in and eventually being rug-pulled.
The only way of not getting rug-pulled is not handing away all of your agency wholesale just for cheap immediate emotional relief.
That's how it works, how it has always worked and how it will always work.
Anyone claiming anything else is in the process of actively scamming you.
This is surprisingly close to a personal theory I've been working on. I've been describing how to use AI to people as engaging the world model in their head, organization, or software.
I'd love to talk more live. I think I have some ideas you'd be interested in. Find me in my profile.
I find connecting understanding between humans and agents is one of the most important parts of the agentic development cycle, and markdown is a great way to handle that.
Not only can you point it at an entire directory, you can point it at multiple projects, quick load a project with a keyboard shortcut, and also easily see recent file that changed to help you find the 75th file your agent just wrote for you.
Recently, I've started to add a review interface where you can track changes, and add comments for your agent, and then instead of trying to do some complicated integration with an agent, it just has a copy button, and it copies all the comments, which context, and instructions for the agent how to reply.
I also find that I generate TONS of markdown junk during development, and I needed a way to handle it and keep it out of the main repository so I built this tool:
Vantage looks great! I’ll try it out this weekend.
To do the job that swarf does, I found that the bwrap sandbox I’d been using is the perfect place to mount a folder to catch markdown junk and keep it out of the project’s actual git repo. Works great.
reply