We created the streams to make it easier to build apps that stream data from an llm to the client, but they’re fundamentally a single-writer, multiple-reader data structure.

The pointer events are stored as json lines in the stream, and then the replay reconstructs shapes from the events as they get played back.

The code lives at https://github.com/instantdb/inkdot

Comments URL: https://news.ycombinator.com/item?id=47414857

Points: 1

# Comments: 0

For example:

  (let [result {:a 1}
        result (assoc result :b 2)]
    ...)

clj-kondo has a :shadowed-var rule, but it will only find cases where you shadow a top-level var (not the case in my example).

https://yt-better-subs.web.app/

I went through quite the hassle to get the app's oauth scopes approved with Google so that it can keep your subscriptions up-to-date as you add or remove YouTube channel subscriptions.

1. Set up logical replication to a new database server. We used https://github.com/2ndQuadrant/pglogical, but maybe you don't need that any more with newer versions of postgres?

2. Flip a feature flag that pauses all database queries and wait for the queue of queries to complete.

3. Wait for the query queue to drain and for replication to catch up.

4. Flip a feature flag that switches the connection from the old db to the new db.

5. Flip the flag to resume queries.

It helped that we were written in OCaml. We had to write our own connection pooling, which meant that we had full control over the query queue. Not sure how you would do it with e.g. Java's Hikari, where the query queue and the connection settings are complected.

We also had no long-running queries, with a default timeout of 30 seconds.

It helped to over-provision servers during the migration, because any requests that came in while the migration was ongoing would have to wait for the migration to complete.

https://g.co/bard/share/7966410c42af

ChatGPT also figured it out, but Bard is much better at displaying information: https://chat.openai.com/share/ba5d5acc-7b40-46e1-ada5-74b4a6...

It's available on GitHub at https://github.com/dwwoelfel/latch

It has a self-hosted UI that reads and writes directly to Google storage and a nodejs client that subscribes to flag changes via Google Pub/Sub.

I made a short video walkthrough to help give a feel for the UI:

https://www.youtube.com/watch?v=GKDekwvPTD4

In the past I've stored feature flags directly in the database, but I wanted something that was more consistent between environments. I also want to be able to use feature flags to trigger db failover during upgrades, so storing them in the db was out.

I've also used two of the main SaaS services at a previous company, which made me wary of lock-in and performance issues.

Google Storage turned out to be well-suited for feature flags. It has a pub/sub integration where changes to a bucket trigger notifications in pub/sub. It has object versioning, which I use to display a flag history. It has an increasing "generation" id for each file, which I use to prevent race conditions and ensure that I'm not making changes based on stale information.

I'm pretty happy with the UI. I built a Relay-compatible GraphQL interface over Google Storage and the UI uses that to read and write from storage using OAuth credentials. If I want to give someone access to the feature flags, I just have to manage their permissions to the bucket.

Comments URL: https://news.ycombinator.com/item?id=36359896

Points: 4

# Comments: 1

1. Make `globalId` part of a "Node" interface that all of the types implement. This will work better with tooling like Relay (used for refetching and caching). It will also let you add a `node` field that can be used to fetch any node in the graph.

2. Make the sort input an enum so that you have `sort: TITLE_DESC` instead of `sort: {by: TITLE, order: DESC}`.

3. Implement the connection spec instead of returning a list of items: https://relay.dev/graphql/connections.htm. This will let you add pagination data to the field and other useful info like totalCount.

4. Spin up a postgraphile instance with the `@graphile-contrib/pg-simplify-inflector` and `postgraphile-plugin-connection-filter` plugins and copy everything they do.

But maybe that's the kind of criticism he was trying to avoid in the first place!

Those affected will gain access to the company’s “generous severance philosophy” and “a talent hub to allow them to opt-in to receive additional support services.” (The details surrounding the severance package are unclear, but some affected workers on Blind alleged they would receive two months worth of base pay; a representative from Coinbase did not provide further comment.)

Comments URL: https://news.ycombinator.com/item?id=30348274

Points: 102

# Comments: 28

I had just started the book when I watched her video, and I counted four separate times where he did mention total energy gain. In one chapter (pg. 142 in my kindle edition), he quotes from a source who lays out the distinction she claims people in fusion research want you to be confused about:

Like First Light Fusion, Tokamak Energy is much more interested in power production than energy breakeven. "Achieving a of one is a scientific goal," Jonathan Carling continues, "but it's nowhere near enough to produce commercial energy, which requires a Q [in the region] of tens." As mentioned previously, Q is the ratio of fusion power out to heating power in. His strong view is that unless other star builders have a credible plan to get to factors of twenty or thirty more power out than they put in, then they're in the science game and not the fusion energy game.

This video by the Improbable Matter channel has a good response to Dr. Hossenfelder's general remarks: https://www.youtube.com/watch?v=KtqC8W0_Ups

It is unethical to leave that up after Fastly pointed out core issues with the benchmarking, like using a free tier that was rate-limited.

It's nice that eastdakota responded here, but he had two weeks since the original tweet thread from Fastly's VP of Eng calling out the problems with their benchmarking. They didn't respond or retract the blog post in those two weeks.

As a cloudflare shareholder (and a fastly shareholder), I want Cloudflare to act ethically and either retract the blog post or issue a correction.