Amazing to me that this billion dollar plus mission (InSight) is now offline because of dust on solar panels. We’ve had solar panels more than four decades now. I don’t assume the challenge is easy, and know that dust is endemic to Mars, but think of the impact from solving this problem on longevity of future missions! After establishing a habitat, Mars visitors need to revive all the offline experiments with a hanky.
Yeah, sounds about right. Real world problems are rarely 1D, constant (or no) velocity, etc.
there are good examples out there, the one by rlabbe on GitHub is pretty dang good. It still didn’t quite get me as far as I needed to go, but it got me a lot further than I was when I started.
That’s a cool story.
I feel like this is so overlooked. I’ve worked with scientists and engineers who’ve actually deployed systems that are incredibly complicated and pretty well known. Those folks are usually skeptical of trendy algorithms and usually start with first principles. 90% of the time, a boring old linear Kalman with simple (or no) motion model will do the trick.
I love ELI5 articles. I work in fields that use Kalman Filters. I have an MS in EE. I’ve been in industry for more than ten years and tried my hand with more than one KF. I know the terms and mostly grok the math. I’m a Senior trending to Principal at a faang, working in robotics. I STILL STRUGGLE and have never deployed a KF to production.
Side story - a friend of mine tried to recruit me to a well known robotics startup in the commercial autonomous air vehicle space (DM me if you’re curious who). The technical portion was basically, “implement a state estimator for radar target tracking, In one day, here’s some data”. Meanwhile, I have a day job. Lol, I told ‘em, “I’m not a good fit if this is what you’re looking for”. Oh yeah, my recruiter/friend also said they were looking for rock stars. I feel like I really dodged a bullet there.
Nicely done Brendan! Thank you, knowing Brendan’s work with eBPF, I take this as a way to more easily monitor and assess performance under different types of performance. Tweaking/tuning comes with trade offs, and I usually end up optimizing one thing to the detriment of others..
Side note, I’ve found btop a super useful replacement for glances, to have an all-in-one TUI view of system performance and loading. Wonder how much those dev(s) are leveraging this, and whether anything out there’s motivated to build better TUI monitoring tools.
Every server I go on, first thing is, start up tmux, dedicate one window to btop.
People forget that getting answers is not what makes a good solution. Computers are still dumb, they only give answers. Asking the right question is everything. With the right question the uncertainty of a problem is resolved. Finishing is a mere exercise. whether hand-crafted assembly or an LLM’s regurgitation and synthesis of things it’s seen before is just a matter of efficiency.
It’s product placement. Happens in retail and grocery all the time. Sellers pay to have their products at eye level, stores optimize placement to boost spending.
Now that it’s digital it’s illegal?
Yeah huh, that’s exactly where they are. We can debate whether the current antitrust laws are broken, we can debate how much government should regulate big tech, but Lena Khan is testing theories and novel legal tactics after the fact. I put more faith in Amazon’s legal chops than government any day, they’re clear of violating. If anything, they’ve taken the existing laws and ran them to their conclusion, faster and more efficiently than any one else.
It’s easy to say, “imagine if there was no Amazon and we were all still living in Walmart and Sears’ retail landscape”, precisely because Amazon came in and changed the game for the betterment of all. Now, FTC seems to be saying, “imagine the little Amazons and Googles that are being suppressed!” It’s hypothetical and not how regulation should be done IMHO.
We’re in need of good regulation of big tech, this is ultimately going to damage that effort. It’s not in good faith.
For my use cases the asyncio wrapper makes it really easy to stack up a bunch of tasks, let the OS it’s thing, and then collect the results when they’re ready.
The mess more reflects supporting a programmatic interface to processes in a cross platform manner, coupled with the actual complexity of parallel processing.
You didn’t mention the recommended high level option for subprocess, ‘subprocess. run’.
People are busy and often times will be aware of the broad strokes of the topic at hand, but not necessarily the latest details. A well written narrative will bring everyone back to the salient choices.
I have a stellar contract SDE from Morocco. Aside from the humanitarian tragedy of this earthquake, my eyes have been opened to the challenges young professionals with Visas face in the US. They tend to be ripe for exploitation by hiring firms and contract agencies.
One aspect of Amazon culture I’ve enjoyed is that meetings are document driven, and PPTs are rare, even non existent. Meetings start with dedicated reading time. It feels like PPT engendered this false sense of communication by making it easier to bullet-ize and visualize. PPT in practice lowers the critical thinking required of presenters and presentees.
I’ve used make like this in the past, it’s handy. But zsh auto suggestion obviates the extra maintenance IMO. I type the command one time and it’s in my history. So my common dev actions are more flexibly maintained by recency than editing a file.
You could say the Makefile also serves as a form of (not great) documentation.
Do you mind sharing what make (vendor) of radar being used? The “AI” in these types of systems is usually more likely classical ML than deep learning, in my experience. The problem is solvable, but it needs more training data and investment to solve to some acceptable performance level. There are a lot of companies trying to cash in on counter drone, and hype-speak like “AI” abounds.
People are diverse and I feel that tech, in the whole, is coming around to that reality in practice and not only in PR. I work in big tech on “cool projects”. Our particular area values experience and intelligently failing fast, which also takes experience. My early career was in the military and I didn’t start working in engineering until my early 30s. Now I’m pushing mid-40s. In a certain way, I got to skip alot of years as a Junior dev and transition right into architect and design roles.
It’s indispensable in some workflows. Scoring history can help with otherwise difficult debugging, or looking up a variant to a complicated invocation that’s not the one suggested by auto complete.
Children are complaining, but this is exactly right. Yes, I can work around the device connecting me to the internet. What I’d really prefer though is to exchange whatever cost went into the shit add ons, the Bluetooth, the speakers, the wifi, the usb, and put that into a better display. Done. I’m sure there’s a good business reason why this simple, obvious project is not what’s in the market.
Doppler is measured across multiple chirps or pulses. So doppler is not really encoded in the same signal used for range. Relative motion induces a phase rotation across multiple chirps, which can be coherently combined with a Fourier transform.
Your comment seems skeptical. Practically, It’s a night and day difference if you’ve integrated a linter with your editor. Plus, ruff is able to fix as well as lint.