jinpan

claude code and friends will make it much easier to rewrite python/js/$memory-hungry code in rust.

not too far from an onion video https://www.youtube.com/watch?v=CkDKmSMvfmk

Most of their tech is stolen or trickles down from innovation elsewhere

Overheard at a battery conference: western battery suppliers are actually copying (or in your words, stealing) the Chinese electrolyte formulas and other designs.

Great artists steal, the US and Europe play the same game.

Pretty good guess but not quite. Email me if you want to chat more!

Wilkie Way Research | Infrastructure Engineer | Remote (US Only) + Peninsula

We're a small (<5 person) quantitative prop trading shop looking for a midcareer engineer to take ownership of core infrastructure. We place a premium on work life balance and offer profit sharing.

Stack is mostly Python (Django, scientific stack, celery) + Rust.

More details at https://www.wilkiewayresearch.com/careers.

Email me (CTO): jin@wilkiewayresearch.com

It vastly increased the number of end users (20M+) using apps built on our SDK. This gave us significant leverage with Google. They are super supportive of the SDK and give us early access to several builds to ensure the SDK doesn’t break when they make updates to Gmail.

One of my favorite essays on "vs effective altruism" - https://astralcodexten.substack.com/p/effective-altruism-as-...

I jot down my notes in a google doc and paste links that I want to save.

If I really like an idea, I'll use these notes to write out a larger form memo that touches on GTM, product strategies, tech stacks, competition, etc.

At the risk of defending a faceless megacorporation, I'm curious what injuries per package look like at Amazon vs "Industry Average".

According to the article, productivity expectations per worker doubled with robotics, and Amazon has about double the injury rate per worker, so my super rough back of the envelope calculation says they should fare similarly on that metric.

If the industry alternative takes twice as many workers to do the same job, ultimately leading to approximately the same number of injured workers, has Amazon really created a safety crisis?

Have you heard of this thing called quantum computing?

The article did say that the motivation for doing so was because AWS credits were running out .. why prematurely optimize a free resource? :)

I built some coronavirus simulations to explore the impacts of our actions in a virtual environment - let me know what you think!

Goals of this project are to spread high quality information around and use some basic monte carlo / bayesian ideas to show the impact of individual actions on the individual.

You may also be interested in https://coronavirus.simrnd.com/shopping_solo/, an earlier simulation that explores the impact of sending out 1 person to shop vs 2 per household.

The simulation engine is written in Rust at https://github.com/jinpan/covid-simulations, which is compiled to WASM to run on the browser.

Yeah that would be incredibly useful for testing policy in a simulated environment before releasing into the wild :)

One of the hardest challenges is "How do we choose realistic parameters for how the virus spreads? Infection duration? Mortality?"

But if we can incorporate the latest research and Monte Carlo simulate with ranges for the above parameters and find policies that are robust against those parameters, I think it would be an enormously powerful policy tool. Additionally, I think it could bring transparency into the decision process, something the public lacks.

If you want to contribute to the effort I started, the simulation code is open source at https://github.com/jinpan/covid-simulations. The engine is in Rust, which is then compiled to wasm.

Semi-related: I'm working on agent-based simulations for coronavirus, exploring our individual actions affect outcomes.

In the first simulation, I take a look at single/dual shopper households and explore how choice affects viral spread on a societal level, and then on am individual household level.

You can check them out at https://coronavirus.simrnd.com/shopping_solo/

Legendary investors Stan Druckenmiller and David Tepper were the latest to weigh in after a historic market rebound, saying the risk-reward of holding shares is the worst they’ve encountered in years.

True, but what are they going to do about it?

They would be hard-pressed to find a rewarding asset class today that has billions in liquidity, with less risk.

If they hold cash/tbills, they risk losing a good chunk to inflation from the new monetary policies.

They could try to hedge for a market crash, but timing it seems difficult - the market can remain irrational longer than they can remain solvent. And hedges are aren't cheap with the higher IV levels today.

I've been working on a coronavirus simulator in rust/wasm.

https://coronavirus.simrnd.com/about/ is a draft of the intentions of the project, and the source code is at https://github.com/jinpan/covid-simulations.

https://imgur.com/a/wFTq7lD is a screen recording of a shopping scenario.

I'm aiming to publish a blog post with some initial simulation results by the end of the week.

I gradually graduate a notebook by extracting what I can into libraries, and importing those utilities into the notebook. Eventually the logic in the notebook becomes trivial enough to easily convert into a production program.

"And the next thing you know, you’re at the zoo, shaving a yak, all so you can wax your car."

I'm interpreting this as (maybe unintentionally) creating new forms of debt - next time you want to visit the zoo, you may need to resolve some favors to the zookeeper that let you into the yak exhibit.

To blend in with humans, these fake profiles could take advantages of AI progress on the Turing test to converse with real people, ultimately befriending (some high fraction of) us.

The persuasive effects of these fake friends on (political) advertising are terrifying.

Another fun surprising result from queuing theory is that if you have a buffer of pending RPC's that have some deadline, it is better to service them in a "unfair" LIFO order, as opposed to a fairer FIFO order. [0] goes into more gory details.

In the bank example from the article, LIFO would dramatically shorten the median wait time (~5 minutes), at the cost of really upsetting customers [1]. In the case of a support queue, where the waiters are blind to each other, this cost could be reconsidered ...

[0] https://arxiv.org/pdf/1008.4895.pdf

[1] from the article, "We really, really hate it when someone shows up after us but gets served before us.

Given that the A12X runs at < 10% of the TDP of an Intel i7 chip, would it be feasible to utilize many of these A12X's in a 10+ socket motherboard to significantly increase NUMA-compatible workloads?

IMHO, it is too soon for Python 4.

It would be interesting to augment an image recognition dataset with additional inputs generated by minor transformations under these scripts.

Rivest invented the 3-ballot voting protocol (see https://en.wikipedia.org/wiki/ThreeBallot)

The trick to the 3-ballot voting protocol is that not everyone can validate their vote. However, when a group of voters band together, they should be able to validate that a subset of their votes (1/3) were registered with high probability.

With a sufficiently large group of voters checking results, the law of large numbers comes into play and they can statistically detect the presence of voting fraud.

Yes, I agree this is a terrible idea in some respects, and we're not certain if we're set on actually donating the jackets. One route we're actively considering is contacting the donors and asking them if they're open to the possibility of us directly giving the funds towards more traditional nonprofits, and refunding those against the idea.

Yes, this does represent a certain level of immaturity. The idea behind this originated as a prank, but it was meant to be a harmless one. We didn't want to dance around being politically correct so we broke some social norms.

However, we are not giving the homeless jackets because we want to laugh at them. We are trying to put a spin on the gamification of charity to see if we can get people to do more than they would otherwise.

It is certainly not our hope that people use this as an avenue to dehumanize the unfortunate.

But is it more acceptable to see the unfortunate clearly wearing thin clothing when it's 20 degrees out and maintain the idea that they are people who must maintain their pride, or offer them the chance to sell out? It's a hard question to answer, but I think society has already answered that for us. Fast food restaurants already offer young professionals a chance to sell out by offering higher wages than many starting jobs in alternative industries [1].

1 - http://www.nytimes.com/2012/11/25/magazine/skills-dont-pay-t...

Well, you have this enormous amount of energy in the form of college rivalries going towards (in my opinion) a relatively silly ball game. Our hypothesis is that if you could tap just 1% of that energy, you could accomplish something great. This is our first foray into tapping that intensity.

I don't see why high school students can't use LinkedIn. I was accepted to a summer internship during high school via LinkedIn.