All System Prompts For Anthropic’s Claude, Revealed

October 12, 2024 by Donald Papp 5 Comments

For as long as AI Large Language Models have been around (well, for as long as modern ones have been accessible online, anyway) people have tried to coax the models into revealing their system prompts. The system prompt is essentially the model’s fundamental directives on what it should do and how it should act. Such healthy curiosity is rarely welcomed, however, and creative efforts at making a model cough up its instructions is frequently met with a figurative glare and stern tapping of the Terms & Conditions sign.

Anthropic have bucked this trend by making system prompts public for the web and mobile interfaces of all three incarnations of Claude. The prompt for Claude Opus (their flagship model) is well over 1500 words long, with different sections specifically for handling text and images. The prompt does things like help ensure Claude communicates in a useful way, taking into account the current date and an awareness of its knowledge cut-off, or the date after which Claude has no knowledge of events. There’s some stylistic stuff in there as well, such as Claude being specifically told to avoid obsequious-sounding filler affirmations, like starting a response with any form of the word “Certainly.”

Continue reading “All System Prompts For Anthropic’s Claude, Revealed” →

Solar Planes Are Hard

October 12, 2024 by Danie Conradie 12 Comments

A regular comment we see on electric aircraft is to “just add solar panels to the wings.” [James] from Project Air has been working on just such a solar plane, and as he shows in the video after the break, it is not a trivial challenge.

A solar RC plane has several difficult engineering challenges masquerading as one. First, you need a solid, efficient airframe with enough surface area for solar panels. Then, you need a reliable, lightweight, and efficient solar charging system and, finally, a well-tuned autopilot to compensate for a human pilot’s limited endurance and attention span.

In part one of this project, a fault in the electrical system caused a catastrophe so James started by benching all the electricals. He discovered the MPPT controller had a battery cutoff feature that he was unaware of, which likely caused the crash. His solution was to connect the solar panels to the input of a 16.7 V voltage regulator—just under the fully charged voltage of a 4S LiPo battery— and wire the ESC, control electronics, and battery in parallel to the output. This should keep the battery charged as long as the motor doesn’t consume too much power.

After rebuilding the airframe and flight testing without the solar system, [James] found the foam wing spars were not up to the task, so he added aluminum L-sections for stiffness. The solar panels and charging system were next, followed by more bench tests. On the test flight, it turned out the aircraft was now underpowered and struggled to gain altitude thanks to the added weight of the solar system. With sluggish control responses,[James] eventually lost sight of it behind some trees, which led to a flat spin and unplanned landing.

Fortunately, the aircraft didn’t sustain any damage, but [James] plans to redesign it anyway to reduce the weight and make it work with the existing power system.

We’ve seen several solar planes from [rctestflight] and meticulously engineered versions from [Bearospace Industrues]. If long flight times is primarily what you are after, you can always ditch the panels and use a big battery for 10+ hour flights.

Continue reading “Solar Planes Are Hard” →

Remembering John Wheeler: You’ve Definitely Heard Of His Work

October 12, 2024 by Donald Papp 9 Comments

Physicist John Archibald Wheeler made groundbreaking contributions to physics, and [Amanda Gefter] has a fantastic writeup about the man. He was undeniably brilliant, and if you haven’t heard of him, you have certainly heard of some of his students, not to mention his work.

Ever heard of wormholes? Black holes? How about the phrase “It from Bit”? Then you’ve heard of his work. All of those terms were coined by Wheeler; a knack for naming things being one of his talents. His students included Richard Feynman and Kip Thorne (if you enjoyed The Martian, you at least indirectly know of Kip Thorne) and more. He never won a Nobel prize, but his contributions were lifelong and varied.

Continue reading “Remembering John Wheeler: You’ve Definitely Heard Of His Work” →

Cockroaches In Space: Waste Processing And A Healthy Protein Source Combined

October 12, 2024 by Maya Posch 51 Comments

As the current frontier of humanity in space, the International Space Station is heavily reliant on Earth not only for fresh supplies but also as a garbage disposal service for the various types of waste produced on the ISS by its human occupants. As future manned missions take humans further away from Earth, finding ways to reprocess this waste rather than chucking it out of the nearest airlock becomes a priority. One suggested solution comes from a Polish company, Astronika, with their insect bioreactor that can process organic material into useful biomass.

Interestingly, the cockroach species picked was the Madagascar hissing cockroach, one of the largest (5 – 7.5 cm) species. This is also a cockroach species which is often kept as a pet. In this closed-loop bioreactor that Astronika has developed, these cockroaches would chew their way through up to 3.6 kg of waste per week in the large version, with the adult cockroaches presumably getting turned into fresh chow and various materials at some point. Beyond the irrational ‘yuck’ factor that comes with eating insect protein, one of the biggest issues we can see with this system is that the long-duration mission crew may get attached to the cockroaches, as they are rather cute.

Continue reading “Cockroaches In Space: Waste Processing And A Healthy Protein Source Combined” →

Approximating An ADC With Successive Approximation

October 12, 2024 by Al Williams 17 Comments

[Igor] made a VU meter with LEDs using 8 LEDs and 8 comparators. This is a fast way to get one of 8 bits to indicate an input voltage, but that’s only the equivalent of a 3-bit analog to digital converter (ADC). To get more bits, you have to use a smarter technique, such as successive approximation. He shows a chip that uses that technique internally and then shows how you can make one without using the chip.

The idea is simple. You essentially build a specialized counter and use it to generate a voltage that will perform a binary search on an unknown input signal. For example, assuming a 5 V reference, you will guess 2.5 V first. If the voltage is lower, your next guess will be 1.25 V. If 2.5 was the low voltage, your next guess will be 3.75 V.

The process repeats until you get all the bits. You can do this with a microcontroller or, as [Igor] shows, with a shift register quite simply. Of course, you can also buy the whole function on a chip like the one he shows at the start of the video. The downside, of course, is the converter is relatively slow, requiring some amount of time for each bit. The input voltage also needs to stay stable over the conversion period. That’s not always a problem, of course.

If that explanation didn’t make sense, watch the video. An oscilloscope trace is often worth at least 1,000 words.

There are, of course, many ways to do such a conversion. Of course, when you start trying to really figure out how many bits of resolution you have or need, it gets tricky pretty fast.

Continue reading “Approximating An ADC With Successive Approximation” →

If You Can’t Say Anything Nice

October 12, 2024 by Elliot Williams 399 Comments

[Editor’s Note: After we posted this, we got hit by a comment-report attack, and about 1,000 (!) comments across the whole site got sent back into the moderation queue on Saturday. We’ve since re-instated them all, but that took a lot of work.

About halfway down the comments in this article, the majority of comments are “hey, why did you delete this?” We didn’t, and they should be all good now. We debated removing the “try deleting this!” comments, but since we didn’t delete them in the first place, we thought we should just leave them. It makes a royal mess of any discussion, and created a lot more heat than light, which is unfortunate.]

You know what your mom would say, right? This week, we got an above average number of useless negative comments. A project was described as looking like a “turd” – for the record I love the hacker’s angular and futuristic designs, but it doesn’t have to be to your taste. Then someone else is like “you don’t even need a computer case.” Another commenter informed us that he doesn’t like to watch videos for the thirtieth time. (Yawn!)

What all of these comments have in common is that they’re negative, low value, non-constructive, and frankly have no place on Hackaday. The vast majority are just kind of Eeyorey complaining about how someone else is enjoying a chocolate ice cream, and the commenter prefers strawberry. But then some of them turn nasty. Why? If someone makes a project that you don’t like, they didn’t do it to offend you. Just move on quietly to one you do like. We publish a hack every three hours like a rubidium clockwork, with a couple of original content pieces scattered in-between on weekdays.

And don’t get us wrong: we love comments that help improve a project. There’s a not-so-fine line between “why didn’t you design it with trusses to better hold the load?” and “why did you paint it black, because blue is the superior color”. You know what we mean. Constructive criticism, good. Pointless criticism, bad.

It was to the point that we were discussing just shutting down the comments entirely. But then we got gems! [Maya Posch]’s fantastic explainer about the Lagrange points had an error: one of the satellites that Wikipedia said was at an earth-moon Lagrange point is actually in normal orbit around the moon. It only used the Lagrange point as a temporary transit orbit. Says who? One of the science instrument leads on the space vehicle in question. Now that is a high-value comment, both because it corrects a mistake and enlightens us all, but also because it shows who is reading Hackaday!

Or take [Al Williams]’s article on mold-making a cement “paper” airplane. It was a cool technique, but the commenters latched onto his assertion that you couldn’t fly a cement plane, and the discussions that ensued are awesome. Part of me wanted to remind folks about the nice mold-making technique on display, but it was such a joy to go down that odd rabbit hole, I forgive you all!

We have an official “be nice” policy about the comments, and that extends fairly broadly. We really don’t want to hear what you don’t like about someone’s project or the way they presented it, because it brings down the people out there who are doing the hard work of posting their hacks. And hackers have the highest priority on Hackaday.

Have You Heard Of The Liquid Powder Display?

October 12, 2024 by Jenny List 10 Comments

Over the decades the technology behind flat panel displays has continuously evolved, and we’ve seen many of them come and go. Among the popular ones there are a few that never quite made the big time, usually because a contemporary competitor took their market. An example is in a recent [Wenting Zhang] video, a mystery liquid powder display. We’d never heard of it, so we were intrigued.

The first segment of the video is an examination of the device, and a comparison with similar-looking ones such as a conventional LCD, or a Sharp Memory LCD. It’s clearly neither of those, and the answer finally came after a lot of research. A paper described a “Quick response liquid powder” as a mechanism for a novel display, and thus it was identified. It works by moving black and white electrically charged powder to flip a pixel from black to white, and its operation is not dissimilar to the liquid-based e-ink displays which evidently won that particular commercial battle.

The process of identifying the driver chip and pinout should be an essential watch for anyone with an interest in display reverse engineering. After a lot of adjusting timing and threshold voltages the dead pixels and weird effects fall away, and then it’s possible to display a not-too-high-quality image on this unusual display, through a custom PCB with an RP2040. Take a look at the video below the break.

We’ve seen [Wenting Zhang]’s work here a few times before, most recently in a very impressive mirror-less camera project.

Continue reading “Have You Heard Of The Liquid Powder Display?” →