nvartolomei

How? Have you tried contributing a reasonable implementation with test coverage and it was rejected?

> changes are aggregated and committed back to S3 roughly every 60 seconds as a single PUT

Single PUT per file I assume?

Upgraded to one of the latest iPhone recently. First time I clicked on “transfer data from old phone”. I’m used to reinstalling the operating system every couple of months from when I used Windows. It took maybe 15 minutes with close to 0 interactions. Everything was transferred. I was already authenticated in apps. What took manual steps was eSIM transfers.

I don’t remember exact steps so there could have been a bit more. But it was an impressive experience and I told my geek friends about it. They were surprised this is the first time I used this feature.

Wrap gdb in a shell script and you’ve got yourself an actual profiler: https://poormansprofiler.org/

I love P and have used it for personal projects (like OSWALD^1) and at work (for some Redpanda components).

I find it very approachable. If you know some basic theory of distributed systems and have been exposed to actors and state machines before, you should be able to have some non trivial models in a few days at days.

Things get slightly more complicated when you begin to ask yourself whether or not you have validated enough of the state space to have confidence in the design. The verifier has a much of options who care poorly documented.

Recently, P was extended to support exploring the entire state space (PEX mode) as well as a prover. These have only rudimentary documentation. I learned about them more by looking at code.

1: https://nvartolomei.com/oswald/

According to public information, AWS uses both and not only. AWS wrote a paper on the topic with some details about their applications of formal methods^1. AWS is a large org though, it would be hard to generalise or reduce their approach to any single method.

1: https://www.amazon.science/publications/using-lightweight-fo...

Later add: I believe P creator is employed by AWS at this moment.

Love that it makes you spell the mangled symbols. Forces the brain to do some proper work.

Please share more about Ray Data use case.

I do often ask Claude Code or Gemini CLI to write commits. I agree with you on why being important. Majority of these being bug fixes accompanied tests where the why is easily inferred from the change/newly added tests and their comments.

*Engines That Move Markets: Technology Investing from Railroads to the Internet and Beyond* by Alasdair G. M. Nairn

I’ve had a few aha moments while reading this book. Although it's primarily written from an investor's perspective, it does contain a fair share of insight about creation and commercialisation of technology, the mechanics of monopolies, government involvement, foreign affairs, etc.

How does something like this scale for more than 10 people? 100? 1000?

While I was writing a tool for myself to summarise daily the top N posts from HN, Google Trends, and RSS feed subscriptions I had the same problem.

The quick solution was to use beautiful soup and readability-lxml to try and get the main article contents and then send it to an LLM.

The results are ok when the markup is semantic. Often it is not. Then you have tables, images, weirdly positioned footnotes, etc.

I believe the best way to extract information the way it was intended to be presented is to screenshot the page and send it to a multimodal LLM for “interpretation”. Anyone experimented with that approach?

——

The aspiration goal for the tool is to be the Presidential Daily Brief but for everyone.

Power Nap is another feature rather than a bug https://support.apple.com/en-gb/guide/mac-help/mh40773/mac

Example: https://delta.io/blog/2022-05-18-multi-cluster-writes-to-del...

You’ll love to learn more about Chernoff faces! https://en.wikipedia.org/wiki/Chernoff_face

The page says they hold the trademark to “Amazon Athena”. Does this come into conflict with a project named “AthenaDB”?

This sentence is explained on their homepage:

> it takes 2 RTTs to complete a consensus request from the view of a client. One RTT takes place between the client and the leader server, and the leader server takes another RTT to broadcast the message to the follower servers

> Existing distributed KV stores mostly adopt the Raft consensus protocol, which takes two RTTs to complete a request.

Why 2? Isn’t a single AppendEntries request from the leader enough?

Adam

dlemire, you note that the read ”overflows”. Why can’t you copy just `len` bytes? Does it slow too much because of the branch/more load/store operations?

Except Oppenheimer movie, who got it right.

Seek to 4m46s https://youtu.be/kqYenzBwlkU?si=y87R0XvuWL8_7t6Q&t=286

If my memory is right, the actual scene, and delay, is longer than it is in this clip.

Commit Mono font does something similar and calls it “Smart kerning”. Visit https://commitmono.com/ and click on the “04 Intelligent” tab for details.

In practice this is unusable. Because the width of the letters now depends on the succeeding character, the text jumps as you write it. Super annoying.

Cool for reading. Awful for writing.

Thread-per-core software architectures are doing this https://penberg.org/papers/tpc-ancs19.pdf

Real world examples are scylladb and Redpanda, both built on the seastar framework (C++ https://seastar.io/message-passing/).

And for rust there is glommio https://www.datadoghq.com/blog/engineering/introducing-glomm...

What’s inside:

- Importers: create LifeLog entries from several data sources (already available: Apple Health, Amazon, Amazon Kindle, Spotify, Venmo, Libby, Google Photos, Google Location, Facebook posts)

- Data Visualization: a ReactJS-based visualization frontend of the personal timeline

- Question Answering: a LLM-based QA engine over the personal timeline

“The vision of creating a timeline of your life experiences has been around for decades. We're releasing TimelineBuilder to spur more work towards the vision. It includes a few importers (photos, maps, shopping), episode inference and basic visualization.”

https://twitter.com/AlonHalevy/status/1673384756883636224

This topic was recently brought to attention by David Friedberg in the latest episode of the All-In podcast. https://youtu.be/5cQXjboJwg0?t=6645 (1:50:45)

You can get access to the playground (Jupiter notebook with Mojo kernel): https://docs.modular.com/mojo/get-started.html

My access was approved within one hour.

A “note taking” app after spending years looking for “the right one”. Nothing ground breaking, similar to most other “connected notes” apps but with one small difference: everything is built for my brain workflow rather than the other way around.

https://nvartolomei.com/omniverse/

Maybe, one day, after I’m satisfied with its functionality I’ll make it open(-source). For now, in the interest of keeping friction low, moving fast and breaking things, it’s pretty private.