maho

The simple explanation: The positrons fly around the trap center. In the z-direction, this is a simple "up and down" oscillation. There are electrodes in the top and bottom of the trap called the "end-caps". The end-caps are connected to each other with a large resistor (technically, a coil forming an LC resonator with the parasitic capacitance, but on resonance it looks like a large resistor).

When the positrons are at the top, they attract negative charges into the top end-cap. This is called an "image charge". When the positrons are close to the bottom end-cap, the image charge into bottom end-cap. This forms a current, the "image current". The more positrons you have, the larger the image current.

Since the image current (on the order of femtoamperes) has to flow through the large resistor, it is turned into a voltage that can be amplified and measured by cryogenic amplifiers. But it has a second effect: The voltage "back-acts" on the positrons, damping their motion. If you have more positrons in the trap, this damping gets stronger, because each positron experiences the back-action of all other positrons.

In the paper, you can see pictures of the driven axial motion, where you take an RF excitation and sweep the frequency over the resonance, recording the positrons' amplitude as a response. The more positrons you have, the "wider" this resonance, because of the damping effect.

As you reduce the number of positrons by doing a "dip" (making the trap really shallow, hoping that some, but not all, positrons escape), you can see the width getting smaller and smaller in discrete steps. That's how you know you are dealing with individual particles.

I haven't worked on Gabrielse's setup, but I have worked on other, very similar (better, ha!) experiments 10 years ago. I don't see what's so special about this trap, but putting antimatter into it always makes for a nice headline.

I had a look under the hood of browser-use. I wanted to to run some tests with a local browser and a trusted LLM provider of my choice, but the (enabled by default) telemetry they send to PostHog is CRAZY. Especially the browser-harness component can leak credentials to PostHog left and right.

This is absolutely not cool.

Which model best allows me to transcribe speech that uses a lot of domain-specific terms? For example, when I say "Claude Code", it often gets transcribed as "Cloud Code", and I have to go back and edit or do a second pass with a traditional LLM (which can introduce additional errors).

The author only compared output token costs -- but for typical agentic workloads, input tokens dominate the costs by a large margin. Running inference locally, input tokens are, to first order, free. (They only generate implicit costs through higher time-to-first-token, higher power use, and lower token output speed).

+1 for introducing them as real-valued functions over cartesian coordinates!

Typically, spherical harmonics are introduced as a complex function over spherical coordinates, which makes them much easier to derive, but imo hides their beauty.

The real-valued, cartesian form of regular spherical harmonics is also called "solid harmonics" or "harmonic polynomials", in case you want to dig deeper.

> At a pragmatic level, can't you say, hey here's something thats probably nothing, let's scan it again in 6 months

If a doctor even _hints_ there might be cancer, the patient will have a terrible 6 months (with actual, measurable negative health impacts of the added stress). Also, at some uncertainy-level (say, 10% chance of cancer) the doctor _has_ to say something and has to schedule expensive followups to not risk liability, even though in 90% of the cases it is not only unnecessary, but actively harmful to the patients.

When, on average, the cost of the screening + the harm done by a false positive outweighs the benefits of an early detection, you shouldn't do the screening in the first place.

At $enterprise, we were just looking for a proper term that sets "responsible vibing" apart from "YOLO vibe coding". We landed on "agent assisted coding".

It's a bit more technical. And it has a three-letter acronym. Gotta have a three letter acronym.

I need this, but for heatpump noise.

It's mostly broadband noise that can be simulated by simpler methods, but visualizing possible resonance patterns for the low-frequency emissions from the compressor (which typically runs at 20Hz, 40Hz, ..., 120 Hz) would be good to know.

Although I am not sure how the 2d simulation result carries over to the 3d world...

The pronounciation of math symbols is hilarious, but not super useful. Prompt: "Give me Maxwell's equations".

I really like nools, which is a drools clone, but for JavaScript. It's fantastic for quick hacks and for getting to know how to write code for rule engines.

Sadly, it is no longer maintained.

Let's say I want to have a somewhat genereric CRUD-app: A database, some application logic, a login system, logging/telemetry and a UI.

What kind of stack should I use so that tools like Darwin (or just plain old copilot) can be most effective?

This is insane. The project includes the hardware (GHz-capable RF-generation and measurement), firmware (FPGA) and Software (a cpp GUI). Surely that can't be all from one person?